v2.0 system for definition extractor

Develop v2.0 system for doc-level user interfaces and doc-level evaluation

(See dykang/definition_extractor_v2.0)

• Model ensemble
  • Development/Testing in the internal NLP repo (+2.0 F1)
  • Deployment to frontend system (not sure to do this or not, because of scalability)
• [ x] Fine-tuning roberta-large with S2ORC (with @kyleclo ) (deprecated due to low performance)
  • Initial development in the internal NLP repo
  • Scale-up for the entire S2ORC dataset: data split and caching
  • Deployment to frontend system
• Categorization of term types: (with @andrewhead )
  • Types:
    • Symbols,
    • Protologisms (i.e., new words defined in the paper)
    • Abbreviations
  • Entities for each type:
    • Symbols: nicknames, definitions
    • Protologisms: definitions
    • Abbreviations: expansions
  • [x ] Extend the acronym/nickname detector to non-definitional sentences
• Add document-level features (e.g., term position, document frequency)
  • Added position_ratio for relative position of a term in the entire document
  • Added section_name for the name of section that the term appears
• Add cross-document-level features (e.g., term position, document frequency, cross-distance)
  • Make a dictionary of previously detected terms and symbols
• within-document term grouping based on their definitions (deprecated)
  • Add coreference links
  • Analysis between term occurrences and detected term definitions
• add confidence score
  • Development/Testing in the internal NLP repo
  • Deployment to frontend system
• fix BERT tokenization error

[dykang]

Closing this issue as we approach merging #146 to master. The remaining features mentioned here will be merged to v3 system or deprecated due to the lack of usefulness.

[dykang]

Target papers for general user study

• 1804.08199, 2002.04138, 1909.13433
• 1704.05572, 1609.05143, 1611.01603

Target papers for comparison study

• ?