Datasets:
Dataset Viewer
ID
large_stringlengths 10
61
| year
int64 1.96k
2.03k
| title
large_stringlengths 4
560
| abstract
large_stringlengths 0
12.8k
|
|---|---|---|---|
buhnila-etal-2025-chain | 2,025 | Chain-of-MetaWriting: Linguistic and Textual Analysis of How Small Language Models Write Young Students Texts | Large Language Models (LLMs) have been used to generate texts in response to different writing tasks: reports, essays, story telling. However, language models do not have a metarepresentation of the text writing process, nor inherent communication learning needs, comparable to those of young human students. This paper introduces a fine-grained linguistic and textual analysis of multilingual Small Language Models' (SLMs) writing. With our method, Chain-of-MetaWriting, SLMs can imitate some steps of the human writing process, such as planning and evaluation. We mainly focused on short story and essay writing tasks in French for schoolchildren and undergraduate students respectively. Our results show that SLMs encounter difficulties in assisting young students on sensitive topics such as violence in the schoolyard, and they sometimes use words too complex for the target audience. In particular, the output is quite different from the human produced texts in term of text cohesion and coherence regarding temporal connectors, topic progression, reference. |
shi-penn-2025-semantic | 2,025 | Semantic Masking in a Needle-in-a-haystack Test for Evaluating Large Language Model Long-Text Capabilities | In this paper, we introduce the concept of Semantic Masking, where semantically coherent surrounding text (the haystack) interferes with the retrieval and comprehension of specific information (the needle) embedded within it. We propose the Needle-in-a-Haystack-QA Test, an evaluation pipeline that assesses LLMs' long-text capabilities through question answering, explicitly accounting for the Semantic Masking effect. We conduct experiments to demonstrate that Semantic Masking significantly impacts LLM performance more than text length does. By accounting for Semantic Masking, we provide a more accurate assessment of LLMs' true proficiency in utilizing extended contexts, paving the way for future research to develop models that are not only capable of handling longer inputs but are also adept at navigating complex semantic landscapes. |
khallaf-etal-2025-reading | 2,025 | Reading Between the Lines: A dataset and a study on why some texts are tougher than others | Our research aims at better understanding what makes a text difficult to read for specific audiences with intellectual disabilities, more specifically, people who have limitations in cognitive functioning, such as reading and understanding skills, an IQ below 70, and challenges in conceptual domains. We introduce a scheme for the annotation of difficulties which is based on empirical research in psychology as well as on research in translation studies. The paper describes the annotated dataset, primarily derived from the parallel texts (standard English and Easy to Read English translations) made available online. we fine-tuned four different pre-trained transformer models to perform the task of multiclass classification to predict the strategies required for simplification. We also investigate the possibility to interpret the decisions of this language model when it is aimed at predicting the difficulty of sentences in this dataset. |
jourdan-etal-2025-pararev | 2,025 | ParaRev : Building a dataset for Scientific Paragraph Revision annotated with revision instruction | Revision is a crucial step in scientific writing, where authors refine their work to improve clarity, structure, and academic quality. Existing approaches to automated writing assistance often focus on sentence-level revisions, which fail to capture the broader context needed for effective modification. In this paper, we explore the impact of shifting from sentence-level to paragraph-level scope for the task of scientific text revision. The paragraph level definition of the task allows for more meaningful changes, and is guided by detailed revision instructions rather than general ones. To support this task, we introduce ParaRev, the first dataset of revised scientific paragraphs with an evaluation subset manually annotated with revision instructions. Our experiments demonstrate that using detailed instructions significantly improves the quality of automated revisions compared to general approaches, no matter the model or the metric considered. |
maggi-vitaletti-2025-towards | 2,025 | Towards an operative definition of creative writing: a preliminary assessment of creativeness in AI and human texts | Nowadays, AI is present in all our activities. This pervasive presence is perceived as a threat by many category of users that might be substituted by their AI counterpart. While the potential of AI in handling repetitive tasks is clear, the potentials of its creativeness is still misunderstood. We believe that understanding this aspects of AI can transform a threat into an opportunity. This paper is a first attempt to provide a measurable definition of creativeness. We applied our definition to AI and human generated texts, proving the viability of the proposed approach. Our preliminary experiments show that human texts are more creative. |
sato-kobayashi-2025-decoding | 2,025 | Decoding Semantic Representations in the Brain Under Language Stimuli with Large Language Models | Brain decoding technology is paving the way for breakthroughs in the interpretation of neural activity to recreate thoughts, emotions, and movements. Tang et al. (2023) introduced a novel approach that uses language models as generative models for brain decoding based on functional magnetic resonance imaging (fMRI) data. Building on their work, this study explored the use of three additional language models along with the GPT model used in previous research to improve decoding accuracy. Furthermore, we added an evaluation metric using an embedding model, providing higher-level semantic similarity than the BERTScore. By comparing the decoding performance and identifying the factors contributing to good performance, we found that high decoding accuracy does not solely depend on the ability to accurately predict brain activity. Instead, the type of text (e.g., web text, blogs, news articles, and books) that the model tends to generate plays a more significant role in achieving more precise sentence reconstruction. |
lamsiyah-etal-2025-arabicsense | 2,025 | ArabicSense: A Benchmark for Evaluating Commonsense Reasoning in Arabic with Large Language Models | Recent efforts in natural language processing (NLP) commonsense reasoning research have led to the development of numerous new datasets and benchmarks. However, these resources have predominantly been limited to English, leaving a gap in evaluating commonsense reasoning in other languages. In this paper, we introduce the ArabicSense Benchmark, which is designed to thoroughly evaluate the world-knowledge commonsense reasoning abilities of large language models (LLMs) in Arabic. This benchmark includes three main tasks: first, it tests whether a system can distinguish between natural language statements that make sense and those that do not; second, it requires a system to identify the most crucial reason why a nonsensical statement fails to make sense; and third, it involves generating explanations for why statements do not make sense. We evaluate several Arabic BERT-based models and causal LLMs on these tasks. Experimental results demonstrate improvements after fine-tuning on our dataset. For instance, AraBERT v2 achieved an 87{\%} F1 score on the second task, while Gemma and Mistral-7b achieved F1 scores of 95.5{\%} and 94.8{\%}, respectively. For the generation task, LLaMA-3 achieved the best performance with a BERTScore F1 of 77.3{\%}, closely followed by Mistral-7b at 77.1{\%}. All codes and the benchmark will be made publicly available at https://github.com/. |
hamed-etal-2025-lahjawi | 2,025 | Lahjawi: Arabic Cross-Dialect Translator | In this paper, we explore the rich diversity of Arabic dialects by introducing a suite of pioneering models called Lahjawi. The primary model, Lahjawi-D2D, is the first designed for cross-dialect translation among 15 Arabic dialects. Furthermore, we introduce Lahjawi-D2MSA, a model designed to convert any Arabic dialect into Modern Standard Arabic (MSA). Both models are fine-tuned versions of Kuwain-1.5B an in-house built small language model, tailored for Arabic linguistic characteristics. We provide a detailed overview of Lahjawi`s architecture and training methods, along with a comprehensive evaluation of its performance. The results demonstrate Lahjawi`s success in preserving meaning and style, with BLEU scores of 9.62 for dialect-to-MSA and 9.88 for dialect-to- dialect tasks. Additionally, human evaluation reveals an accuracy score of 58{\%} and a fluency score of 78{\%}, underscoring Lahjawi`s robust handling of diverse dialectal nuances. This research sets a foundation for future advancements in Arabic NLP and cross-dialect communication technologies. |
bezancon-etal-2025-lost | 2,025 | Lost in Variation: An Unsupervised Methodology for Mining Lexico-syntactic Patterns in Middle Arabic Texts | While MSA and some dialects of Arabic have been extensively studied in NLP, Middle Arabic is still very much unknown to the field. However, Middle Arabic holds issues that are still not covered: it is characterized by variation since it mixes standard features, colloquial ones, as well as features that belong to neither of the two. Here, we introduce a methodology to identify, extract and rank variations of 13 manually retrieved formulas. Those formulas come from the nine first booklets of S ̄IRAT AL-MALIK AL-Z. ̄AHIR BAYBAR S., a corpus of Damascene popular literature written in Middle Arabic and composed of 53,843 sentences. In total, we ranked 20, sequences according to their similarity with the original formulas on multiple linguistic layers. We noticed that the variations in these formulas occur in a lexical, morphological and graphical level, but in opposition, the semantic and syntactic levels remain strictly invariable. |
alahmari-2025-sadslyc | 2,025 | SADSLyC: A Corpus for Saudi Arabian Multi-dialect Identification through Song Lyrics | This paper presents the Saudi Arabian Dialects Song Lyrics Corpus (SADSLyC), the first dataset featuring song lyrics from the five major Saudi dialects: Najdi (Central Region), Hijazi (Western Region), Shamali (Northern Region), Janoubi (Southern Region), and Shargawi (Eastern Region). The dataset consists of 31,358 sentences, with each sentence representing a self-contained verse in a song, totaling 151,841 words. Additionally, we present a baseline experiment using the SaudiBERT model to classify the fine-grained dialects in the SADSLyC Corpus. The model achieved an overall accuracy of 73{\%} on the test dataset. |
hossain-etal-2025-enhancing | 2,025 | Enhancing Dialectal Arabic Intent Detection through Cross-Dialect Multilingual Input Augmentation | Addressing the challenges of Arabic intent detection amid extensive dialectal variation, this study presents a crossdialtectal, multilingual approach for classifying intents in banking and migration contexts. By augmenting dialectal inputs with Modern Standard Arabic (MSA) and English translations, our method leverages cross-lingual context to improve classification accuracy. We evaluate single-input (dialect-only), dual-input (dialect + MSA), and triple-input (dialect + MSA + English) models, applying language-specific tokenization for each. Results demonstrate that, in the migration dataset, our model achieved an accuracy gain of over 50{\%} on Tunisian dialect, increasing from 43.3{\%} with dialect-only input to 94{\%} with the full multilingual setup. Similarly, in the PAL (Palestinian dialect) dataset, accuracy improved from 87.7{\%} to 93.5{\%} with translation augmentation, reflecting a gain of 5.8 percentage points. These findings underscore the effectiveness of our approach for intent detection across various Arabic dialects. |
khered-etal-2025-dial2msa | 2,025 | Dial2MSA-Verified: A Multi-Dialect Arabic Social Media Dataset for Neural Machine Translation to Modern Standard Arabic | Social media has become an essential focus for Natural Language Processing (NLP) research due to its widespread use and unique linguistic characteristics. Normalising social media content, especially for morphologically rich languages like Arabic, remains a complex task due to limited parallel corpora. Arabic encompasses Modern Standard Arabic (MSA) and various regional dialects, collectively termed Dialectal Arabic (DA), which complicates NLP efforts due to their informal nature and variability. This paper presents Dial2MSA-Verified, an extension of the Dial2MSA dataset that includes verified translations for Gulf, Egyptian, Levantine, and Maghrebi dialects. We evaluate the performance of Seq2Seq models on this dataset, highlighting the effectiveness of state-of-the-art models in translating local Arabic dialects. We also provide insights through error analysis and outline future directions for enhancing Seq2Seq models and dataset development. The Dial2MSA-Verified dataset is publicly available to support further research. |
el-ghawi-2025-web | 2,025 | Web-Based Corpus Compilation of the Emirati Arabic Dialect | This paper displays some initial efforts conducted in the compilation pursuits of Arabic dialectal corpora in the form of raw text, the end purpose of which is to fine-tune existing Arabic large language models (LLM) to better understand and generate text in the Emirati dialect as instructed. The focus of the paper is on the process of compiling corpora from the web, which includes the exploration of possible methods, tools and techniques specific to web search, as well as examples of genres and domains to explore. The results of these efforts and the importance of native speaker contributions to corpus compilation for low-resource languages are also touched upon. |
