面向查询的XML文本自动文摘研究

项目名称： 面向查询的XML文本自动文摘研究

项目编号： No.60803105

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 金属学与金属工艺

项目作者： 刘德喜

作者单位： 江西财经大学

项目金额： 19万元

中文摘要： 对信息检索结果进行自动摘要可以提高用户获取信息的效率。本项目围绕面向查询的XML文本摘要开展研究，取得的主要研究成果有：①#25552;出了以内容查询为主、结构查询为辅、不依赖于结构约束来查找答案结点并评价的方法；②#25552;出了面向用户查询意图的检索结果聚类方法；③#26500;建了3000余篇面向查询的XML文本摘要语料库；④#25552;出了基于三类特征的XML摘要句抽取方法及基于PageRank的摘要句重排方法；⑤#25552;出了基于主题概括强度的结点(标签/路径)语义权重模型；⑥#23558;结点语义权重用于改进BM25模型，提出了基于BM25SF的XML片段检索策略；⑦#21442;加了国际INEX 2011的片段检索评测，并取得了第二名的优异成绩；⑧#25552;出了不确定数据库上两种新的Top-k查询语义和快速搜索算法。获得的重要结论有：①#23545;XML查询结果的聚类和摘要要充分考虑用户的信息需求；②ML结构信息和结点的语义信息对提高XML检索和摘要的质量很有帮助；③#21033;用结点语义权重进行XML片段检索有助于用户快速了解原文档的相关性； ④ML摘要内容的重排需要考虑更多的因素。

中文关键词： XML检索；XML文本摘要；语料库；结点语义；片段检索

英文摘要： Automatic summarization of retrieval results can improve the efficiency of the user access to information. Supported by this project, we studied the query-oriented summarization of the XML text and the main contributions are list as follows: ①e proposed a method to find and evaluate the answer nodes, which is content query-based, supplemented by structured query and does not depend on the structural constraints; ②e put forward a query-oriented clustering method for the retrieved results; ③e built a query-oriented XML text summarization corpus that includes more than 3,000 manual articles; ④e proposed a sentence extraction method for XML summarization based on three types of features, and a sentence reordering method based on PageRank; ⑤e put forward a automatically weighting method for XML node (tag/path) based on the strength of topic generalization; ⑥e employed node weights to BM25 model and proposed BM25SF model for XML snippet retrieval; ⑦e participated in the INEX 2011 and one of our runs ranked 2 among 50 runs from 10 participators; ⑧we proposed two novel Top-k query semantics and the corresponding query processing methods for uncertain database. The main conclusions of this project are: ①user's information needs should be taken into consideration sufficiently when clustering and summarizing results of XML retrieval system; ②tructural and semantic information of XML nodes are beneficial to improve the quality of the XML retrieval and XML summarization; ③he XML snippet, retrieved based on the node semantic weights, is helpful for users to quickly decide which document is related; ④ore factors should be considered for automation summarization of XML text.

英文关键词： XML Retrieval; XML Text Summarization; Corpus; Node Semantic; Snippet Retrieval