Extracting Relations, Events, and Time

Extracting Relations, Events, and Time

Growing an extractor from a few seeds

Bootstrapping a home-port extractor

Events, tenses, and timelines

A relation extractor built purely from hand-written lexico-syntactic patterns is run over a large corpus. Which combination of properties should you expect?

A bootstrapping relation extractor learns a pattern from a sentence about a vessel and then harvests the tuple $\langle \text{Pilot Cutter}, \text{Skarvholm} \rangle$ from it, which is wrong for the intended home-port relation. What is this failure called, and what is the standard guard?

An open relation extractor produces too many tuples to check by hand, so a sample of $42$ extracted tuples is drawn and $35$ of them turn out correct on inspection. Compute the estimated precision $\hat P$, rounding to three decimals.

Which statement correctly describes the distant supervision method for relation extraction?

Consider the sentence By the time the pilot boarded, the ferry had already cast off, spoken now. Under Reichenbach's analysis, which ordering of the event time E, reference time R and speech time U is correct?