al-laith-kebdani-2025-evaluating | 2,025 | Evaluating Calibration of Arabic Pre-trained Language Models on Dialectal Text | While pre-trained language models have made significant progress in different classification tasks, little attention has been given to the reliability of their confidence scores. Calibration, how well model confidence aligns with actual accuracy, is essential for real-world applications where decisions rely on probabilistic outputs. This study addresses this gap in Arabic dialect identification by assessing the calibration of eight pre-trained language models, ensuring their predictions are not only accurate but also reliable for practical applications. We analyze two datasets: one with over 1 million text samples and the Nuanced Arabic Dialect Identification dataset(NADI-2023). Using Expected Calibration Error (ECE) as a metric, we reveal substantial variation in model calibration across dialects in both datasets, showing that prediction confidence can vary significantly depending on regional data. This research has implications for improving the reliability of Arabic dialect models in applications like sentiment analysis and social media monitoring. |
aftiss-etal-2025-empirical | 2,025 | Empirical Evaluation of Pre-trained Language Models for Summarizing Moroccan Darija News Articles | Moroccan Dialect (MD), or {\textquotedblleft}Darija,{\textquotedblright} is a primary spoken variant of Arabic in Morocco, yet remains underrepresented in Natural Language Processing (NLP) research, particularly in tasks like summarization. Despite a growing volume of MD textual data online, there is a lack of robust resources and NLP models tailored to handle the unique linguistic challenges posed by MD. In response, we introduce .MA{\_}v2, an expanded version of the GOUD.MA dataset, containing over 50k articles with their titles across 11 categories. This dataset provides a more comprehensive resource for developing summarization models. We evaluate the application of large language models (LLMs) for MD summarization, utilizing both fine-tuning and zero-shot prompting with encoder-decoder and causal LLMs, respectively. Our findings demonstrate that an expanded dataset improves summarization performance and highlights the capabilities of recent LLMs in handling MD text. We open-source our dataset, fine-tuned models, and all experimental code, establishing a foundation for future advancements in MD NLP. We release the code at https://github.com/AzzedineAftiss/Moroccan-Dialect-Summarization. |
chafik-etal-2025-dialect2sql | 2,025 | Dialect2SQL: A Novel Text-to-SQL Dataset for Arabic Dialects with a Focus on Moroccan Darija | The task of converting natural language questions into executable SQL queries, known as text-to-SQL, has gained significant interest in recent years, as it enables non-technical users to interact with relational databases. Many benchmarks, such as SPIDER and WikiSQL, have contributed to the development of new models and the evaluation of their performance. In addition, other datasets, like SEDE and BIRD, have introduced more challenges and complexities to better map real-world scenarios. However, these datasets primarily focus on high-resource languages such as English and Chinese. In this work, we introduce Dialect2SQL, the first large-scale, cross-domain text-to-SQL dataset in an Arabic dialect. It consists of 9,428 NLQ-SQL pairs across 69 databases in various domains. Along with SQL-related challenges such as long schemas, dirty values, and complex queries, our dataset also incorporates the complexities of the Moroccan dialect, which is known for its diverse source lan-guages, numerous borrowed words, and unique expressions. This demonstrates that our dataset will be a valuable contribution to both the text-to-SQL community and the development of resources for low-resource languages. |
bouomar-abbas-2025-arasim | 2,025 | AraSim: Optimizing Arabic Dialect Translation in Children`s Literature with LLMs and Similarity Scores | The goal of the paper is to address the linguistic gap faced by young Egyptian Arabic speakers through translating children stories from Modern Standard Arabic to the Egyptian Cairo dialect. Claude is used for initial translation, and a fine-tuned AraT5 model is used for backtranslation. The translation quality is assessed using semantic similarity and BLUE scores to compare the original texts and the translations. The resulting corpus contains 130 stories which were revised by native Egyptian speakers who are professional translators. The strengths of this paper are multiple: working on a less-resourced variety, addressing an important social issue, creating a dataset with potential real-life applications, and ensuring the quality of the produced dataset through human validation. |
haj-ahmed-etal-2025-navigating | 2,025 | Navigating Dialectal Bias and Ethical Complexities in Levantine Arabic Hate Speech Detection | Social media platforms have become central to global communication, yet they also facilitate the spread of hate speech. For underrepresented dialects like Levantine Arabic, detecting hate speech presents unique cultural, ethical, and linguistic challenges. This paper explores the complex sociopolitical and linguistic landscape of Levantine Arabic and critically examines the limitations of current datasets used in hate speech detection. We highlight the scarcity of publicly available, diverse datasets and analyze the consequences of dialectal bias within existing resources. By emphasizing the need for culturally and contextually informed natural language processing (NLP) tools, we advocate for a more nuanced and inclusive approach to hate speech detection in the Arab world. |
scherrer-etal-2025-findings | 2,025 | Findings of the VarDial Evaluation Campaign 2025: The NorSID Shared Task on Norwegian Slot, Intent and Dialect Identification | The VarDial Evaluation Campaign 2025 was organized as part of the twelfth workshop on Natural Language Processing for Similar Languages, Varieties and Dialects (VarDial), colocated with COLING 2025. It consisted of one shared task with three subtasks: intent detection, slot filling and dialect identification for Norwegian dialects. This report presents the results of this shared task. Four participating teams have submitted systems with very high performance ({\ensuremath{>}} 97{\%} accuracy) for intent detection, whereas slot detection and dialect identification showed to be much more challenging, with respectively span-F1 scores up to 89{\%}, and weighted dialect F1 scores of 84{\%}. |
alves-2025-information | 2,025 | Information Theory and Linguistic Variation: A Study of Brazilian and European Portuguese | We present a general analysis of the lexical and grammatical differences between Brazilian and European Portuguese by applying entropy measures, including Kullback-Leibler divergence and word order entropy, across various linguistic levels. Using a parallel corpus of BP and EP sentences translated from English, we quantified these differences and identified characteristic phenomena underlying the divergences between the two varieties. The highest divergence was observed at the lexical level due to word pairs unique to each variety but also related to grammatical distinctions. Furthermore, the analysis of parts-of-speech (POS), dependency relations, and POS tri-grams provided information concerning distinctive grammatical constructions. Finally, the word order entropy analysis revealed that while most of the syntactic features analysed showed similar patterns across BP and EP, specific word order preferences were still apparent. |
ng-markov-2025-leveraging | 2,025 | Leveraging Open-Source Large Language Models for Native Language Identification | Native Language Identification (NLI) {--} the task of identifying the native language (L1) of a person based on their writing in the second language (L2) {--} has applications in forensics, marketing, and second language acquisition. Historically, conventional machine learning approaches that heavily rely on extensive feature engineering have outperformed transformer-based language models on this task. Recently, closed-source generative large language models (LLMs), e.g., GPT-4, have demonstrated remarkable performance on NLI in a zero-shot setting, including promising results in open-set classification. However, closed-source LLMs have many disadvantages, such as high costs and undisclosed nature of training data. This study explores the potential of using open-source LLMs for NLI. Our results indicate that open-source LLMs do not reach the accuracy levels of closed-source LLMs when used out-of-the-box. However, when fine-tuned on labeled training data, open-source LLMs can achieve performance comparable to that of commercial LLMs. |
torgbi-etal-2025-adapting | 2,025 | Adapting Whisper for Regional Dialects: Enhancing Public Services for Vulnerable Populations in the United Kingdom | We collect novel data in the public service domain to evaluate the capability of the state-of-the-art automatic speech recognition (ASR) models in capturing regional differences in accents in the United Kingdom (UK), specifically focusing on two accents from Scotland with distinct dialects. This study addresses real-world problems where biased ASR models can lead to miscommunication in public services, disadvantaging individuals with regional accents particularly those in vulnerable populations. We first examine the out-of-the-box performance of the Whisper large-v3 model on a baseline dataset and our data. We then explore the impact of fine-tuning Whisper on the performance in the two UK regions and investigate the effectiveness of existing model evaluation techniques for our real-world application through manual inspection of model errors. We observe that the Whisper model has a higher word error rate (WER) on our test datasets compared to the baseline data and fine-tuning on a given data improves performance on the test dataset with the same domain and accent. The fine-tuned models also appear to show improved performance when applied to the test data outside of the region it was trained on suggesting that fine-tuned models may be transferable within parts of the UK. Our manual analysis of model outputs reveals the benefits and drawbacks of using WER as an evaluation metric and fine-tuning to adapt to regional dialects. |
alam-anastasopoulos-2025-large | 2,025 | Large Language Models as a Normalizer for Transliteration and Dialectal Translation | NLP models trained on standardized language data often struggle with variations. We assess various Large Language Models (LLMs) for transliteration and dialectal normalization. Tuning open-source LLMs with as little as 10,000 parallel examples using LoRA can achieve results comparable to or better than closed-source LLMs. We perform dialectal normalization experiments for twelve South Asian languages and dialectal translation experiments for six language continua worldwide. The dialectal normalization task can also be a preliminary step for the downstream dialectal translation task. Among the six languages used in dialectal translation, our approach enables Italian and Swiss German to surpass the baseline model by 21.5 and 25.8 BLEU points, respectively. |
faisal-anastasopoulos-2025-testing | 2,025 | Testing the Boundaries of LLMs: Dialectal and Language-Variety Tasks | This study evaluates the performance of large language models (LLMs) on benchmark datasets designed for dialect-specific NLP tasks. Dialectal NLP is a low-resource field, yet it is crucial for evaluating the robustness of language models against linguistic diversity. This work is the first to systematically compare state-of-the-art instruction-tuned LLMs{---}both open-weight multilingual and closed-weight generative models{---}with encoder-based models that rely on supervised task-specific fine-tuning for dialectal tasks. We conduct extensive empirical analyses to provide insights into the current LLM landscape for dialect-focused tasks. Our findings indicate that certain tasks, such as dialect identification, are challenging for LLMs to replicate effectively due to the complexity of multi-class setups and the suitability of these tasks for supervised fine-tuning. Additionally, the structure of task labels{---}whether categorical or continuous scoring{---}significantly affects model performance. While LLMs excel in tasks like machine reading comprehension, their instruction-following ability declines in simpler tasks like POS tagging when task instructions are inherently complex. Overall, subtle variations in prompt design can greatly impact performance, underscoring the need for careful prompt engineering in dialectal evaluations. |
plum-etal-2025-text | 2,025 | Text Generation Models for Luxembourgish with Limited Data: A Balanced Multilingual Strategy | This paper addresses the challenges in developing language models for less-represented languages, with a focus on Luxembourgish. Despite its active development, Luxembourgish faces a digital data scarcity, exacerbated by Luxembourg`s multilingual context. We propose a novel text generation model based on the T5 architecture, combining limited Luxembourgish data with equal amounts, in terms of size and type, of German and French data. We hypothesise that a model trained on Luxembourgish, German, and French will improve the model`s cross-lingual transfer learning capabilities and outperform monolingual and large multilingual models. To verify this, the study at hand explores whether multilingual or monolingual training is more beneficial for Luxembourgish language generation. For the evaluation, we introduce LuxGen, a text generation benchmark that is the first of its kind for Luxembourgish. |
lendvai-etal-2025-retrieval | 2,025 | Retrieval of Parallelizable Texts Across Church Slavic Variants | The goal of our study is to identify parallelizable texts for Church Slavic, across chronological and regional variants. Next to using a benchmark text, we utilize a recently digitized, large text collection and compile new resources for the retrieval of similar texts: a ground truth dataset holding a small amount of manually aligned sentences in Old Church Slavic and in Old East Slavic, and a large unaligned dataset that has a subset of ground truth (GT) quality texts but contains noise from handwritten text recognition (HTR) for the majority of the collection. We discuss preprocessing challenges in the data and the impact of sentence segmentation on retrieval performance. We evaluate sentence snippets mapped across these two diachronic variants of Church Slavic, expressed by mean reciprocal rank, using embedding representations from large language models (LLMs) as well as classical string similarity based approaches combined with k-nearest neighbor (kNN) search. Experimental results indicate that in the current setup (short text snippets, off-the-shelf multilingual embeddings), classical string similarity based retrieval can still outperform embedding based retrieval. |
