IBM's PIQUANT II in TREC 2004
نویسندگان
چکیده
PIQUANT II, the system we used for TREC 2004, is a completely reengineered system whose core functionalities for answering factoid and list questions remain largely unchanged from previous years [Chu-Carroll et al, 2003, Prager et al, 2004]. We continue to address these questions using our multi-strategy and multi-source approach. For “other” questions, we experimented with two alternative approaches, one that uses statistical collocation information for extracting prominent passages related to the target, and the other which is a slight variation of the QA-by-Dossier approach we employed last year [Prager et al, 2004] that asks a set of subquestions of interest about the target and returns a set of relevant passages that answer these subquestions. In addition, to address this year’s new question format, we developed a question preprocessing component to interpret individual questions against the given target to generate a selfcontained natural language question as expected by subsequent components of our QA system. NIST assessed scores showed substantial improvement of our new PIQUANT II system over earlier versions of our QA system both in terms of absolute scores as well as relative improvement compared to the best and median scores in each of the three component subtasks.
منابع مشابه
IBM's PIQUANT II in TREC 2005
Introduction This year, the PIQUANT II system we used in the TREC QA track is an improved version over the reengineered system we used in last year's entry [Chu-Carroll et al., 2005]. Our system adopts a multi-agent approach to question answering. In this framework, a question is submitted to multiple agents, each adopting a different question answering strategy and/or consults a different info...
متن کاملIBM's PIQUANT in TREC2003
For the most part, the system we used for TREC2003 was a smooth evolution of the one we ran in TREC2002 [Chu-Carroll et al, 2003b]. We continued to use our multi-source and multi-agent architecture. For Factoid questions we used all of our previous answering agents with an additional pattern-based agent, an enhanced answer resolution algorithm, and increased coverage of the Cyc sanity checker. ...
متن کاملIBM's Statistical Question Answering System - TREC-10
In this paper, we document our efforts to extend our statistical question answering system for TREC-11. We incorporated a web search feature, and novel extensions of statistical machine translation as well as extracting lexical patterns for exact answers from a supervised corpus. Without modification to our base set of thirty-one categories, we were able to achieve a confidence weighted score o...
متن کاملAnswerFinder at TREC 2004
AnswerFinder combines lexical, syntactic, and semantic information in various stages of the question answering process. The candidate sentences are preselected on the basis of (i) the presence of named entity types compatible with the expected answer type, and (ii) a score combination of the overlap of words, grammatical relations, and flat logical forms. The candidate answers, in turn, are ext...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2004