Comments on: POTW 3/11/07: Discussion of “COGEX: A Logic Prover for Question Answering” http://www.paperoftheweek.com/2007/03/16/potw-31107-discussion-of-cogex-a-logic-prover-for-question-answering/ Read. Learn. Discuss. Mon, 18 Jun 2007 15:05:50 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Eric Yeh http://www.paperoftheweek.com/2007/03/16/potw-31107-discussion-of-cogex-a-logic-prover-for-question-answering/comment-page-1/#comment-509 Eric Yeh Wed, 04 Apr 2007 16:23:27 +0000 http://www.paperoftheweek.com/2007/03/16/potw-31107-discussion-of-cogex-a-logic-prover-for-question-answering/#comment-509 I haven't given the paper as thorough a reading as I should've, so this might be completely off base, but I didn't see anything that dealt directly with coreference issues, such as pronomial anaphora. I get the sense that the authors may be pushing this to the proof phase, instead of treating it as a separate annotation issue, which would congruent with what Jerry Hobbs was arguing for in "Ontological Promiscuity," where he argued for minimal commitments when logically representing text. What's also interesting is to compare this with the work LCC has done in the textual entailment task, where they (apparently) did not make use of the theorem prover. I believe they had a recent paper outlining how their textual entailment work applies to the question answering task, which may be interesting to look at. I haven’t given the paper as thorough a reading as I should’ve, so this might be completely off base, but I didn’t see anything that dealt directly with coreference issues, such as pronomial anaphora. I get the sense that the authors may be pushing this to the proof phase, instead of treating it as a separate annotation issue, which would congruent with what Jerry Hobbs was arguing for in “Ontological Promiscuity,” where he argued for minimal commitments when logically representing text.

What’s also interesting is to compare this with the work LCC has done in the textual entailment task, where they (apparently) did not make use of the theorem prover. I believe they had a recent paper outlining how their textual entailment work applies to the question answering task, which may be interesting to look at.

]]>