emiltj / DANSK-gold-NER

emiltj/DANSK-gold-NER Issues

Remove merge_inter_with_annotator script after merging in db using the raters_to_db.sh script
Closed 2 years ago
Ensure that each raters data contains no unnecessary duplicates
Closed 2 years ago
Find out what to do with duplicates
Closed 2 years ago
ensure that there are no duplicate docs within rater
Closed 2 years ago
Inkludér GPE
Closed 2 years ago
Fjern all language og product tags (da kvaliteten er for ringe - efter dialog med Kenneth)
Closed 2 years ago
Change the thresholds on the basis of the discussion with Kenneth
Closed 2 years ago
NER_interannotator_annotator3.jsonl files should be merged with NER_annotator3.jsonl
Closed 2 years ago
Ensure that README contains guidelines for the review process
Closed 2 years ago1
In README, no accurate description of the streamline multi has been written yet.
Closed 2 years ago
Raters 7 and 8 (and others(!)) appear multiple times
Closed 2 years ago
The removal of rater 2 and 10 should be applied before the split. multi now contains single documents.
Closed 2 years ago
Ensure that the thresholds are both good and meaningful for multiple_streamline
Closed 2 years ago
See that the infrequent ents and frequent ents removal and addition works as wanted
Closed 2 years ago
Rater8 is being filtered away in streamline_docs
Closed 2 years ago
nlp = dacy.load("medium") for getting vocab yields very slow findings of unique_docs
Closed 2 years ago
Multi_streamline is extremely expensive to run
Closed 2 years ago
The scripts have not been updated to the readme.md description
Closed 2 years ago
streamline_multi wrongly uses a single threshold for finding frequent ents
Closed 2 years ago1
Repo right now implements the old way of converting from .spacy to .jsonl
Closed 2 years ago
Still needs to implement a .py script for converting .spacy to .jsonl
Closed 2 years ago
streamline_multi should be a .py script
Closed 2 years ago
Output of streamline_multi doesn't work with Prodigy
Closed 2 years ago5