lutgen-etal-2025-neural | 2,025 | Neural Text Normalization for Luxembourgish Using Real-Life Variation Data | Orthographic variation is very common in Luxembourgish texts due to the absence of a fully-fledged standard variety. Additionally, developing NLP tools for Luxembourgish is a difficult task given the lack of annotated and parallel data, which is exacerbated by ongoing standardization. In this paper, we propose the first sequence-to-sequence normalization models using the ByT5 and mT5 architectures with training data obtained from word-level real-life variation data. We perform a fine-grained, linguistically-motivated evaluation to test byte-based, word-based and pipeline-based models for their strengths and weaknesses in text normalization. We show that our sequence model using real-life variation data is an effective approach for tailor-made normalization in Luxembourgish. |
kruckl-etal-2025-improving | 2,025 | Improving Dialectal Slot and Intent Detection with Auxiliary Tasks: A Multi-Dialectal Bavarian Case Study | Reliable slot and intent detection (SID) is crucial in natural language understanding for applications like digital assistants. Encoder-only transformer models fine-tuned on high-resource languages generally perform well on SID. However, they struggle with dialectal data, where no standardized form exists and training data is scarce and costly to produce. We explore zero-shot transfer learning for SID, focusing on multiple Bavarian dialects, for which we release a new dataset for the Munich dialect. We evaluate models trained on auxiliary tasks in Bavarian, and compare joint multi-task learning with intermediate-task training. We also compare three types of auxiliary tasks: token-level syntactic tasks, named entity recognition (NER), and language modelling. We find that the included auxiliary tasks have a more positive effect on slot filling than intent classification (with NER having the most positive effect), and that intermediate-task training yields more consistent performance gains. Our best-performing approach improves intent classification performance on Bavarian dialects by 5.1 and slot filling F1 by 8.4 percentage points. |
coats-etal-2025-regional | 2,025 | Regional Distribution of the /el/-/\ael/ Merger in Australian English | Prelateral merger of /e/ and /{\ae}/ is a salient acoustic feature of speech from Melbourne and the state of Victoria in Australia, but little is known about its presence in other parts of the country. In this study, automated methods of data collection, forced alignment, and formant extraction are used to analyze the regional distribution of the vowel merger within all of Australia, in 4.3 million vowel tokens from naturalistic speech in 252 locations. The extent of the merger is quantified using the difference in Bhattacharyya`s distance scores based on phonetic context, and the regional distribution is assessed using spatial autocorrelation. The principal findings are that the merger is most prominent in Victoria and least prominent in Sydney and New South Wales. We also find preliminary indications that it may be present in other parts of the country. |
khalifa-etal-2025-learning | 2,025 | Learning Cross-Dialectal Morphophonology with Syllable Structure Constraints | We investigate learning surface forms from underlying morphological forms for low-resource language varieties. We concentrate on learning explicit rules with the aid of learned syllable structure constraints, which outperforms neural methods on this small data task and provides interpretable output. Evaluating across one relatively high-resource and two related low-resource Arabic dialects, we find that a model trained only on the high-resource dialect achieves decent performance on the low-resource dialects, useful when no low-resource training data is available. The best results are obtained when our system is trained only on the low-resource dialect data without augmentation from the related higher-resource dialect. We discuss the impact of syllable structure constraints and the strengths and weaknesses of data augmentation and transfer learning from a related dialect. |
lopetegui-etal-2025-common | 2,025 | Common Ground, Diverse Roots: The Difficulty of Classifying Common Examples in Spanish Varieties | Variations in languages across geographic regions or cultures are crucial to address to avoid biases in NLP systems designed for culturally sensitive tasks, such as hate speech detection or dialog with conversational agents. In languages such as Spanish, where varieties can significantly overlap, many examples can be valid across them, which we refer to as common examples. Ignoring these examples may cause misclassifications, reducing model accuracy and fairness. Therefore, accounting for these common examples is essential to improve the robustness and representativeness of NLP systems trained on such data. In this work, we address this problem in the context of Spanish varieties. We use training dynamics to automatically detect common examples or errors in existing Spanish datasets. We demonstrate the efficacy of using predicted label confidence for our Datamaps (CITATION) implementation for the identification of hard-to-classify examples, especially common examples, enhancing model performance in variety identification tasks. Additionally, we introduce a Cuban Spanish Variety Identification dataset with common examples annotations developed to facilitate more accurate detection of Cuban and Caribbean Spanish varieties. To our knowledge, this is the first dataset focused on identifying the Cuban, or any other Caribbean, Spanish variety. |
blaschke-etal-2025-add | 2,025 | Add Noise, Tasks, or Layers? MaiNLP at the VarDial 2025 Shared Task on Norwegian Dialectal Slot and Intent Detection | Slot and intent detection (SID) is a classic natural language understanding task. Despite this, research has only more recently begun focusing on SID for dialectal and colloquial varieties. Many approaches for low-resource scenarios have not yet been applied to dialectal SID data, or compared to each other on the same datasets. We participate in the VarDial 2025 shared task on slot and intent detection in Norwegian varieties, and compare multiple set-ups: varying the training data (English, Norwegian, or dialectal Norwegian), injecting character-level noise, training on auxiliary tasks, and applying Layer Swapping, a technique in which layers of models fine-tuned on different datasets are assembled into a model. We find noise injection to be beneficial while the effects of auxiliary tasks are mixed. Though some experimentation was required to successfully assemble a model from layers, it worked surprisingly well; a combination of models trained on English and small amounts of dialectal data produced the most robust slot predictions. Our best models achieve 97.6{\%} intent accuracy and 85.6{\%} slot F1 in the shared task. |
midtgaard-etal-2025-ltg | 2,025 | LTG at VarDial 2025 NorSID: More and Better Training Data for Slot and Intent Detection | This paper describes the LTG submission to the VarDial 2025 shared task, where we participate in the Norwegian slot and intent detection subtasks. The shared task focuses on Norwegian dialects, which present challenges due to their low-resource nature and variation. We test a variety of neural models and training data configurations, with the focus on improving and extending the available Norwegian training data. This includes automatically re-aligning slot spans in Norwegian Bokm{\r{a}}l, as well as re-translating the original English training data into both Bokm{\r{a}}l and Nynorsk. {\%} to address dialectal diversity. We also re-annotate an external Norwegian dataset to augment the training data. Our best models achieve first place in both subtasks, achieving an span F1 score of 0.893 for slot filling and an accuracy of 0.980 for intent detection. Our results indicate that while translation quality is less critical, improving the slot labels has a notable impact on slot performance. Moreover, adding more standard Norwegian data improves performance, but incorporating even small amounts of dialectal data leads to greater gains. |
bengoetxea-etal-2025-hitz | 2,025 | HiTZ at VarDial 2025 NorSID: Overcoming Data Scarcity with Language Transfer and Automatic Data Annotation | In this paper we present our submission for the NorSID Shared Task as part of the 2025 VarDial Workshop, consisting of three tasks: Intent Detection, Slot Filling and Dialect Identification, evaluated using data in different dialects of the Norwegian language. For Intent Detection and Slot Filling, we have fine-tuned a multitask model in a cross-lingual setting, to leverage the xSID dataset available in 17 languages. In the case of Dialect Identification, our final submission consists of a model fine-tuned on the provided development set, which has obtained the highest scores within our experiments. Our final results on the test set show that our models do not drop in performance compared to the development set, likely due to the domain-specificity of the dataset and the similar distribution of both subsets. Finally, we also report an in-depth analysis of the provided datasets and their artifacts, as well as other sets of experiments that have been carried out but did not yield the best results. Additionally, we present an analysis on the reasons why some methods have been more successful than others; mainly the impact of the combination of languages and domain-specificity of the training data on the results. |
ibrahim-2025-cufe | 2,025 | CUFE@VarDial 2025 NorSID: Multilingual BERT for Norwegian Dialect Identification and Intent Detection | Dialect identification is crucial in enhancing various tasks, including sentiment analysis, as a speaker`s geographical origin can significantly affect their perspective on a topic, also, intent detection has gained significant traction in natural language processing due to its applications in various domains, including virtual assistants, customer service automation, and information retrieval systems. This work describes a system developed for VarDial 2025: Norwegian slot and intent detection and dialect identification shared task (Scherrer et al., 2025), a challenge designed to address the dialect recognition and intent detection problems for a low-resource language like Norwegian. More specifically, this work investigates the performance of different BERT models in solving this problem. Finally, the output of the multilingual version of the BERT model was submitted to this shared task, the developed system achieved a weighted F1 score of 79.64 for dialect identification and an accuracy of 94.38 for intent detection. |
zevallos-etal-2025-first | 2,025 | The First Multilingual Model For The Detection of Suicide Texts | Suicidal ideation is a serious health problem affecting millions of people worldwide. Social networks provide information about these mental health problems through users' emotional expressions. We propose a multilingual model leveraging transformer architectures like mBERT, XML-R, and mT5 to detect suicidal text across posts in six languages - Spanish, English, German, Catalan, Portuguese and Italian. A Spanish suicide ideation tweet dataset was translated into five other languages using SeamlessM4T. Each model was fine-tuned on this multilingual data and evaluated across classification metrics. Results showed mT5 achieving the best performance overall with F1 scores above 85{\%}, highlighting capabilities for cross-lingual transfer learning. The English and Spanish translations also displayed high quality based on perplexity. Our exploration underscores the importance of considering linguistic diversity in developing automated multilingual tools to identify suicidal risk. Limitations exist around semantic fidelity in translations and ethical implications which provide guidance for future human-in-the-loop evaluations. |
lin-etal-2025-crossin | 2,025 | CrossIn: An Efficient Instruction Tuning Approach for Cross-Lingual Knowledge Alignment | Multilingual proficiency presents a significant challenge for large language models (LLMs). English-centric models are usually suboptimal in other languages, particularly those that are linguistically distant from English. This performance discrepancy mainly stems from the imbalanced distribution of training data across languages during pre-training and instruction tuning stages. To address this problem, we propose a novel approach called CrossIn, which utilizes a mixed composition of cross-lingual instruction tuning data. Our method leverages the compressed representation shared by various languages to efficiently enhance the model`s task-solving capabilities and multilingual proficiency within a single process. In addition, we introduce a multi-task and multi-faceted benchmark to evaluate the effectiveness of CrossIn. Experimental results demonstrate that our method substantially improves performance across tasks and languages, and we provide extensive insights into the impact of cross-lingual data volume and the integration of translation data on enhancing multilingual consistency and accuracy. |
srirag-etal-2025-evaluating | 2,025 | Evaluating Dialect Robustness of Language Models via Conversation Understanding | With an evergrowing number of LLMs reporting superlative performance for English, their ability to perform equitably for different dialects of English (i.e., dialect robustness) needs to be ascertained. Specifically, we use English language (US English or Indian English) conversations between humans who play the word-guessing game of {\textquoteleft}taboo{\textquoteleft}. We formulate two evaluative tasks: target word prediction (TWP) (i.e., predict the masked target word in a conversation) and target word selection (TWS) (i.e., select the most likely masked target word in a conversation, from among a set of candidate words). Extending MD3, an existing dialectic dataset of taboo-playing conversations, we introduce M-MD3, a target-word-masked version of MD3 with the en-US and en-IN subsets. We create two subsets: en-MV (where en-US is transformed to include dialectal information) and en-TR (where dialectal information is removed from en-IN). We evaluate three multilingual LLMs{--}one open source (Llama3) and two closed-source (GPT-4/3.5). LLMs perform significantly better for US English than Indian English for both TWP and TWS tasks, for all settings, exhibiting marginalisation against the Indian dialect of English. While GPT-based models perform the best, the comparatively smaller models work more equitably after fine-tuning. Our evaluation methodology exhibits a novel and reproducible way to examine attributes of language models using pre-existing dialogue datasets with language varieties. Dialect being an artifact of one`s culture, this paper demonstrates the gap in the performance of multilingual LLMs for communities that do not use a mainstream dialect. |
