Appendix: Suggested evaluation datasets: SemEval WSD, OntoNotes, OpenSubtitles (for colloquial usage), Common Crawl-derived corpora for frequency and trend detection.
