README.md

---
tags:
- marathi
- translation
- text-classification
- semantic-search
- nlp
- ai
- deep-learning
- language-models
- text-processing
- multilingual
- huggingface-dataset
- dataset
language:
- mr
license: cc-by-nc-nd-4.0
task_categories:
- text-classification
- text-retrieval
- sentence-similarity
- token-classification
- question-answering
- zero-shot-classification
- summarization
- text-generation
- multiple-choice
- document-question-answering
size_categories:
- 100K<n<1M
pretty_name: Specter Marathi Dataset
dataset_info:
  features:
  - name: anchor
    dtype: string
  - name: positive
    dtype: string
  - name: negative
    dtype: string
  splits:
  - name: train
    num_bytes: 393484095
    num_examples: 684098
  download_size: 97559561
  dataset_size: 393484095
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# **Specter Marathi Dataset: High-Quality Marathi NLP Corpus**

## 📌 **Overview**
The **Specter Marathi dataset** is a meticulously curated collection of **684098** rows of Marathi text, ensuring linguistic accuracy and natural flow. Every sentence has been **verified by native Marathi speakers** to maintain **contextual integrity and correctness**.

This dataset is designed for **semantic search, text classification, and various NLP tasks**, making it a valuable resource for **machine learning models dealing with Marathi language understanding**.

---

## 🔥 **Key Features**
✅ **High-Quality Marathi Data** – Professionally curated and reviewed.  
✅ **Native Speaker Verification** – Every row is manually checked for accuracy.  
✅ **Optimized for NLP Tasks** – Useful for **semantic modeling, text classification, and retrieval-based applications**.  
✅ **Clean and Structured Data** – Ready-to-use for deep learning and transformer models.  
✅ **Supports Diverse Use Cases** – Ideal for **zero-shot classification, sentiment analysis, and language comprehension models**.  

---

## 📂 **Dataset Structure**
- **Language**: Marathi (`mr`)  
- **Rows**: 684098  
- **Columns**: anchor, positive, negative  
- **Size**: 375.26 MB (0.37 GB)  
- **Format**: Text corpus (structured sentences & paragraphs)  
- **Validation**: Human-reviewed for correctness  
- **Intended Use**: Training NLP models for **classification, retrieval, and understanding tasks**  

---

## 🚀 **Usage Guide**
To integrate this dataset into your NLP workflow, use the following code:

```python
from datasets import load_dataset

dataset = load_dataset("Singhchandann/specter_marathi")
```

The dataset can be directly used for **training, fine-tuning, and evaluating NLP models**.

---

Here’s a properly structured and refined version of your license section:  

---

## 🔒 **License & Accessibility**  

### **📜 License: CC BY-NC-ND 4.0**  
This dataset is licensed under the **Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 (CC BY-NC-ND 4.0)** license.  

- **✅ Attribution Required:** You must provide proper credit when using this dataset after getting permission.  
- **⛔ No Commercial Use:** This dataset **cannot** be used for commercial purposes.  
- **⛔ No Modifications:** You **cannot** modify, adapt, or create derivative datasets from this work.  

For any special permissions or exceptions, please contact on: **[Author Email](mailto:[email protected])**  

🔗 **License Details:** [CC BY-NC-ND 4.0](https://creativecommons.org/licenses/by-nc-nd/4.0/)  

---

### **🔐 Access & Permissions**  
- This dataset is **private** and **not publicly available**.  
- Only **authorized users** with explicit permission can access and utilize this dataset.  

📩 **For access requests or further inquiries, contact the dataset author.**  

---

## 📧 **Author & Contact**
- **Author**: Chandan Singh  
- **Hugging Face Profile**: [Singhchandann](https://huggingface.co/Singhchandann)  
- **Intended Users**: Researchers, NLP practitioners, and AI developers working with **Marathi text data**.  
- **github Profile**: [Singhchandann](https://github.com/Singhchandann)  
- **Gmail**: [Send an email](mailto:[email protected])

For inquiries, please reach out via Hugging Face or Github or gmail.

---

## 📊 **Example Data Samples**
### **Example 1**
```text
anchor:
काल माझा कार्यक्रम यशस्वी झाला. आज नाही. का?

positive:
शब्दार्थ मार्गदर्शित प्रतिगमन चाचणी खर्चात कपात

negative:
गैरहजेरीच्या जप्तीमध्ये चेतनेची उपस्थिती
```

### **Example 2**
```text
anchor:
नैसर्गिक वातावरणातील कार्यात्मक नियर-इन्फ्रारेड स्पेक्ट्रोस्कोपी मोजमापांचा आढावा

positive:
प्रत्यक्ष डोळ्यांच्या संपर्कात असताना मेंदूच्या आत आणि त्यापलीकडे फ्रंटल टेम्पोरल आणि पॅरिएटल प्रणाली समक्रमित होतात.

negative:
एपिलेप्सीसाठी कॅनाबिनॉइड्सः आपल्याला काय माहित आहे आणि आपण कुठे जातो?
```

### **Example 3**
```text
anchor:
द मीडिया फ्रेम्स कॉर्पसः सर्व समस्यांच्या चौकटीवरील टिप्पण्या

positive:
रचनेचा सिद्धांत

negative:
आमोंग नायजेरियन प्रौढांच्या सल्लामसलतीसाठीच्या धोरणांचा गैरवापर करा
```

---