tashu-etal-2025-cross | 2,025 | Cross-Lingual Document Recommendations with Transformer-Based Representations: Evaluating Multilingual Models and Mapping Techniques | Recommendation systems, for documents, have become tools for finding relevant content on the Web. However, these systems have limitations when it comes to recommending documents in languages different from the query language, which means they might overlook resources in non-native languages. This research focuses on representing documents across languages by using Transformer Leveraged Document Representations (TLDRs) that are mapped to a cross-lingual domain. Four multilingual pre-trained transformer models (mBERT, mT5 XLM RoBERTa, ErnieM) were evaluated using three mapping methods across 20 language pairs representing combinations of five selected languages of the European Union. Metrics like Mate Retrieval Rate and Reciprocal Rank were used to measure the effectiveness of mapped TLDRs compared to non-mapped ones. The results highlight the power of cross-lingual representations achieved through pre-trained transformers and mapping approaches suggesting a promising direction for expanding beyond language connections, between two specific languages. |
nozaki-etal-2025-vrcp | 2,025 | VRCP: Vocabulary Replacement Continued Pretraining for Efficient Multilingual Language Models | Building large language models (LLMs) for non-English languages involves leveraging extensively trained English models through continued pre-training on the target language corpora. This approach harnesses the rich semantic knowledge embedded in English models, allowing superior performance compared to training from scratch. However, tokenizers not optimized for the target language may make inefficiencies in training. We propose Vocabulary Replacement Continued Pretraining (VRCP), a method that optimizes the tokenizer for the target language by replacing unique (solely available) vocabulary from the source tokenizer while maintaining the overall vocabulary size. This approach preserves the semantic knowledge of the source model while enhancing token efficiency and performance for the target language. We evaluated VRCP using the Llama-2 model on Japanese and Chinese corpora. The results show that VRCP matches the performance of vocabulary expansion methods on benchmarks and achieves superior performance in summarization tasks. Additionally, VRCP provides an optimized tokenizer that balances token efficiency, task performance, and GPU memory footprint, making it particularly suitable for resource-constrained environments. |
bernardo-estuar-2025-bai | 2,025 | bAI-bAI: A Context-Aware Transliteration System for Baybayin Scripts | Baybayin, a pre-colonial writing system from the Philippines, has seen a resurgence in recent years. Research in computational linguistics has shown an increasing interest in Baybayin OCR, which focuses on the recognition and classification of script characters. However, existing studies face challenges with ambiguous Baybayin words that have multiple possible transliterations. This study introduces a disambiguation technique that employs word embeddings (WE) for contextual analysis and uses part-of-speech (POS) tagging as an initial filtering step. This approach is compared with an LLM method that prompts GPT-4o mini to determine the most appropriate transliteration given a sentence input. The proposed disambiguation process is integrated into existing Baybayin OCR systems to develop bAI-bAI, a context-aware Baybayin transliteration system capable of handling ambiguous words. Results show that incorporating POS as a filter does not significantly affect performance. The WE-Only method yields an accuracy of 77.46{\%} and takes 5.35ms to process one sample while leveraging GPT-4o mini peaks at a higher accuracy of 90.52{\%} but with a much longer runtime of 3280ms per sample. These findings present an opportunity to further explore and improve NLP approaches in disambiguation methods. |
wongso-etal-2025-nusabert | 2,025 | NusaBERT: Teaching IndoBERT to be Multilingual and Multicultural | We present NusaBERT, a multilingual model built on IndoBERT and tailored for Indonesia`s diverse languages. By expanding vocabulary and pre-training on a regional corpus, NusaBERT achieves state-of-the-art performance on Indonesian NLU benchmarks, enhancing IndoBERT`s multilingual capability. This study also addresses NusaBERT`s limitations and encourages further research on Indonesia`s underrepresented languages. |
boonsarngsuk-etal-2025-evaluating | 2,025 | Evaluating Sampling Strategies for Similarity-Based Short Answer Scoring: a Case Study in Thailand | Automatic short answer scoring is a task whose aim is to help grade written works by learners of some subject matter. In niche subject domains with small examples, existing methods primarily utilized similarity-based scoring, relying on predefined reference answers to grade each student`s answer based on the similarity to the reference. However, these reference answers are often generated from a randomly selected set of graded student answer, which may fail to represent the full range of scoring variations. We propose a semi-automatic scoring framework that enhances the selective sampling strategy for defining the reference answers through a K-center-based and a K-means-based sampling method. Our results demonstrate that our framework outperforms previous similarity-based scoring methods on a dataset with Thai and English. Moreover, it achieves competitive performance compared to human reference performance and LLMs. |
artkaew-2025-thai | 2,025 | Thai Winograd Schemas: A Benchmark for Thai Commonsense Reasoning | Commonsense reasoning is one of the important aspects of natural language understanding, with several benchmarks developed to evaluate it. However, only a few of these benchmarks are available in languages other than English. Developing parallel benchmarks facilitates cross-lingual evaluation, enabling a better understanding of different languages. This research introduces a collection of Winograd Schemas in Thai, a novel dataset designed to evaluate commonsense reasoning capabilities in the context of the Thai language. Through a methodology involving native speakers, professional translators, and thorough validation, the schemas aim to closely reflect Thai language nuances, idioms, and cultural references while maintaining ambiguity and commonsense challenges. We evaluate the performance of popular large language models on this benchmark, revealing their strengths, limitations, and providing insights into the current state-of-the-art. Results indicate that while models like GPT-4 and Claude-3-Opus achieve high accuracy in English, their performance significantly drops in Thai, highlighting the need for further advancements in multilingual commonsense reasoning. |
hakim-etal-2025-anak | 2,025 | Anak Baik: A Low-Cost Approach to Curate Indonesian Ethical and Unethical Instructions | This study explores the ethical challenges faced by Indonesian Large Language Models (LLMs), particularly focusing on their ability to distinguish between ethical and unethical instructions. As LLMs become increasingly integrated into sensitive applications, ensuring their ethical operation is crucial. A key contribution of this study is the introduction of the Anak Baik dataset, a resource designed to enhance the ethical reasoning capabilities of Indonesian LLMs. The phrase {\textquotedblleft}Anak Baik{\textquotedblright}, meaning {\textquotedblleft}Good Boy{\textquotedblright}, symbolizes the ideal of ethical behavior, as a well-behaved child refrains from engaging in harmful actions. The dataset comprises instruction-response pairs in Indonesian, crafted for Supervised Fine-Tuning (SFT) tasks. It includes examples of both ethical and unethical responses to guide models in learning to generate responses that uphold moral standards. Leveraging Low-Rank Adaptation (LoRA) on models such as Komodo and Cendol shows a significant improvement in ethical decision-making processes. This enhanced performance is quantitatively validated through substantial increases in BLEU and ROUGE scores, indicating a stronger alignment with socially responsible behavior. |
abdjul-etal-2025-indonesian | 2,025 | Indonesian Speech Content De-Identification in Low Resource Transcripts | Advancements in technology and the increased use of digital data threaten individual privacy, especially in speech containing Personally Identifiable Information (PII). Therefore, systems that can remove or process privacy-sensitive data in speech are needed, particularly for low-resource transcripts. These transcripts are minimally annotated or labeled automatically, which is less precise than human annotation. However, using them can simplify the development of de-identification systems in any language. In this study, we develop and evaluate an efficient speech de-identification system. We create an Indonesian speech dataset containing sensitive private information and design a system with three main components: speech recognition, information extraction, and masking. To enhance performance in low-resource settings, we incorporate transcription data in training, use data augmentation, and apply weakly supervised learning. Our results show that our techniques significantly improve privacy detection performance, with approximately 29{\%} increase in F1 score, 20{\%} in precision, and 30{\%} in recall with minimally labeled data. |
kamajaya-moeljadi-2025-indomorph | 2,025 | IndoMorph: a Morphology Engine for Indonesian | Indonesian is an agglutinative language and rich in morphology. Although it has more than 250 million speakers, it is a low resource language in NLP field. Many Indonesian NLP resources are scattered, undocumented, and not publicly available. In this paper we address the issue of analyzing morphology as well as generating Indonesian words. We introduce IndoMorph, a morphology analyzer and word generator for Indonesian. In an agglutinative language, morphology deconstruction can be crucial to understand the structure and meaning of words. IndoMorph can be useful for language modeling and testing certain analyses. In addition, it can be employed to make a new Indonesian subword representation resource such as Indonesian morphology dictionary (IMD), used as a language education tool, or embedded in various applications such as text analysis applications. We hope that IndoMorph can be employed not only in the Indonesian NLP research development, but also in the NLP research of any agglutinative languages. |
purwarianti-etal-2025-nusadialogue | 2,025 | NusaDialogue: Dialogue Summarization and Generation for Underrepresented and Extremely Low-Resource Languages | Developing dialogue summarization for extremely low-resource languages is a challenging task. We introduce NusaDialogue, a dialogue summarization dataset for three underrepresented languages in the Malayo-Polynesian language family: Minangkabau, Balinese, and Buginese. NusaDialogue covers 17 topics and 185 subtopics, with annotations provided by 73 native speakers. Additionally, we conducted experiments using fine-tuning on a specifically designed medium-sized language model for Indonesian, as well as zero- and few-shot learning on various multilingual large language models (LLMs). The results indicate that, for extremely low-resource languages such as Minangkabau, Balinese, and Buginese, the fine-tuning approach yields significantly higher performance compared to zero- and few-shot prompting, even when applied to LLMs with considerably larger parameter sizes. |
gokhan-etal-2025-shared | 2,025 | Shared Task RIRAG-2025: Regulatory Information Retrieval and Answer Generation | This paper provides an overview of the Shared Task RIRAG-2025, which focused on advancing the field of Regulatory Information Retrieval and Answer Generation (RIRAG). The task was designed to evaluate methods for answering regulatory questions using the ObliQA dataset. This paper summarizes the shared task, participants' methods, and the results achieved by various teams. |
chikati-etal-2025-challenges | 2,025 | Challenges in Technical Regulatory Text Variation Detection | We present a preliminary study on the feasibility of using current natural language processing techniques to detect variations between the construction codes of different jurisdictions. We formulate the task as a sentence alignment problem and evaluate various sentence representation models for their performance in this task. Our results show that task-specific trained embeddings perform marginally better than other models, but the overall accuracy remains a challenge. We also show that domain-specific fine-tuning hurts the task performance. The results highlight the challenges of developing NLP applications for technical regulatory texts. |
lotfi-etal-2025-bilingual | 2,025 | Bilingual BSARD: Extending Statutory Article Retrieval to Dutch | Statutory article retrieval plays a crucial role in making legal information more accessible to both laypeople and legal professionals. Multilingual countries like Belgium present unique challenges for retrieval models due to the need for handling legal issues in multiple languages. Building on the Belgian Statutory Article Retrieval Dataset (BSARD) in French, we introduce the bilingual version of this dataset, bBSARD. The dataset contains parallel Belgian statutory articles in both French and Dutch, along with legal questions from BSARD and their Dutch translation. Using bBSARD, we conduct extensive benchmarking of retrieval models available for Dutch and French. Our benchmarking setup includes lexical models, zero-shot dense models, and fine-tuned small foundation models. Our experiments show that BM25 remains a competitive baseline compared to many zero-shot dense models in both languages. We also observe that while proprietary models outperform open alternatives in the zero-shot setting, they can be matched or surpassed by fine-tuning small language-specific models. Our dataset and evaluation code are publicly available. |
