Description

The Text Clustering repository by HuggingFace serves as a minimal codebase for clustering texts. It contains a pipeline that uses existing standard methods such as Sentence Transformers, UMAP, Faiss, Plotly, Matplotlib, and Scikit-learn. The pipeline consists of several distinct blocks that can be customized and run in a few minutes on a consumer laptop. This repository also provides an example of clustering and topic labeling applied to the AutoMathText dataset, utilizing Cosmopedia's web labeling approach.