This research delves into the nuanced analysis of parliamentary discourse by focusing on two central tasks: identifying the political ideology of speakers and categorizing their party affiliation. We leveraged machine learning models including Linear SVC, Logistic Regression, and DistilBERT to address these classification challenges using a multilingual corpus from ParlaMint.

Our findings reveal that Linear SVC outperforms other models, demonstrating effectiveness in interpreting complex political texts. This research underscores the value of integrating both traditional and advanced machine learning methods for political discourse analysis.

We developed a model to identify disaster-related tweets on Twitter, crucial for emergency response organizations. The model distinguishes genuine disaster announcements from regular tweets using advanced NLP techniques and machine learning algorithms trained on a diverse dataset.

The system achieved an accuracy of 0.783, effectively minimizing false positives/negatives. This approach enhances the efficiency of disaster monitoring systems by reliably filtering crucial information from real-time social media data.

Developed two models for Trauma Detection App: one to predict if a person has trauma based on their conversations, and another to classify the type of trauma. These models utilize advanced NLP techniques and psychological markers to provide accurate assessments.