vanapalli-etal-2025-unifying | 2,025 | Unifying Large Language Models and Knowledge Graphs for efficient Regulatory Information Retrieval and Answer Generation | In a rapidly changing socio-economic land-scape, regulatory documents play a pivotal role in shaping responses to emerging challenges. An efficient regulatory document monitoring system is crucial for addressing the complexi ties of a dynamically evolving world, enabling prompt crisis response, simplifying compliance, and empowering data-driven decision-making. In this work, we present a novel comprehensive analytical framework, PolicyInsight, which is based on a specialized regulatory data model and state-of-the-art NLP techniques of Large Language Models (LLMs) and Knowledge Graphs to derive timely insights, facilitating data-driven decision-making and fostering a more transparent and informed governance ecosystem for regulators, businesses, and citizens. |
rayo-mosquera-etal-2025-hybrid | 2,025 | A Hybrid Approach to Information Retrieval and Answer Generation for Regulatory Texts | |
purbey-etal-2025-1-800-shared | 2,025 | 1-800-SHARED-TASKS at RegNLP: Lexical Reranking of Semantic Retrieval (LeSeR) for Regulatory Question Answering | This paper presents the system description of our entry for the COLING 2025 RegNLP RIRAG (Regulatory Information Retrieval and Answer Generation) challenge, focusing on leveraging advanced information retrieval and answer generation techniques in regulatory domains. We experimented with a combination of embedding models, including Stella, BGE, CDE, and Mpnet, and leveraged fine-tuning and reranking for retrieving relevant documents in top ranks. We utilized a novel approach, LeSeR, which achieved competitive results with a recall@10 of 0.8201 and map@10 of 0.6655 for retrievals. This work highlights the transformative potential of natural language processing techniques in regulatory applications, offering insights into their capabilities for implementing a retrieval augmented generation system while identifying areas for future improvement in robustness and domain adaptation. |
malviya-etal-2025-mst | 2,025 | MST-R: Multi-Stage Tuning for Retrieval Systems and Metric Evaluation | Regulatory documents are rich in nuanced terminology and specialized semantics. FRAG systems: Frozen retrieval-augmented generators utilizing pre-trained (or, frozen) components face consequent challenges with both retriever and answering performance. We present a system that adapts the retriever performance to the target domain using a multi-stage tuning (MST) strategy. Our retrieval approach, called MST-R (a) first fine-tunes encoders used in vector stores using hard negative mining, (b) then uses a hybrid retriever, combining sparse and dense retrievers using reciprocal rank fusion, and then (c) adapts the cross-attention encoder by fine-tuning only the top-k retrieved results. We benchmark the system performance on the dataset released for the RIRAG challenge (as part of the RegNLP workshop at COLING 2025). We achieve significant performance gains obtaining a top rank on the RegNLP challenge leaderboard. We also show that a trivial answering approach *games* the RePASs metric outscoring all baselines and a pre-trained Llama model. Analyzing this anomaly, we present important takeaways for future research. We also release our [code base](https://github.com/Indic-aiDias/MST-R) |
chasandras-etal-2025-aueb | 2,025 | AUEB-Archimedes at RIRAG-2025: Is Obligation concatenation really all you need? | This paper presents the systems we developed for RIRAG-2025, a shared task that requires answering regulatory questions by retrieving relevant passages. The generated answers are evaluated using RePASs, a reference-free and model-based metric. Our systems use a combination of three retrieval models and a reranker. We show that by exploiting a neural component of RePASs that extracts important sentences ({\textquoteleft}obligations') from the retrieved passages, we achieve a dubiously high score (0.947), even though the answers are directly extracted from the retrieved passages and are not actually generated answers. We then show that by selecting the answer with the best RePASs among a few generated alternatives and then iteratively refining this answer by reducing contradictions and covering more obligations, we can generate readable, coherent answers that achieve a more plausible and relatively high score (0.639). |
abbas-etal-2025-structured | 2,025 | Structured Tender Entities Extraction from Complex Tables with Few-short Learning | Extracting structured text from complex tables in PDF tender documents remains a challenging task due to the loss of structural and positional information during the extraction process. AI-based models often require extensive training data, making development from scratch both tedious and time-consuming. Our research focuses on identifying tender entities in complex table formats within PDF documents. To address this, we propose a novel approach utilizing few-shot learning with large language models (LLMs) to restore the structure of extracted text. Additionally, handcrafted rules and regular expressions are employed for precise entity classification. To evaluate the robustness of LLMs with few-shot learning, we employ data-shuffling techniques. Our experiments show that current text extraction tools fail to deliver satisfactory results for complex table structures. However, the few-shot learning approach significantly enhances the structural integrity of extracted data and improves the accuracy of tender entity identification. |
sun-etal-2025-two | 2,025 | A Two-Stage LLM System for Enhanced Regulatory Information Retrieval and Answer Generation | This technical report describes our methodology for the Regulatory Information Retrieval and Answer Generation (RIRAG) Shared Task, a component of the RegNLP workshop at COLING 2025. The challenge aims to effectively navigate and extract relevant information from regulatory texts to generate precise, coherent answers for compliance and obligation-related queries. To tackle subtask1, we introduce a two-stage approach comprising an initial output stage and a subsequent refinement stage. Initially, we fine-tune the LLaMa-2-7B model using LoRA to produce a preliminary output. This is followed by the application of an expert mechanism to enhance the results. For subtask2, we design specific prompt to facilitate the generation of high-quality answers. Consequently, our approach has achieved state-of-the-art performance on the leaderboard, which serves as a testament to the effectiveness and competitiveness of our proposed methodology. |
khan-etal-2025-nust | 2,025 | NUST Nova at RIRAG 2025: A Hybrid Framework for Regulatory Information Retrieval and Question Answering | NUST Nova participates in RIRAG Shared Task, addressing two critical challenges: Task 1 involves retrieving relevant subsections from regulatory documents based on user queries, while Task 2 focuses on generating concise, contextually accurate answers using the retrieved information. We propose a Hybrid Retrieval Framework that combines graph-based retrieval, vector-based methods, and keyword matching BM25 to enhance relevance and precision in regulatory QA. Using score-based fusion and iterative refinement, the framework retrieves the top 10 relevant passages, which are then used by an LLM to generate accurate, context-aware answers. After empirical evaluation, we also conduct an error analysis to identify our framework`s limitations. |
faisal-etal-2025-nust | 2,025 | NUST Alpha at RIRAG 2025: Fusion RAG for Bridging Lexical and Semantic Retrieval and Question Answering | NUST Alpha participates in the Regulatory Information Retrieval and Answer Generation (RIRAG) shared task. We propose FusionRAG that combines OpenAI embeddings, BM25, FAISS, and Rank-Fusion to improve information retrieval and answer generation. We also explores multiple variants of our model to assess the impact of each component in overall performance. FusionRAG strength comes from our rank fusion and filter strategy. Rank fusion integrates semantic and lexical relevance scores to optimize retrieval accuracy and result diversity, and Filter mechanism remove irrelevant passages before answer generation. Our experiments demonstrate that FusionRAG offers a robust and scalable solution for automating the analysis of regulatory documents, improving compliance efficiency, and mitigating associated risks. We further conduct an error analysis to explore the limitations of our model`s performance. |
ameer-etal-2025-nust | 2,025 | NUST Omega at RIRAG 2025: Investigating Context-aware Retrieval and Answer Generations-Lessons and Challenges | NUST Omega participates in Regulatory Information Retrieval and Answer Generation (RIRAG) Shared Task. Regulatory documents poses unique challenges in retrieving and generating precise and relevant answers due to their inherent complexities. We explore the task by proposing a progressive retrieval pipeline and investigate its performance with multiple variants. Some variants include different embeddings to explore their effects on the retrieval score. Some variants examine the inclusion of keyword-driven query matching technique. After exploring such variations, we include topic modeling in our pipeline to investigate its impact on the performance. We also study the performance of various prompt techniques with our proposed pipeline. With empirical experiments, we find some strengths and limitations in the proposed pipeline. These findings will help the research community by offering valuable insights to make advancements in tackling this complex task. |
umar-etal-2025-enhancing | 2,025 | Enhancing Regulatory Compliance Through Automated Retrieval, Reranking, and Answer Generation | This paper explains a Retrieval-Augmented Generation (RAG) pipeline that optimizes reg- ularity compliance using a combination of em- bedding models (i.e. bge-m3, jina-embeddings- v3, e5-large-v2) with reranker (i.e. bge- reranker-v2-m3). To efficiently process long context passages, we introduce context aware chunking method. By using the RePASS met- ric, we ensure comprehensive coverage of obli- gations and minimizes contradictions, thereby setting a new benchmark for RAG-based regu- latory compliance systems. The experimen- tal results show that our best configuration achieves a score of 0.79 in Recall@10 and 0.66 in MAP@10 with LLaMA-3.1-8B model for answer generation. |
bayer-etal-2025-regnlp | 2,025 | A REGNLP Framework: Developing Retrieval-Augmented Generation for Regulatory Document Analysis | This study presents the development of a Retrieval-Augmented Generation (RAG) framework tailored for analyzing regulatory documents from the Abu Dhabi Global Markets (ADGM). The methodology encompasses comprehensive data preprocessing, including extraction, cleaning, and compression of documents, as well as the organization of the ObliQA dataset. The embedding model is utilized for generating embeddings during the retrieval phase, facilitated by the txtai library for managing embeddings and streamlining testing. The training process incorporated innovative strategies such as duplicate recognition, dropout implementation, pooling adjustments, and label modifications to enhance retrieval performance. Hyperparameter tuning further refined the retrieval component, with improvements validated using the recall@10 metric, which measures the proportion of relevant passages among the top-10 results. The refined retrieval component effectively identifies pertinent passages within regulatory documents, expediting information access and supporting compliance efforts. |
quinn-etal-2025-regulatory | 2,025 | Regulatory Question-Answering using Generative AI | Although retrieval augmented generation (RAG) has proven to be an effective approach for creating question-answering systems on a corpus of documents, there is a need to improve the performance of these systems, especially in the regulatory domain where clear and accurate answers are required. This paper outlines the methodology used in our submission to the Regulatory Information Retrieval and Answer Generation (RIRAG) shared task at the Regulatory Natural Language Processing Workshop (RegNLP 2025). The goal is to improve document retrieval (Shared Task 1) and answer generation (Shared Task 2). Our pipeline is constructed as a two-step process for Shared Task 1. In the first step, we utilize a text-embedding-ada-002-based retriever, followed by a RankGPT-based re-ranker. The ranked results of Task 1 are then used to generate responses to user queries in Shared Task 2 through a prompt-based approach using GPT-4o. For Shared Task 1, we achieved a recall rate of 75{\%}, and with the prompts we developed, we were able to generate coherent answers for Shared Task 2. |
zhang-etal-2025-rirag | 2,025 | RIRAG: A Bi-Directional Retrieval-Enhanced Framework for Financial Legal QA in ObliQA Shared Task | In professional financial-legal consulting services, accurately and efficiently retrieving and answering legal questions is crucial. Although some breakthroughs have been made in information retrieval and answer generation, few frameworks have successfully integrated these tasks. Therefore, we propose RIRAG (Retrieval-In-the-loop Response and Answer Generation), a bi-directional retrieval-enhanced framework for financial-legal question answering in ObliQA Shared Task. The system introduces BDD-FinLegal, which means Bi-Directional Dynamic finance-legal, a novel retrieval mechanism specifically designed for financial-legal documents, combining traditional retrieval algorithms with modern neural network methods. Legal answer generation is implemented through large language models retrained on expert-annotated datasets. Our method significantly improves the professionalism and interpretability of the answers while maintaining high retrieval accuracy. Experiments on the ADGM dataset show that the system achieved a significant improvement in the Recall@10 evaluation metric and was recognized by financial legal experts for the accuracy and professionalism of the answer generation. This study provides new ideas for building efficient and reliable question-answering systems in the financial-legal domain. |
