基于冗余的仿真问答系统的轻量级局部文本分析

发布时间：2018-05-01 21:25

本文选题：问答系统 + 局部文本分析　；参考：《图书情报知识》2009年01期

【摘要】：近年来,基于冗余的仿真问答系统因其轻量性及高效性受到越来越多的关注。与分析法不同的是,冗余法不需要大量的语料库也能得到满意的检索结果。这些方法通常重视全文信息而忽视局部环境,而局部环境往往可能包含了大量关于自然语言问题和答案间语义及语法关系的信息。本文中,我们介绍了一种新的轻量级局部文本分析技术以及实现该技术的Askdragon问答系统。该系统只需向搜索引擎提交一个不太严谨或较为模糊的问题,然后选取小部分高质量的检索结果进行答案抽取即可。问答系统在保持较高精度的同时还显著提升了效率。文章最后使用TREC(Text Retrieval Conference)问题集和四个时间类问题集对该系统进行了评测,评测结果显示系统的性能可以匹敌甚至超越了目前最先进的问答系统,如ARANEA和QUANTA。
[Abstract]:In recent years, more and more attention has been paid to the redundant simulation question answering system because of its lightweight and high efficiency. Different from the analytical method, redundancy method can obtain satisfactory retrieval results without a large number of corpus. These methods usually attach importance to the full text information and ignore the local environment, which may contain a lot of information about the semantic and grammatical relations between natural language questions and answers. In this paper, we introduce a new lightweight local text analysis technology and Askdragon question answering system. The system only needs to submit a less precise or fuzzy question to the search engine, and then select a small number of high quality retrieval results to extract the answers. The Q & A system not only keeps high precision, but also improves the efficiency significantly. Finally, the system is evaluated using the TREC(Text Retrieval reference) problem set and four time class problem sets. The results show that the system can match or even surpass the most advanced question answering systems, such as ARANEA and Quanta.
【作者单位】：德雷塞尔大学信息科学与技术学院;武汉大学信息资源研究中心;
【分类号】：TP391.1

【相似文献】