Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix small grammatical errors with coref_resolved #332

Open
wants to merge 3 commits into
base: master
Choose a base branch
from

Conversation

sam2k13
Copy link

@sam2k13 sam2k13 commented Jan 9, 2022

Happy to get feedback, but this handles 2 issues I found reducing the quality of the coref_resolved property on the output.

  • Don't replace Coref spans with PRONOMINAL spans.

    • It appears "PRONOMINAL" spans (i.e. He, She, They) are contextually lacking and replacing a span with a PRONOMINAL does not improve quality
    • Original Sentence: He likes to eat chocolate. His favorite chocolate is Hershey's.
    • Previously: He likes to eat chocolate. He favorite chocolate is Hershey's.
    • New: He likes to eat chocolate. His favorite chocolate is Hershey's.
  • Don't replace spans including a "Determiner" POS tag (DET)

    • Possessive Determiner's are words like: My, your, his, her, their
    • Without resolving punctuation, replacing a "Determiner" with a pronoun can lead to grammatically incorrect sentences. Also, replacing a Determiner in the middle of a sentence often degrades its quality.
      • Example: Joe likes to bring his boat to the lake.
      • Previously: Joe likes to bring Joe boat to the lake.
      • New: Joe likes to bring his boat to the lake.

If any of these changes go against the grain of the desired Coreference resolution, I am open to changes or no need to merge this PR. For my use case, having the resolved Coref text stay high quality is very important.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

1 participant