AutomaticAnnotator implementations should save their training data

comphist / cora

A web-based, token-level annotation tool for non-standard language data

http://www.linguistics.rub.de/comphist/resources/cora/

MIT License

10 stars 6 forks source link

AutomaticAnnotator implementations should save their training data #74

Open mbollmann opened 8 years ago

mbollmann commented 8 years ago

Originally reported by: fpetran (Bitbucket: fpetran, GitHub: fpetran)

The AutomaticAnnotators may need to regenerate their parameter files when they are updated or recompiled. For such a case we should have record of the training data for each parameter files. Since the AutomaticAnnotators are external tools, they need to generate some input file anyway.

Bitbucket: https://bitbucket.org/mbollmann/cora/issue/71

mbollmann commented 8 years ago

Original comment by Marcel Bollmann (Bitbucket: mbollmann, GitHub: mbollmann):

Is that really a CorA issue, though?

Parameter files that were generated by CorA can simply be re-generated by each user with the click of a button, and the "fixed" parameter files (e.g., Tiger/TüBa) are generated outside of CorA.