|
--- |
|
license: cc-by-nc-nd-4.0 |
|
metrics: |
|
- f1 |
|
- accuracy |
|
base_model: |
|
- FacebookAI/xlm-roberta-base |
|
language: |
|
- en |
|
- hi |
|
- mr |
|
- bn |
|
- ta |
|
- te |
|
- ml |
|
- ur |
|
pipeline_tag: text-classification |
|
tags: |
|
- sexism |
|
- hate |
|
- indic |
|
- empowerment |
|
- gender |
|
--- |
|
# Model Card for Model ID |
|
|
|
Classifies polarised gendered discourse for all indic languages. |
|
|
|
0=Neutral |
|
1=Sexist and misogynistic |
|
2=Empowering |
|
|
|
## Model Details |
|
|
|
genAMI, paper forthcoming |
|
|
|
## Author Details |
|
|
|
Praachi Kumar |
|
|
|
Research Fellow |
|
|
|
United Nations University - MERIT |
|
|
|
### Model Description |
|
|
|
|
|
- **Developed by:** Praachi Kumar |
|
- **Model type:** Fine-tuned XLM-RoBERTa base for sequence classification |
|
- **Language(s) (NLP):** Multi, focus on Indic |
|
- **License:** Non commercial, no derrivatives |
|
- **Paper:** Forthcoming |
|
|
|
## Uses |
|
|
|
Social science research, intended for academic and nonacademic use |
|
|
|
## Bias, Risks, and Limitations |
|
|
|
Social science approaches to annotation, single annotator coded |
|
|
|
### Recommendations |
|
|
|
Please contact me at [email protected] for instructions on further use |
|
|
|
## How to Get Started with the Model |
|
|
|
Forthcoming |
|
|
|
## Training Details |
|
|
|
### Training Data |
|
|
|
English language Tweets |
|
|
|
#### Metrics |
|
|
|
## English Tweets: |
|
|
|
Macro Average F1 Score: 0.83 |
|
|
|
Balanced Accuracy: 0.88 |
|
|
|
## Multilingual Tweets: |
|
|
|
Macro Average F1 Score: 0.76 |
|
|
|
Balanced Accuracy: 0.76 |
|
|
|
### Results |
|
|
|
Forthcoming |
|
|
|
## Citation |
|
|
|
**Model** |
|
|
|
**BibTeX:** |
|
|
|
@misc{genami2025, |
|
author = {Praachi Kumar}, |
|
title = {genAMI}, |
|
year = {2025}, |
|
month = {March}, |
|
day = {13}, |
|
howpublished = {\url{https://doi.org/10.57967/hf/5784}} |
|
} |
|
|
|
|
|
**APA:** Kumar, P. (2025). genAMI. Hugging Face. https://doi.org/10.57967/hf/5784 |
|
|
|
**Paper**: Forthcoming |