Hugging Face's logo

poltextlab
/
xlm-roberta-large-pooled-cap-media2-v2

Text Classification
Transformers
Safetensors
PyTorch
English
xlm-roberta
zero-shot-classification
Eval Results
Model card Files Community

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Our models are intended for academic use only. If you are not affiliated with an academic institution, please provide a rationale for using our models. Please allow us a few business days to manually review subscriptions.

Log in or Sign Up to review the conditions and access this model content.

xlm-roberta-large-pooled-cap-media2-v2

Model description

An xlm-roberta-large model finetuned on english training data labelled with major topic codes from the Comparative Agendas Project. Furthermore we used the following 18 media codes:

  • State and Local Government Administration (24)
  • Weather (25)
  • Fires, emergencies and natural disasters (26)
  • Crime and trials (27)
  • Arts, culture, entertainment and history (28)
  • Style and fashion (29)
  • Food (30)
  • Travel (31)
  • Wellbeing and learning (32)
  • Personal finance and real estate (33)
  • Personal technology and popular science (34)
  • Churches and Religion (35)
  • Celebrities and human interest (36)
  • Obituaries and death notices (37)
  • Sports (38)
  • Crosswords, puzzles, comics (39)
  • Media production/internal, letters (40)
  • Advertisements (41)

How to use the model 

from transformers import AutoTokenizer, pipeline

tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-large")
pipe = pipeline(
    model="poltextlab/xlm-roberta-large-pooled-cap-media2-v2",
    task="text-classification",
    tokenizer=tokenizer,
    token="<your_hf_read_only_token>"
)

text = "Tapenade, when combined with chopped tomatoes and basil, makes for one of the best simple summer dishes in existence."
pipe(text)

Gated access

Due to the gated access, you must pass the token parameter when loading the model. In earlier versions of the Transformers package, you may need to use the use_auth_token parameter instead.

Model performance

The model was evaluated on a test set of 67,681 english examples.

  • Accuracy: 0.8.
  • Precision: 0.8.
  • Recall: 0.8
  • Weighted Average F1-score: 0.8
Label precision recall f1-score support
Macroeconomics (1) 0.72 0.73 0.72 2471
Civil Rights (2) 0.7 0.68 0.69 1886
Health (3) 0.82 0.83 0.83 2471
Agriculture (4) 0.77 0.78 0.77 811
Labor (5) 0.74 0.7 0.72 1277
Education (6) 0.86 0.88 0.87 2080
Environment (7) 0.77 0.8 0.78 1283
Energy (8) 0.8 0.83 0.81 1370
Immigration (9) 0.73 0.79 0.76 514
Transportation (10) 0.81 0.81 0.81 2375
Law and Crime (12) 0.71 0.69 0.7 2471
Social Welfare (13) 0.66 0.67 0.67 683
Housing (14) 0.72 0.72 0.72 1023
Banking, Finance, and Domestic Commerce (15) 0.72 0.69 0.71 2471
Defense (16) 0.76 0.78 0.77 2471
Technology (17) 0.73 0.75 0.74 1375
Foreign Trade (18) 0.7 0.66 0.68 533
International Affairs (19) 0.69 0.67 0.68 2471
Government Operations (20) 0.72 0.66 0.69 2507
Public Lands (21) 0.64 0.66 0.65 554
Culture (23) 0.74 0.76 0.75 2142
State and Local Government Administration (24) 0.6 0.67 0.63 81
Weather (25) 0.8 0.72 0.76 39
Fires, emergencies and natural disasters (26) 0.96 0.97 0.97 2471
Crime and trials (27) 0.69 0.88 0.78 293
Arts, culture, entertainment and history (28) 0.78 0.77 0.77 2265
Style and fashion (29) 0.81 0.69 0.74 2389
Food (30) 0.79 0.84 0.81 2168
Travel (31) 0.8 0.87 0.83 1999
Wellbeing and learning (32) 0.75 0.78 0.77 1963
Personal finance and real estate (33) 0.86 0.85 0.86 2222
Personal technology and popular science (34) 0.81 0.81 0.81 2054
Churches and Religion (35) 0.7 0.62 0.66 74
Celebrities and human interest (36) 0.69 0.75 0.72 1109
Obituaries and death notices (37) 0.74 0.86 0.79 521
Sports (38) 0.92 0.92 0.92 2220
Crosswords, puzzles, comics (39) 0 0 0 7
Media production/internal, letters (40) 0.91 0.92 0.91 763
Advertisements (41) 1 0.99 0.99 7333
No Policy and No Media Content (998) 0.83 0.78 0.8 2471

Inference platform

This model is used by the CAP Babel Machine, an open-source and free natural language processing tool, designed to simplify and speed up projects for comparative research.

Cooperation

Model performance can be significantly improved by extending our training sets. We appreciate every submission of CAP-coded corpora (of any domain and language) at poltextlab{at}poltextlab{dot}com or by using the CAP Babel Machine.

Debugging and issues

This architecture uses the sentencepiece tokenizer. In order to run the model before transformers==4.27 you need to install it manually.

If you encounter a RuntimeError when loading the model using the from_pretrained() method, adding ignore_mismatched_sizes=True should solve the issue.

Downloads last month
2,928
Safetensors
Model size
560M params
Tensor type
F32
·
Inference Providers NEW
Text Classification
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for poltextlab/xlm-roberta-large-pooled-cap-media2-v2

Base model

FacebookAI/xlm-roberta-large
Finetuned
(448)
this model

Space using poltextlab/xlm-roberta-large-pooled-cap-media2-v2 1

Collection including poltextlab/xlm-roberta-large-pooled-cap-media2-v2

Evaluation results

View on Papers With Code