List resources to release and where

[mcollardanuy]

To do:

• Upload NER model to HuggingFace: Upload our NER models to HuggingFace #193
• Upload DeezyMatch training set to BL repo: Upload DeezyMatch OCR training set to the BL repo #242
• Upload NER- and ED- formatted topres19th data to BL repo: Upload NER and ED formatted topres19th data #243
• Decide where to upload wikipedia/wikidata resources: Upload Wiki-based resources to BL repo #259

List of resources to release (currently in toponymVM2.0):

To run T-Res for the current configuration

resources/models/blb_lwm-ner-fine.model/
resources/wikidata/wikidata_gazetteer.csv
resources/wikidata/entity2class.txt
resources/wikidata/mentions_to_wikidata.json
resources/wikidata/mentions_to_wikidata_normalized.json
resources/wikidata/wikidata_to_mentions_normalized.json
resources/deezymatch/combined/wkdtalts_w2v_ocr/
resources/deezymatch/models/w2v_ocr/
resources/rel_db/embeddings_database.db
resources/models/disambiguation/deezymatch+3+25_Ashton1860+wmtops

Also, resources/models/blb_lwm-ner-fine.model/ unless it's uploaded to HF (see issue).

To train a new fine-grained NER model (with topRes19th)

experiments/outputs/data/lwm/ner_fine_train.json
experiments/outputs/data/lwm/ner_fine_dev.json

To train a new DeezyMatch model (for English OCR)

resources/deezymatch/data/w2v_ocr_pairs.txt

To generate a new DeezyMatch training set (for English OCR)

Nothing, embeddings already uploaded here, store them under resources/models/w2v/ in folders named w2v_18XXs_news, where XX is the decade.

To train a new entity disambiguation model (with topRes19th)

experiments/outputs/data/lwm/linking_df_split.tsv