Normalize entities and relations from annotated data with Wikidata

[apiad]

In the data/output folder, two files have been created:

• entities.tsv contains all keyphrases annotated with their corresponding label.
• relations.tsv contains all relation triplets.

Both files are TSV (tab-separated values), so there should no problems with ,, ", etc. Simply opening each file and splitting by \t should do.

The idea would be trying to normalize these mentions with their appearances in Wikidata. For that I would propose creating another two files (data/output/(entities|relations)-normalization.tsv for which all the matches found are logged together with their corresponding Wikidata metadata (i.e., IDs, etc.)

[csisc]

I developed Python codes to solve this issue:

• https://github.com/knowledge-learning/cord19-ann/blob/master/data/output/wdreconciliation/entities.py
• https://github.com/knowledge-learning/cord19-ann/blob/master/data/output/wdreconciliation/wikidataann.py
  The outputs are in XLSX format