hierarchical-taxonomy-clustering

Build a unified category taxonomy by using embedding-based recursive clustering on hierarchical product category paths. This process automates grouping similar category paths and generates an N-level hierarchy for e-commerce datasets.

Embedding-based recursive clustering works by converting hierarchical category paths into text embeddings, then recursively grouping them into a multi-level taxonomy. It also uses weighted word frequency analysis to generate meaningful category names for each cluster.

Yes, you need sentence-transformers for generating text embeddings and scipy for clustering operations. The pipeline also requires pandas, numpy, nltk, and tqdm to process multiple e-commerce datasets, remove duplicates, and clean text.

The best way to clean e-commerce datasets is to remove duplicates and process text data before clustering. This data preparation step ensures that the embedding-based recursive clustering generates a unified multi-level taxonomy accurately.

Yes, you can automatically generate meaningful category names from hierarchical product paths. The pipeline uses weighted word frequency analysis on the clustered category texts to derive representative names for the unified multi-level taxonomy.