aushev-etal-2025-ragulator | 2,025 | RAGulator: Effective RAG for Regulatory Question Answering | Regulatory Natural Language Processing (RegNLP) is a multidisciplinary domain focused on facilitating access to and comprehension of regulatory regulations and requirements. This paper outlines our strategy for creating a system to address the Regulatory Information Retrieval and Answer Generation (RIRAG) challenge, which was conducted during the RegNLP 2025 Workshop. The objective of this competition is to design a system capable of efficiently extracting pertinent passages from regulatory texts (ObliQA) and subsequently generating accurate, cohesive responses to inquiries related to compliance and obligations. Our proposed method employs a lightweight BM25 pre-filtering in retrieving relevant passages. This technique efficiently shortlisting candidates for subsequent processing with Transformer-based embeddings, thereby optimizing the use of resources. |
lee-etal-2025-chain | 2,025 | Chain of Knowledge Graph: Information-Preserving Multi-Document Summarization for Noisy Documents | With the advent of large language models, the complexity of multi-document summarization task has been substantially reduced. The summarization process must effectively handle noisy documents that are irrelevant to the main topic while preserving essential information. Recently, Chain-of-Density (COD) and Chain-of-Event (CoE) have proposed prompts to effectively handle the noisy documents by using entity-centric approaches for the summarization. However, CoD and CoE are prone to information loss during entity extraction due to their tendency to overly filter out entities perceived as less critical but that could still be important. In this paper, we propose a novel instruction prompt termed as Chain of Knowledge Graph (CoKG) for multi-document summarization. Our prompt extracts entities and constructs relationships between entities to form a Knowledge Graph (KG). Next, the prompt enriches these relationships to recognize potentially important entities and assess the strength of each relation. If the acquired KG meets a predefined quality level, the KG is used to summarize the given documents. This process helps alleviate the information loss in multi-document summarization. Experimental results demonstrate that our prompt effectively preserves key entities and is robust to noisy documents. |
sun-etal-2025-cegrl | 2,025 | CEGRL-TKGR: A Causal Enhanced Graph Representation Learning Framework for Temporal Knowledge Graph Reasoning | Temporal knowledge graph reasoning (TKGR) is increasingly gaining attention for its ability to extrapolate new events from historical data, thereby enriching the inherently incomplete temporal knowledge graphs. Existing graph-based representation learning frameworks have made significant strides in developing evolving representations for both entities and relational embeddings. Despite these achievements, there`s a notable tendency in these models to inadvertently learn biased data representations and mine spurious correlations, consequently failing to discern the causal relationships between events. This often leads to incorrect predictions based on these false correlations. To address this, we propose an innovative \textbf{C}ausal \textbf{E}nhanced \textbf{G}raph \textbf{R}epresentation \textbf{L}earning framework for TKGR (named CEGRL-TKGR). This framework introduces causal structures in graph-based representation learning to unveil the essential causal relationships between events, ultimately enhancing the performance of the TKGR task. Speci{\~A}{\^A}{\~A}{\^A} ̄{\~A}{\^A}{\~A}{\^A}{\textlnot}{\~A}{\^A}{\~A}{\^A}cally, we first disentangle the evolutionary representations of entities and relations in a temporal knowledge graph sequence into two distinct components, namely causal representations and confounding representations. Then, drawing on causal intervention theory, we advocate the utilization of causal representations for predictions, aiming to mitigate the effects of erroneous correlations caused by confounding features, thus achieving more robust and accurate predictions. Finally, extensive experimental results on six benchmark datasets demonstrate the superior performance of our model in the link prediction task. |
bai-etal-2025-reasoning | 2,025 | Reasoning Knowledge Filter for Logical Table-to-Text Generation | Logical table-to-text generation (LT2T) seeks to produce logically faithful textual descriptions base on tables. Current end-to-end LT2T models, which use descriptions directly as learning objectives, frequently face challenges in maintaining logical faithfulness due to the lack of a reasoning knowledge. Recent research have introduced reasoning knowledge generated by models for LT2T task, but the noise along with it limited its performance. We therefore propose a framework reasoning knowledge filter that leverages the collaboration between large language models and smaller models to filter data points with high-quality reasoning knowledge. This framework aims to provide highly matched table, description and reasoning knowledge triplets for LT2T. The results obtained on LogicNLG database demonstrate that the efficiencies of the method in this paper has achieved optimal performance with a reduced amount of data. Specifically, it enhances SP-Acc by 1.4 points and NLI-Acc by 0.7 points compared to the current state-of-the-art model. |
sun-etal-2025-chain | 2,025 | From Chain to Tree: Refining Chain-like Rules into Tree-like Rules on Knowledge Graphs | With good explainability and controllability, rule-based methods play an important role in the task of Knowledge Graph Completion (KGC). However, existing studies primarily focused on learning chain-like rules, whose chain-like structure limits their expressive power. Consequently, chain-like rules often exhibit lower Standard Confidence, and are prone to the incorrect grounding values during reasoning, thus producing erroneous reasoning results. In this paper, we propose the concept of tree-like rules on knowledge graphs to expand the scope of the application and improve the reasoning ability of rule-based methods. To achieve this, we formalize the problem of tree-like rule refinement and propose an effective framework for refining chain-like rules into tree-like rules. Experimental evaluations on four public datasets demonstrate that the proposed framework can seamlessly adapt to various chain-like rule induction methods and the refined tree-like rules consistently exhibit higher Standard Confidence and achieve better performances than the original chain-like rules on link prediction tasks. Furthermore, we illustrate that the improvements brought by tree-like rules are positively correlated with the density of the knowledge graphs. The data and code of this paper can be available at https://github.com/forangel2014/tree-rule. |
guo-etal-2025-lab | 2,025 | LAB-KG: A Retrieval-Augmented Generation Method with Knowledge Graphs for Medical Lab Test Interpretation | Laboratory tests generate structured numerical data, which a clinician must interpret to justify diagnoses and help patients understand the outcomes of the tests. LLMs have the potential to assist with the generation of interpretative comments, but legitimate concerns remain about the accuracy and reliability of the generation process. This work introduces LAB-KG, which conditions the generation process of an LLM on information retrieved from a knowledge graph of relevant patient conditions and lab test results. This helps to ground the text-generation process in accurate medical knowledge and enables generated text to be traced back to the knowledge graph. Given a dataset of laboratory test results and associated interpretive comments, we show how an LLM can build a KG of the relationships between laboratory test results, reference ranges, patient conditions and demographic information. We further show that the interpretive comments produced by an LLM conditioned on information retrieved from the KG are of higher quality than those from a standard RAG method. Finally, we show how our KG approach can improve the interpretability of the LLM generated text. |
dong-etal-2025-bridging | 2,025 | Bridging Language and Scenes through Explicit 3-D Model Construction | We introduce the methodology of explicit model construction to bridge linguistic descriptions and scene perception and demonstrate that in Visual Question-Answering (VQA) using MC4VQA (Model Construction for Visual Question-Answering), a method developed by us. Given a question about a scene, our MC4VQA first recognizes objects utilizing pre-trained deep learning systems. Then, it constructs an explicit 3-D layout by repeatedly reducing the difference between the input scene image and the image rendered from the current 3-D spatial environment. This novel {\textquotedblleft}iterative rendering{\textquotedblright} process endows MC4VQA the capability of acquiring spatial attributes without training data. MC4VQA outperforms NS-VQA (the SOTA system) by reaching 99.94{\%} accuracy on the benchmark CLEVR datasets, and is more robust than NS-VQA on new testing datasets. With newly created testing data, NS-VQA`s performance dropped to 97.60{\%}, while MC4VQA still kept the 99.0{\%} accuracy. This work sets a new SOTA performance of VQA on the benchmark CLEVR datasets, and shapes a new method that may solve the out-of-distribution problem. |
bai-etal-2025-vcrmner | 2,025 | VCRMNER: Visual Cue Refinement in Multimodal NER using CLIP Prompts | With the continuous growth of multi-modal data on social media platforms, traditional Named Entity Recognition has rendered insufficient for handling contemporary data formats. Consequently, researchers proposed Multi-modal Named Entity Recognition (MNER). Existing studies focus on capturing the visual regions corresponding to entities to assist in entity recognition. However, these approaches still struggle to mitigate interference from visual regions that are irrelevant to the entities. To address this issue, we propose an innovative framework, Visual Cue Refinement in MNER(VCRMNER) using CLIP Prompts, to accurately capture visual cues (object-level visual regions) associated with entities. We leverage prompts to represent the semantic information of entity categories, which helps us assess visual cues and minimize interference from those irrelevant to the entities. Furthermore, we designed an interaction transformer that operates in two stages{---}first within each modality and then between modalities{---}to refine visual cues by learning from a frozen image encoder, thereby reducing differences between text and visual modalities. Comprehensive experiments were conducted on two public datasets, Twitter15 and Twitter17. The results and detailed analyses demonstrate that our method exhibits robust and competitive performance. |
kang-etal-2025-neuro | 2,025 | Neuro-Conceptual Artificial Intelligence: Integrating OPM with Deep Learning to Enhance Question Answering Quality | Knowledge representation and reasoning are critical challenges in Artificial Intelligence (AI), particularly in integrating neural and symbolic approaches to achieve explainable and transparent AI systems. Traditional knowledge representation methods often fall short of capturing complex processes and state changes. We introduce Neuro-Conceptual Artificial Intelligence (NCAI), a specialization of the neuro-symbolic AI approach that integrates conceptual modeling using Object-Process Methodology (OPM) ISO 19450:2024 with deep learning to enhance question-answering (QA) quality. By converting natural language text into OPM models using in-context learning, NCAI leverages the expressive power of OPM to represent complex OPM elements{---}processes, objects, and states{---}beyond what traditional triplet-based knowledge graphs can easily capture. This rich structured knowledge representation improves reasoning transparency and answer accuracy in an OPM-QA system. We further propose transparency evaluation metrics to quantitatively measure how faithfully the predicted reasoning aligns with OPM-based conceptual logic. Our experiments demonstrate that NCAI outperforms traditional methods, highlighting its potential for advancing neuro-symbolic AI by providing rich knowledge representations, measurable transparency, and improved reasoning. |
al-saeedi-harma-2025-emergence | 2,025 | Emergence of symbolic abstraction heads for in-context learning in large language models | Large Language Models (LLMs) based on self-attention circuits are able to perform, at inference time, novel reasoning tasks, but the mechanisms inside the models are currently not fully understood. We assume that LLMs are able to generalize abstract patterns from the input and form an internal symbolic internal representation of the content. In this paper, we study this by analyzing the performance of small LLM models trained with sequences of instantiations of abstract sequential symbolic patterns or templates. It is shown that even a model with two layers is able to learn an abstract template and use it to generate correct output representing the pattern. This can be seen as a form of symbolic inference taking place inside the network. In this paper, we call the emergent mechanism abstraction head. Identifying mechanisms of symbolic reasoning in a neural network can help to find new ways to merge symbolic and neural processing. |
zinova-etal-2025-linking | 2,025 | Linking language model predictions to human behaviour on scalar implicatures | We explore the behaviour of language models on adjectival scales in connection with negation when prompted with material used in human experiments. We propose several metrics extracted from the model predictions and analyze those metrics in relation to human data as well as use them to propose new items to be tested in human experiments. |
tayyar-madabushi-etal-2025-generative | 2,025 | Generative FrameNet: Scalable and Adaptive Frames for Interpretable Knowledge Storage and Retrieval for LLMs Powered by LLMs | Frame semantics provides an explanation for how we make use of conceptual frames, which encapsulate background knowledge and associations, to more completely understand the meanings of words within a context. Unfortunately, FrameNet, the only widely available implementation of frame semantics, is limited in both scale and coverage. Therefore, we introduce a novel mechanism for generating task-specific frames using large language models (LLMs), which we call Generative FrameNet. We demonstrate its effectiveness on a task that is highly relevant in the current landscape of LLMs: the interpretable storage and retrieval of factual information. Specifically, Generative Frames enable the extension of Retrieval-Augmented Generation (RAG), providing an interpretable framework for reducing inaccuracies in LLMs. We conduct experiments to demonstrate the effectiveness of this method both in terms of retrieval effectiveness as well as the relevance of the automatically generated frames and frame relations. Expert analysis shows that Generative Frames capture a more suitable level of semantic specificity than the frames from FrameNet. Thus, Generative Frames capture a notion of frame semantics that is closer to Fillmore`s originally intended definition, and offer potential for providing data-driven insights into Frame Semantics theory. Our results also show that this novel mechanism of Frame Semantic-based interpretable retrieval improves RAG for question answering with LLMs{---}outperforming a GPT-4 based baseline by up to 8 points. We provide open access to our data, including prompts and Generative FrameNet. |
