Version 4.2

Lecture: Simulating Lexical Semantic Change from Sense-Annotated Data

Evaluation in automatic lexical change detection is a notorious challenge. Existing test sets have the downside of being very small and hence do not allow for generalization of the results obtained on them. Artificial data, on the other hand, has the downside of relying on various assumptions that may or may not be correct. The large advantage of artificial data is, though, that it allows the precise controlling of potentially influencing variables such as frequency or polysemy. We present a new way to simulate lexical semantic change with little assumptions from synchronic sense-annotated data and demonstrate its usefulness for evaluation of lexical semantic change detection models.

Info

Day: 2019-05-25
Start time: 14:00
Duration: 00:30
Room: 100 / Hörsaal V
Track: Computational Linguistics
Language: en

Links:

Files

Feedback

Click here to let us know how you liked this event.

Concurrent Events