Sentiment Analysis
—Classification of informal social media text
——— Work in Progress ———
This project explores how natural language processing can be used to recognise sarcastic intent in informal text. By training neural network models on a dataset of labelled Reddit comments, the goal is to distinguish between sarcastic and sincere statements. In future iterations the scope will expand to general sentiment analysis and even multimodal input (e.g. combining images and text).
Project Overview
Goal- Build models that can classify text based on sentiment.
- Explore the effectiveness of different embeddings and neural architectures for sentiment tasks.
- Develop a pipeline that can extend to sarcasm detection and more nuanced emotional analysis.
- Implemented preprocessing and feature extraction using GloVe embeddings.
- Trained and evaluated FNN on labeled sentiment datasets.
- Expand experiments to include transformer-based architectures such as BERT.
- Extend project to handle multi-modal input.
- Create a browser app that analyzes webpage in real time.