Fix streamlit accuracy that suppose to tally with the actual model accuracy

feat: add bilingual support with XLM-RoBERTa and classical bilingual models

add colab notebook, distilbert, xai, inference module

checkpoint 2: classical models, streamlit app, visualisations