Using A Named Entity Tagger To Generalise Surface Matching使用命名实体标注概括表面匹配.pptVIP

April 14th 2003 NLP4QA an EACL 2003 Workshop Using a Named Entity Tagger to Generalise Surface Matching Text Patterns for Question Answering Mark A. Greenwood and Robert Gaizauskas Natural Language Processing Group Department of Computer Science University of Sheffield, UK Outline of Presentation Introduction and overview The basic approach Limitations of the approach Generalising using an NE tagger Results to date Discussion and Conclusions Future Work Introduction and Overview The basic question answering system outlined in this talk is similar to those developed by Soubbotin and Soubbotin (2001) and Ravichandran and Hovy (2002). These systems use a single resource consisting of a large number of surface matching text patterns. These patterns are derived using simple machine learning techniques from a set of question-answer pairs and a corpus of documents in which the answer is known to occur. While agreeing that these systems have enormous potential we aim to highlight, and provided a solution to, a problem that can be caused by overly specific patterns: They mainly result in systems which may be unable to answer specific types of questions. The Basic Approach Learning patterns which can be used to find answers involves a two stage process: The first stage is to learn a set of patterns from a set of question-answer pairs. The second stage involves assigning a precision to each pattern and discarding those patterns which are tied to a specific question-answer pair. To explain the process we will use questions of the form “When was X born?”: As a concrete example we will use “When was Mozart born?”. For which the question-answer pair is: Mozart 1756 The Basic Approach The first stage is to learn a set of patterns from the question-answer pairs for a specific question type: For each example the question and answer terms are submitted to Google and the top ten documents are downloaded. Each document then has the question and answer terms replaced by AnCHoR and AnSW