sabra-2025-deciphering | 2,025 | Deciphering Implicatures: On NLP and Oral Testimonies | The utterance of a word does not intrinsically convey its intended force. The semantic of utterances is not shaped by the precise references of the words used. Asserting that {\textquotedblleft}it is shameful to abandon our country{\textquotedblright} does not merely convey information; rather, it asserts an act of resilience. In most of our exchanges, we rarely utilize sentences to describe reality or the world around us. More frequently, our statements aim to express opinions, to influence, or be influenced by others. Words carry more than just their syntax and semantics; they also embody a pragmatic normative force. This divergence between literal and conveyed meaning was depicted in the literature of philosophy of language as the difference between sentence meaning and speaker meaning. Where the former is the literal understanding of the words combined in a sentence, the latter is what the speaker is trying to convey through her expression. In order to derive the speaker meaning from the sentence meaning, J.L. Austin (the author of How To Do Things with Words) relied on conventions, whereas H.P. Grice (the author of Logic and Conversations) relied on conventional and non conventional implicatures. This paper aims to decipher how we can infer speaker meaning from sentence meaning and thereby capture the force of what has been articulated, focusing specifically on oral testimonies. I argue that oral testimonies are forms of speech acts that aim to produce normative changes. Following this discussion, I will examine various natural language processing (NLP) models that make explicit what is implicit in oral testimonies with its benefits and limitations. Lastly, I will address two challenges, the former is related to implicatures that are not governed by conventions and the latter is concerned with the biases inherent in hermeneutical approaches. |
regier-khalidi-2025-cultural | 2,025 | A cultural shift in Western perceptions of Palestine | We argue that a cultural shift in Western perceptions of Palestine began in the late 1990s to 2000s, leading to increased openness to Palestinian perspectives, including awareness of the Nakba. We present 3 computational analyses designed to test this idea against data from the 2020 Google Books English dataset. The results support the claim of a cultural shift, and help to characterize that shift. |
lamar-etal-2025-cognitive | 2,025 | Cognitive Geographies of Catastrophe Narratives: Georeferenced Interview Transcriptions as Language Resource for Models of Forced Displacement | We present a machine-understandable geotagged dataset of translated interviews from the Nakba Archive alongside a complete georeferenced dataset of named locations mentioned in the interviews. In a preliminary analysis of this dataset, we find that the cognitive relationship of interviewees to place and spatiality is significantly correlated with gender. Our data also shows that interviewees with birthplaces depopulated in the 1948 Nakba incorporate references to named places in their interviews in substantially different ways than other interviewees. This suggests that the status of the interviewee`s birthplace may impact the way they narrate their experiences. Our work serves as a foundation for continued and expanded statistical and cognitive models of Palestinian forced displacement. |
ashqar-2025-sentiment | 2,025 | Sentiment Analysis of Nakba Oral Histories: A Critical Study of Large Language Models | This study explores the use of Large Language Models (LLMs), specifically ChatGPT, for sentiment analysis of Nakba oral histories, which document the experiences of Palestinian refugees. The study compares sentiment analysis results from full testimonies (average 2500 words) and their summarized versions (300 words). The findings reveal that summarization increased positive sentiment and decreased negative sentiment, suggesting that the process may highlight more hopeful themes while oversimplifying emotional complexities. The study highlights both the potential and limitations of using LLMs for analyzing sensitive, trauma-based narratives and calls for further research to improve sentiment analysis in such contexts. |
abuhaija-etal-2025-nakba | 2,025 | The Nakba Lexicon: Building a Comprehensive Dataset from Palestinian Literature | This paper introduces the Nakba Lexicon, a comprehensive dataset derived from the poetry collection \textit{Asifa {\textquoteleft}Ala al-Iz{\textquoteleft}aj} (Sorry for the Disturbance) by Istiqlal Eid, a Palestinian poet from El-Birweh. Eid`s work poignantly reflects on themes of Palestinian identity, displacement, and resilience, serving as a resource for preserving linguistic and cultural heritage in the context of post-Nakba literature. The dataset is structured into ten thematic domains, including political terminology, memory and preservation, sensory and emotional lexicon, toponyms, nature, and external linguistic influences such as Hebrew, French, and English, thereby capturing the socio-political, emotional, and cultural dimensions of the Nakba. The Nakba Lexicon uniquely emphasises the contributions of women to Palestinian literary traditions, shedding light on often-overlooked narratives of resilience and cultural continuity. Advanced Natural Language Processing (NLP) techniques were employed to analyse the dataset, with fine-tuned pre-trained models such as ARABERT and MARBERT achieving F1-scores of 0.87 and 0.68 in language and lexical classification tasks, respectively, significantly outperforming traditional machine learning models. These results highlight the potential of domain-specific computational models to effectively analyse complex datasets, facilitating the preservation of marginalised voices. By bridging computational methods with cultural preservation, this study enhances the understanding of Palestinian linguistic heritage and contributes to broader efforts in documenting and analysing endangered narratives. The Nakba Lexicon paves the way for future interdisciplinary research, showcasing the role of NLP in addressing historical trauma, resilience, and cultural identity. |
hamed-zaidkilani-2025-arabic | 2,025 | Arabic Topic Classification Corpus of the Nakba Short Stories | In this paper, we enrich Arabic Natural Language Processing (NLP) resources by introducing the {\textquotedblleft}Nakba Topic Classification Corpus (NTCC),{\textquotedblright} a novel annotated Arabic corpus derived from narratives about the Nakba. The NTCC comprises approximately 470 sentences extracted from eight short stories and captures the thematic depth of the Nakba narratives, providing insights into both historical and personal dimensions. The corpus was annotated in a two-step process. One third of the dataset was manually annotated, achieving an IAA of 87{\%} (later resolved to 100{\%}), while the rest was annotated using a rule-based system based on thematic patterns. This approach ensures consistency and reproducibility, enhancing the corpus`s reliability for NLP research. The NTCC contributes to the preservation of the Palestinian cultural heritage while addressing key challenges in Arabic NLP, such as data scarcity and linguistic complexity. By like topic modeling and classification tasks, the NTCC offers a valuable resource for advancing Arabic NLP research and fostering a deeper understanding of the Nakba narratives |
hamed-zaidkilani-2025-exploring | 2,025 | Exploring Author Style in Nakba Short Stories: A Comparative Study of Transformer-Based Models | Measuring semantic similarity and analyzing authorial style are fundamental tasks in Natural Language Processing (NLP), with applications in text classification, cultural analysis, and literary studies. This paper investigates the semantic similarity and stylistic features of Nakba short stories, a key component of Palestinian literature, using transformer-based models, AraBERT, BERT, and RoBERTa. The models effectively capture nuanced linguistic structures, cultural contexts, and stylistic variations in Arabic narratives, outperforming the traditional TF-IDF baseline. By comparing stories of similar length, we minimize biases and ensure a fair evaluation of both semantic and stylistic relationships. Experimental results indicate that RoBERTa achieves slightly higher performance, highlighting its ability to distinguish subtle stylistic patterns. This study demonstrates the potential of AI-driven tools to provide more in-depth insights into Arabic literature, and contributes to the systematic analysis of both semantic and stylistic elements in Nakba narratives. |
hamarsheh-etal-2025-detecting | 2,025 | Detecting Inconsistencies in Narrative Elements of Cross Lingual Nakba Texts | This paper suggests a methodology for contradiction detection in cross lingual texts about the Nakba. We propose a pipeline that includes text translation using Google`s Gemini for context-aware translations, followed by a fact extraction task using either Gemini or the TextRank algorithm. We then apply Natural Language Inference (NLI) by using models trained for this task, such as XLM-RoBERTa and BART to detect contradictions from different texts about the Nakba. We also describe how the performance of such NLI models is affected by the complexity of some sentences as well as the unique syntactic and semantic characteristics of the Arabic language. Additionally, we introduce a method using cosine similarity of vector embeddings of facts for identifying missing or underrepresented topics among historical narrative texts. The approach we propose in this paper provides insights into biases, contradictions, and gaps in narratives surrounding the Nakba, offering a deeper understanding of historical perspectives. |
ragab-etal-2025-multilingual | 2,025 | Multilingual Propaganda Detection: Exploring Transformer-Based Models mBERT, XLM-RoBERTa, and mT5 | This research investigates multilingual propaganda detection by employing transformer-based models, specifically mBERT, XLM-RoBERTa, and mT5. The study utilizes a balanced dataset from the BiasFigNews corpus, annotated for propaganda and bias across five languages. The models were finely tuned to generate embeddings for classification tasks. The evaluation revealed mT5 as the most effective model, achieving an accuracy of 99.61{\%} and an F1-score of 0.9961, followed by mBERT and XLM-RoBERTa with accuracies of 92{\%} and 91.41{\%}, respectively. The findings demonstrate the efficacy of transformer-based embeddings in detecting propaganda while also highlighting challenges in subtle class distinctions. Future work aims to enhance cross-lingual adaptability and explore lightweight models for resource-constrained settings. |
awad-etal-2025-collective | 2,025 | Collective Memory and Narrative Cohesion: A Computational Study of Palestinian Refugee Oral Histories in Lebanon | This study uses the Palestinian Oral History Archive (POHA) to investigate how Palestinian refugee groups in Lebanon sustain a cohesive collective memory of the Nakba through shared narratives. Grounded in Halbwachs' theory of group memory, we employ statistical analysis of pairwise similarity of narratives, focusing on the influence of shared gender and location. We use textual representation and semantic embeddings of narratives to represent the interviews themselves. Our analysis demonstrates that shared origin is a powerful determinant of narrative similarity across thematic keywords, landmarks, and significant figures, as well as in semantic embeddings of the narratives. Meanwhile, shared residence fosters cohesion, with its impact significantly amplified when paired with shared origin. Additionally, women`s narratives exhibit heightened thematic cohesion, particularly in recounting experiences of the British occupation, underscoring the gendered dimensions of memory formation. This research deepens the understanding of collective memory in diasporic settings, emphasizing the critical role of oral histories in safeguarding Palestinian identity and resisting erasure. |
garcia-corral-etal-2025-missing | 2,025 | The Missing Cause: An Analysis of Causal Attributions in Reporting on Palestine | Missing cause bias is a specific type of bias in media reporting that relies on consistently omitting causal attribution to specific events, for example when omitting specific actors as causes of incidents. Identifying these patterns in news outlets can be helpful in assessing the level of bias present in media content. In this paper, we examine the prevalence of this bias in reporting on Palestine by identifying causal constructions in headlines. We compare headlines from three main news media outlets: CNN, the BBC, and AJ (AlJazeera), that cover the Israel-Palestine conflict. We also collect and compare these findings to data related to the Ukraine-Russia war to analyze editorial style within press organizations. We annotate a subset of this data and evaluate two causal language models (UniCausal and GPT-4o) for the identification and extraction of causal language in news headlines. Using the top performing model, GPT-4o, we machine annotate the full corpus and analyze missing bias prevalence within and across news organizations. Our findings reveal that BBC headlines tend to avoid directly attributing causality to Israel for the violence in Gaza, both when compared to other news outlets, and to its own reporting on other conflicts. |
