Skip to content

Latest commit

 

History

History
13 lines (11 loc) · 732 Bytes

README.md

File metadata and controls

13 lines (11 loc) · 732 Bytes
Raw

LexTermEval

Automatic legal terminology evaluation for South Tyrolean German

Given a test set of source-reference sentence pairs, respective MT hypothesis sentences and a MultiTermXML export file:

  • Termbase pre-processing
    • converting termbase (MultiTermXML export file) to python data structure for reference in terminology evaluation (xml2dict.py)
  • Dataset pre-processing
    • pre-processing the dataset and creating final testset file (create-testset.py)
  • Terminology evaluation
    • fine-grained automatic evaluation of legal terminology in MT output (LexTermEval.py)
  • Evaluating LexTermEval precision
    • creating a tab-separated file for manual evaluation of LexTermEval precision (precision_evaluation.py)