Full-corpus SemCor relabeling with GPT-5.5 and Gemma, controlled BEM/ESCHER/ConSeC retraining, and the Glite LENS bi-encoder - reproducibility record for the lexEn/SenseBench WSD paper
-
Updated
Jul 18, 2026 - Python
Full-corpus SemCor relabeling with GPT-5.5 and Gemma, controlled BEM/ESCHER/ConSeC retraining, and the Glite LENS bi-encoder - reproducibility record for the lexEn/SenseBench WSD paper
To associate your repository with the semcor topic, visit your repo's landing page and select "manage topics."