mohammed-etal-2025-bias | 2,025 | Bias Detection in Media: Traditional Models vs. Transformers in Analyzing Social Media Coverage of the Israeli-Gaza Conflict | Bias in news reporting significantly influences public perception, particularly in sensitive and polarized contexts like the Israel-Gaza conflict. Detecting bias in such cases presents unique challenges due to political, cultural, and ideological complexities, often amplifying disparities in reporting. While prior research has addressed media bias and dataset fairness, these approaches inadequately capture the nuanced dynamics of the Israel-Gaza conflict. To address this gap, we propose an NLP-based framework that leverages Nakba narratives as linguistic resources for bias detection in news coverage. Using a multilingual corpus focusing on Arabic texts, we apply rigorous data cleaning, pre-processing, and methods to mitigate imbalanced class distributions that could skew classification outcomes. Our study explores various approaches, including Machine Learning (ML), Deep Learning (DL), Transformer-based architectures, and generative models. The findings demonstrate promising advancements in automating bias detection, and enhancing fairness and accuracy in politically sensitive reporting. |
bilgin-tasdemir-ozates-2025-nakbatr | 2,025 | NakbaTR: A Turkish NER Dataset for Nakba Narratives | This paper introduces a novel, annotated Named Entity Recognition (NER) dataset derived from a collection of 181 news articles about the Nakba and its witnesses. Given their prominence as a primary source of information on the Nakba in Turkish, news articles were selected as the primary data source. Some 4,032 news sentences are collected from web sites of two news agencies, Anadolu Ajans{\i} and TRTHaber. We applied a filtering process to make sure that only the news which contain witness testimonies regarding the ongoing Nakba are included in the dataset. After a semi-automatic annotation for entities of type Person, Location, and Organization, we obtained a NER dataset of 2,289 PERSON, 5,875 LOCATION, and 1,299 ORGANIZATION tags. We expect the dataset to be useful in several NLP tasks such as sentiment analysis and relation extraction for Nakba event while providing a new language resource for Turkish. As a future work, we aim to improve the dataset by increasing the number of news and entity types. |
nabhani-etal-2025-integrating | 2,025 | Integrating Argumentation Features for Enhanced Propaganda Detection in Arabic Narratives on the Israeli War on Gaza | Propaganda significantly shapes public opinion, especially in conflict-driven contexts like the Israeli-Palestinian conflict. This study explores the integration of argumentation features, such as claims, premises, and major claims, into machine learning models to enhance the detection of propaganda techniques in Arabic media. By leveraging datasets annotated with fine-grained propaganda techniques and employing crosslingual and multilingual NLP methods, along with GPT-4-based annotations, we demonstrate consistent performance improvements. A qualitative analysis of Arabic media narratives on the Israeli war on Gaza further reveals the model`s capability to identify diverse rhetorical strategies, offering insights into the dynamics of propaganda. These findings emphasize the potential of combining NLP with argumentation features to foster transparency and informed discourse in politically charged settings. |
bennie-etal-2025-panda | 2,025 | PANDA - Paired Anti-hate Narratives Dataset from Asia: Using an LLM-as-a-Judge to Create the First Chinese Counterspeech Dataset | Despite the global prevalence of Modern Standard Chinese language, counterspeech (CS) resources for Chinese remain virtually nonexistent. To address this gap in East Asian counterspeech research we introduce the a corpus of Modern Standard Mandarin counterspeech that focuses on combating hate speech in Mainland China. This paper proposes a novel approach of generating CS by using an LLM-as-a-Judge, simulated annealing, LLMs zero-shot CN generation and a round-robin algorithm. This is followed by manual verification for quality and contextual relevance. This paper details the methodology for creating effective counterspeech in Chinese and other non-Eurocentric languages, including unique cultural patterns of which groups are maligned and linguistic patterns in what kinds of discourse markers are programmatically marked as hate speech (HS). Analysis of the generated corpora, we provide strong evidence for the lack of open-source, properly labeled Chinese hate speech data and the limitations of using an LLM-as-Judge to score possible answers in Chinese. Moreover, the present corpus servers as the first East Asian language based CS corpus and provides an essential resource for future research on counterspeech generation and evaluation. |
v-2025-rssn | 2,025 | RSSN at Multilingual Counterspeech Generation: Leveraging Lightweight Transformers for Efficient and Context-Aware Counter-Narrative Generation | This paper presents a system for counter-speech generation, developed for the COLING 2025 shared task. By leveraging lightweight transformer models, DistilBART and T5-small, we optimize computational efficiency while maintaining strong performance. The work includes an in-depth analysis of a multilingual dataset, addressing hate speech instances across diverse languages and target groups. Through systematic error analysis, we identify challenges such as lack of specificity and context misinterpretation in generated counter-narratives. Evaluation metrics like BLEU, ROUGE, and BERTScore demonstrate the effectiveness of our approaches, while comparative insights highlight complementary strengths in fluency, contextual integration, and creativity. Future directions focus on enhancing preprocessing, integrating external knowledge sources, and improving scalability. |
wadhwa-etal-2025-northeastern | 2,025 | Northeastern Uni at Multilingual Counterspeech Generation: Enhancing Counter Speech Generation with LLM Alignment through Direct Preference Optimization | The automatic generation of counter-speech (CS) is a critical strategy for addressing hate speech by providing constructive and informed responses. However, existing methods often fail to generate high-quality, impactful, and scalable CS, particularly across diverse lin- guistic contexts. In this paper, we propose a novel methodology to enhance CS generation by aligning Large Language Models (LLMs) using Supervised Fine-Tuning (SFT) and Di- rect Preference Optimization (DPO). Our ap- proach leverages DPO to align LLM outputs with human preferences, ensuring contextu- ally appropriate and linguistically adaptable responses. Additionally, we incorporate knowl- edge grounding to enhance the factual accuracy and relevance of generated CS. Experimental results demonstrate that DPO-aligned models significantly outperform SFT baselines on CS benchmarks while scaling effectively to mul- tiple languages. These findings highlight the potential of preference-based alignment tech- niques to advance CS generation across var- ied linguistic settings. The model supervision and alignment is done in English and the same model is used for reporting metrics across other languages like Basque, Italian, and Spanish. |
marquez-etal-2025-nlp | 2,025 | NLP@IIMAS-CLTL at Multilingual Counterspeech Generation: Combating Hate Speech Using Contextualized Knowledge Graph Representations and LLMs | We present our approach for the shared task on Multilingual Counterspeech Generation (MCG) to counteract hate speech (HS) in Spanish, English, Basque, and Italian. To accomplish this, we followed two different strategies: 1) a graph-based generative model that encodes graph representations of knowledge related to hate speech, and 2) leveraging prompts for a large language model (LLM), specifically GPT-4o. We find that our graph-based approach tends to perform better in terms of traditional evaluation metrics (i.e., RougeL, BLEU, BERTScore), while the JudgeLM evaluation employed in the shared task favors the counter-narratives generated by the LLM-based approach, which was ranked second for English and third for Spanish on the leaderboard. |
bennie-etal-2025-codeofconduct | 2,025 | CODEOFCONDUCT at Multilingual Counterspeech Generation: A Context-Aware Model for Robust Counterspeech Generation in Low-Resource Languages | This paper introduces a context-aware model for robust counterspeech generation, which achieved significant success in the MCG-COLING-2025 shared task. Our approach particularly excelled in low-resource language settings. By leveraging a simulated annealing algorithm fine-tuned on multilingual datasets, the model generates factually accurate responses to hate speech. We demonstrate state-of-the-art performance across four languages (Basque, English, Italian, and Spanish), with our system ranking first for Basque, second for Italian, and third for both English and Spanish. Notably, our model swept all three top positions for Basque, highlighting its effectiveness in low-resource scenarios. Evaluation of the shared task employs both traditional metrics (BLEU, ROUGE, BERTScore, Novelty) and the LLM-based JudgeLM. We present a detailed analysis of our results, including error cases and potential improvements. This work contributes to the growing body of research on multilingual counterspeech generation, offering insights into developing robust models that can adapt to diverse linguistic and cultural contexts in the fight against online hate speech. |
lyu-etal-2025-hw | 2,025 | HW-TSC at Multilingual Counterspeech Generation | Multilingual counterspeech generation (MCSG) contributes to generating counterspeech with respectful, non-offensive information that is specific and truthful for the given hate speech, especially those for languages other than English. Generally, the training data of MCSG in low-source language is rare and hard to curate. Even with the impressive large language models (LLMs), it is a struggle to generate an appreciative counterspeech under the multilingual scenario. In this paper, we design a pipeline with a generation-reranking mode to effectively generate counterspeech under the multilingual scenario via LLM. Considering the scarcity of training data, we first utilize the training-free strategy, i.e., in-context learning (ICL), to generate the candidate counterspeechs. Then, we propose to rerank those candidate counterspeech via the Elo rating algorithm and a fine-tuned reward model. Experimental results on four languages, including English (EN), Italian (IT), Basque (EU) and Spanish (ES), our system achieves a comparative or even better performance in four metrics compared to the winner in this shared task. |
moscato-etal-2025-mnlp | 2,025 | MilaNLP@Multilingual Counterspeech Generation: Evaluating Translation and Background Knowledge Filtering | We describe our participation in the Multilingual Counterspeech Generation shared task, which aims to generate a counternarrative to counteract hate speech, given a hateful sentence and relevant background knowledge. Our team tested two different aspects: translating outputs from English vs generating outputs in the original languages and filtering pieces of the background knowledge provided vs including all the background knowledge. Our experiments show that filtering the background knowledge in the same prompt and leaving data in the original languages leads to more adherent counternarrative generations, except for Basque, where translating the output from English and filtering the background knowledge in a separate prompt yields better results. Our system ranked first in English, Italian, and Spanish and fourth in Basque. |
farhan-2025-hyderabadi | 2,025 | Hyderabadi Pearls at Multilingual Counterspeech Generation : HALT : Hate Speech Alleviation using Large Language Models and Transformers | This paper explores the potential of using fine- tuned Large Language Models (LLMs) for generating counter-narratives (CNs) to combat hate speech (HS). We focus on English and Basque, leveraging the ML{\_}MTCONAN{\_}KN dataset, which provides hate speech and counter-narrative pairs in multiple languages. Our study compares the performance of Mis- tral, Llama, and a Llama-based LLM fine- tuned on a Basque language dataset for CN generation. The generated CNs are evalu- ated using JudgeLM (a LLM to evaluate other LLMs in open-ended scenarios) along with traditional metrics such as ROUGE-L, BLEU, BERTScore, and other traditional metrics. The results demonstrate that fine-tuned LLMs can produce high-quality contextually relevant CNs for low-resource languages that are comparable to human-generated responses, offering a sig- nificant contribution to combating online hate speech across diverse linguistic settings. |
russo-2025-trenteam | 2,025 | TrenTeam at Multilingual Counterspeech Generation: Multilingual Passage Re-Ranking Approaches for Knowledge-Driven Counterspeech Generation Against Hate | Hate speech (HS) in online spaces poses severe risks, including real-world violence and psychological harm to victims, necessitating effective countermeasures. Counterspeech (CS), which responds to hateful messages with opposing yet non-hostile narratives, offer a promising solution by mitigating HS while upholding free expression. However, the growing volume of HS demands automation, making Natural Language Processing a viable solution for the automatic generation of CS. Recent works have explored knowledge-driven approaches, leveraging external sources to improve the relevance and informativeness of responses. These methods typically involve multi-step pipelines combining retrieval and passage re-ranking modules. While effective, most studies have focused on English, with limited exploration of multilingual contexts. This paper addresses these gaps by proposing a multilingual, knowledge-driven approach to CS generation. We integrate state-of-the-art re-ranking mechanisms into the CS generation pipeline and evaluate them using the MT-CONAN-KN dataset, which includes hate speech, relevant knowledge sentences, and counterspeech in four languages: English, Italian, Spanish, and Basque. Our approach compares reranker-based systems employing multilingual cross-encoders and LLMs to a simpler end-to-end system where the language model directly handles both knowledge selection and CS generation. Results demonstrate that reranker-based systems outperformed end-to-end systems in syntactic and semantic similarity metrics, with LLM-based re-rankers delivering the strongest performance overall. This work is the result of our participation in the Shared Task on Multilingual Counterspeech Generation held at COLING 2025. |
End of preview. Expand
in Data Studio
ACL abstracts extracted from the official ACL Anthology BibTeX.
- Downloads last month
- 20