How AI describes TreeSeg

TreeSeg is an unsupervised algorithm by AugmendTech for hierarchical topic segmentation of large meeting transcripts, using embeddings and divisive clustering. It produces a binary tree of segments by recursively splitting segments into sub-segments. The repository includes code, baselines, datasets (e.g., TinyRec, ICSI/AMI), and tooling for extracting utterances and ground-truth segmentations, with OpenAI embeddings (and other embeddings supported).