Whamit!

The Weekly Newsletter of MIT Linguistics

Ling-Lunch 3/15 - Michael Yoshitaka Erlewine

Title: The Constituency of Hyperlinks in a Hypertext Corpus
Speaker: Michael Yoshitaka Erlewine (MIT)
Date/Time: Thursday, Mar 15, 12:30-1:45p
Location: 32-D461

While many have advocated for the use of Internet corpora in traditional corpus linguistics, no previous work has used one of the greatest characteristics of hypermedia—inline hyperlinks—as a tool in the generative study of syntax. In this talk I argue that the hyperlink-authoring behavior of naive speakers reflects the underlying syntactic constituency of sentences. I show that inline hyperlinks show a highly statistically significant tendency to be constituents in their host sentences (96% in our preliminary sample). In addition, hyperlinking behavior is sensitive to structural distinctions, distinguishing between complement and adjunct PPs. Finally, hyperlinks identified as non-constituents will be discussed as a unique natural class of nonconstituent linguistic structure.