pytextrank

Why we included this project

PyTextRank brings the TextRank family of graph algorithms into spaCy as a pipeline extension, so if you are already doing tokenization and parsing you can pull the top-ranked phrases out of a document in a few lines of code and without spinning up a large language model. That makes it a low-cost route to extractive summarization, topic indexing, and cleaner concepts for a knowledge graph. Since it runs deterministically on plain Python and CPU, it stays fast and reproducible when you are processing text at scale. The PositionRank, TopicRank, and Biased TextRank variants also give you room to tune results toward different kinds of documents.