relatio-nlp / relatio

code base for constructing narrative statements from text
MIT License
98 stars 27 forks source link

Use clustered proper noun phrases as alternative to NER #69

Open elliottash opened 2 years ago

elliottash commented 2 years ago

See here, "The President" is identified as a proper noun phrase: https://explosion.ai/demos/displacy?text=The%20President%20voted%20for%20schools.&model=en_core_web_sm&cpu=1&cph=1

As an alternative to NER entity tagging, run two parallel noun phrase clusterings:

one set of clusters for proper nouns, and another set of clusters for common nouns (based on POS tag).

this will work best after co-reference resolution.