Add pnode filter, use spacy tokenizer for calculating spans - Githubissues

isi-vista / cdse-covid

Claim detection & semantic extraction (Covid-19 domain)

0 stars 0 forks source link

Add pnode filter, use spacy tokenizer for calculating spans #139

Closed elizlee closed 2 years ago

elizlee commented 2 years ago

Addresses the following issues:

"Pnode" (Wikidata property) selections instead of Qnodes; e.g. selecting P2047 for "duration" instead of Q2199864
Making it easier to find spans by using the spacy tokenizer on claims/sentences instead of simple splits
Some type values in the CS file were followed by a number; left those out during ingestion
Some arguments still had trailing/preceding stop words; applied a function to trim those

Before running the pipeline with these changes, remove the files from your kgtk_event_cache and kgtk_refvar_cache.