A multilingual analysis system designed for sentiment classification across four different languages: English
, Hindi
, Marathi
, and Gujarati
.
- Input: Any sentence in
English
,Hindi
,Marathi
, andGujarati
. - Output: Sentiment Class (
POSITIVE
orNEGATIVE
) - The system features an intuitive GUI for seamless web interface interaction.
The model has undergone fine-tuning on target language data, facilitated by the Google Translation API.
Dataset: Sentiment140
Two distinct model architectures have been evaluated:
- DistilBERT
- mBERT
Model | Class | Precision | Recall | F1-Score |
---|---|---|---|---|
DistilBERT | 0 | 0.77 | 0.82 | 0.80 |
1 | 0.81 | 0.76 | 0.78 | |
mBERT | 0 | 0.71 | 0.74 | 0.73 |
1 | 0.74 | 0.71 | 0.72 |