Datasets:

taln-ls2n
/

silk

 - natural-language-processing
 size_categories:
 - 1K<n<10K
+---
+This dataset contains the data files generated by silk, a method that leverages citation contexts to create synthetic samples of documents paired with silver-standard keyphrases for adapting keyphrase generation models to new domains.
+We applied silk on three domains --namely, Natural Language Processing (nlp), Astrophysics (astro)and Paleontology (paleo)--, thereby creating new adaptation data for each domain.
+This dataset also includes three human-labeled test sets to assess the performance of keyphrase generation across these domains.