基于概念语义相似度的长句查询扩展研究

发布时间：2019-05-28 23:01

【摘要】：随着Internet的飞速发展,网络信息检索也快速地发展起来。目前网络信息检索的主要形式是搜索引擎,它己是继电子邮件服务之后的网络第二大服务。当前的搜索引擎主要以用户输入的查询关键词进行信息的检索,但用户查询时输入的有限词语并不能完全准确表达其检索的真正意图,查询本身存在的歧义性导致搜索引擎返回大量与用户需求无关的文档,致使查全率和查准率较低。另一方面,用户有时往往会输入长句查询,提供给搜索引擎的查询词的长度正在不断增长,这也使得搜索引擎由于查询主题偏移等问题导致检索结果不理想。因此,为了解决上述问题,相关学者提出了查询扩展技术,即通过对原查询词进行修改来提高查询检索的准确率和召回率,也确实取得了一定的成果,但大多都是针对较短的查询进行的。而国外近几年正在加大对长句查询扩展技术方面的研究,这主要是由于使用自然语言-长句,能够更好的表达复杂和特定的信息需求,是用户未来查询表达的一个发展趋势,而且,随着查询扩展技术不断向语义查询方向发展,长句查询中包含的丰富语义关联为语义查询扩展的实现提供了更好的研究基础,对于了解用户复杂的语言特点和不同的语法习惯很有帮助。为此,本文针对在现有搜索引擎中长句查询主题偏移、准确率低以及低召回率时相关文档排序靠后等问题,提出了一种基于概念语义相似度的长句查询扩展方法。该方法首先利用AAlesk词义排歧方法确定查询词的正确词义,然后将该词义下WordNet同义词集中的相关语义概念加入到原查询中,通过对查询中不同概念之间进行语义相似度聚类得到查询聚类集,根据聚类集中整体的语义相似度水平和概念自身的语义相关重要性得到最佳候选概念集,最后从概念集中提取出得分最高的关键词来代表原始查询,从而实现对原长句查询的改进。另外,本文还应用了一种KeyGraph关键词抽取方法对原长句查询进行处理,并将这两种不同的长句查询改进结果放到三个不同类型的主流检索模型中分别进行检索实验。实验结果表明,改进后的长句查询检索效率都有了提高,特别是本文提出的长句查询扩展方法能够从语义层面更好的表达用户真实的信息需求,极大的提高了长句查询的准确率和召回率,比较适合在现有的主流语言检索模型上进行应用。
[Abstract]:With the rapid development of Internet, network information retrieval has also developed rapidly. At present, the main form of network information retrieval is search engine, which is the second largest service after email service. The current search engine mainly uses the query keywords entered by the user to retrieve the information, but the limited words entered by the user can not completely accurately express the real intention of the retrieval. The ambiguity of the query itself leads the search engine to return a large number of documents independent of user requirements, resulting in a low recall and precision. On the other hand, users sometimes enter long sentence queries, and the length of query words provided to search engines is increasing, which makes the retrieval results not ideal because of query topic offset and other problems. Therefore, in order to solve the above problems, the relevant scholars have proposed query expansion technology, that is, by modifying the original query words to improve the accuracy and recall rate of query retrieval, and have indeed achieved some results. However, most of them are aimed at short queries. In recent years, foreign countries are increasing the research on long sentence query expansion technology, which is mainly due to the use of natural language-long sentence, which can better express complex and specific information requirements, which is a development trend of user query expression in the future. Moreover, with the continuous development of query extension technology in the direction of semantic query, the rich semantic association contained in long sentence query provides a better research basis for the implementation of semantic query extension. It is very helpful to understand the complex language characteristics and different grammatical habits of users. Therefore, in order to solve the problems of long sentence query topic offset, low accuracy and low recall rate, a long sentence query extension method based on conceptual semantic similarity is proposed in this paper. In this method, the AAlesk word meaning disambiguation method is used to determine the correct word meaning of the query word, and then the related semantic concepts of the WordNet synonym set under the meaning of the word are added to the original query. The query clustering set is obtained by clustering the semantic similarity between different concepts in the query, and the best candidate concept set is obtained according to the overall semantic similarity level of the clustering set and the semantic correlation importance of the concept itself. Finally, the keywords with the highest score are extracted from the concept set to represent the original query, so as to improve the query of the original long sentence. In addition, a KeyGraph keyword extraction method is used to process the original long sentence query, and the improved results of these two different long sentence queries are put into three different types of mainstream retrieval models for retrieval experiments. The experimental results show that the improved long sentence query retrieval efficiency has been improved, especially the long sentence query extension method proposed in this paper can better express the real information needs of users from the semantic level. It greatly improves the accuracy and recall rate of long sentence query, and is more suitable for application in the existing mainstream language retrieval model.
【学位授予单位】：山东理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3

【相似文献】