This project explores unsupervised topic modelling within large text corpora, applying techniques such as Latent Dirichlet Allocation (LDA) and Latent Semantic Analysis (LSA) to extract latent themes and semantic structures from unstructured textual data. The workflow includes data cleaning, tokenization, stop-word removal, lemmatization, and construction of document-term matrices, followed by application of topic modelling algorithms and interpretation of resulting themes. The outputs provide insights into underlying patterns within the text, which can be leveraged for document clustering, trend analysis and content organisation.

This project explores how unsupervised learning can uncover hidden structures within large volumes of text. The workflow begins with loading and cleaning textual data — removing punctuation, stop words, and irrelevant symbols, followed by tokenization and lemmatization using NLTK. The cleaned text is converted into numerical representations using Term Frequency-Inverse Document Frequency (TF-IDF) and Count Vectorizer, preparing it for topic modeling. Two algorithms — Latent Dirichlet Allocation (LDA) and Latent Semantic Analysis (LSA) is what I applied to extract underlying topics. Each algorithm analyzes the word distributions and clusters terms that frequently occur together. Each document is then scored across these topics, visualized through word clouds and interactive charts, enabling easy interpretation of dominant themes and relationships. Through this process, the model turns raw, unstructured text into actionable insights helping researchers, analysts, and organizations understand the underlying insights.