Computational Linguistics Colloquium

Thursday, 18 December, 16:15
Conference Room, Building C7 4

Multilingual Grammar Engineering with the Grammar Matrix: Crosslinguistic Hypothesis Testing

Though linguists strive for elegance in our analyses, the fact remains that the natural objects we study (languages) are highly complex. Grammar engineering allows us to create models of the objects under study that can be manipulated by machines, and therefore verified against a much broader range of data than would ever be possible by hand. Furthermore, by implementing models and running regression tests, we can ensure that each new addition to the grammar is consistent with the previously implemented analyses, and/or update analyses as needed.
The Grammar Matrix project promotes this style of linguistic hypothesis testing at two levels. In the first instance, it lowers the bar to entry, making it easier for linguist/users to create precision grammars and thus use them for hypothesis testing. At another level, however, it is also an exercise in crosslinguistic hypothesis testing, exploring the extent to which one and the same core grammar can be shared across typologically diverse languages.
In this talk, I will describe the architecture of the Grammar Matrix and its system of "libraries" (analyses of crosslinguistically variable phenomena), and reflect on what we have learned about how crosslinguistic variation constrains what can be done in the core grammar.

If you would like to meet with the speaker, please contact Antske Fokkens.