

























This paper presents an original methodology to consider question answering. We noticed that query expansion is often incorrect because of a bad understanding of the question. But the automatic good understanding of an utterance is linked to the context length, and the question are often short. This methodology proposes to analyse the documents and to construct an informative structure from the results of the analysis and from a semantic text expansion. The linguistic analysis identifies words (tokenization and morphological analysis), links between words (syntactic analysis) and word sense (semantic disambiguation). The text expansion adds to each word the synonyms matching its sense and replaces the words in the utterances by derivatives, modifying the syntactic schema if necessary. In this way, whatever enrichment may be, the text keeps the same meaning, but each piece of information matches many realisations. The questioning method consists in constructing a local informative structure without enrichment, and matches it with the documentary structure. If a sentence in the informative structure matches the question structure, this sentence is the answer to the question.
此内容由惯性聚合(RSS阅读器)自动聚合整理,仅供阅读参考。 原文来自 — 版权归原作者所有。