tomotopy

Why we included this project

Tomotopy is a Python wrapper around a C++ Gibbs-sampling engine, which is what makes it fast enough to train topic models on large text collections. It covers a broad set of formulations: plain LDA, labeled and supervised variants, hierarchical Dirichlet processes, dynamic topic models, and Pachinko allocation, so you can match the model to your data instead of being locked into one approach. The SIMD-accelerated core trains noticeably faster than pure-Python libraries, which helps when you are iterating over millions of documents. Trained models can be saved and reloaded, and there is an interactive browser-based viewer for inspecting topic-word distributions, so the library works both for exploratory analysis and for dropping a trained model into a pipeline. Teams doing document clustering, content tagging, or corpus exploration get a self-contained tool without much setup.