DTECT is an end-to-end system for solving the challenge of discovering changing topics and trends in explosively growing text data. It complements the fragmented pipelines and lack of interpretability and user-friendly exploration of existing dynamic topic modeling techniques, providing an integrated workflow that supports data preprocessing, various model architectures, and evaluation metrics for analyzing the quality of temporal topic models. It significantly enhances interpretability through LLM-based automatic topic labeling, trend analysis using temporally salient words, interactive visualizations with document-level summaries, and a natural language chat interface for intuitive data querying. By integrating these capabilities into a single platform, it enables users to effectively track and understand topic dynamics. DTECT is open source and available on GitHub.
