Add part of spech-aware processing #8

LoicGrobol · 2025-01-17T16:22:58Z

allow substracting the average embedding of the POS of the word (pre-clustering)
- Inferring POS from a lexicon (typically SUBTLEX)
- Using POS-tagging from SpaCy (make this optional to avoid a forced spacy dep)
doc
tests

This is an optional extra step/subcommand to avoid having it take over the otherwise straigtforward embedding and clustering code.

…use it in clamp

LoicGrobol and others added 5 commits January 17, 2025 16:53

allow preserving embeddings in clustering outputs

2f27b64

removing umap from dependencies for now since at this point we don't …

1dcde3b

…use it in clamp

bumping freezes

21cf02f

add the spacy extra for pos

6a87a69

🎨

8f329ab

Provide feedback