关键词查询的推荐技术研究
发布时间：2019-01-30 08:55
【摘要】：关键词查询是文本数据(如万维网等)的经典查询方式,它只需用户输入简单的关键词即可得到结果,省去了学习查询语言和了解底层数据的负担。因其良好的易用性,关键词查询在结构化数据(如关系数据库和深度万维网数据库)上同样得到了广泛的应用。然而,随着底层数据越来越复杂,这种简单易用的查询方式在表达能力方面的局限性日益暴露：一方面,语义模糊或表述不准确的关键词查询难以检索到高质量的结果；另一方面,无结构的关键词难以描述结构化的查询需求。针对这些问题,论文提出了文本数据和结构化数据上的查询推荐技术,在保证易用性的前提下,辅助用户准确地表达查询意图。论文的主要研究工作和贡献包括： 1.文本数据上主题相关的查询词推荐：为了辅助用户生成高质量的关键词查询,论文提出了一种主题相关的查询词推荐方法。该方法考虑用户输入的部分查询,分析隐含的查询主题,推荐与主题相关的关键词,从而辅助用户生成高质量的完整查询,准确地表达查询意图。此外,查询词推荐支持实时的自动补全,随着用户逐字母地敲入查询词的前缀,可以实时地进行响应,推荐包含该前缀的查询词,从而方便用户对查询词进行快速地选择或修改,提高了查询的效率。 2.传统关系数据上交互式SQL查询语句推荐：针对传统关键词查询难以准确表达关系数据库结构化查询需求的问题,论文提出了一种交互式的SQL查询语句推荐方法,根据用户输入的关键词实时地推荐相关的SQL语句,并按照相关性对SQL语句进行排序。该方法提出了有效的推荐模型度量关键词与SQL语句之间的相关性,设计了快速的算法支持SQL语句的实时推荐,从而在减轻查询负担的同时,辅助用户准确地表达结构化数据上的查询意图,有效地解决了无结构的关键词查询与结构化数据之间的信息鸿沟问题,提供了一种既好用又有很强表达能力的关系数据库查询方式。 3.深度万维网数据库查询推荐：在深度万维网数据库访问受限的前提下,论文提出了一种基于关键词的深度万维网数据库查询推荐方法。该方法通过查询日志挖掘和数据库采样技术,在访问受限情况下分析用户的查询意图,将关键词映射为数据库上的结构化表单查询并在线地获取相关的结果,从而为深度万维网查询与现有搜索引擎的无缝集成提供了一种有效的手段。
[Abstract]:Keyword query is a classical query method for text data (such as the World wide Web). It only needs users to input simple keywords to get the results, which saves the burden of learning query language and understanding the underlying data. Because of its good usability, keyword query is also widely used in structured data (such as relational database and deep Web database). However, as the underlying data become more and more complex, the limitations of this simple and easy-to-use query in terms of expressive ability are increasingly exposed: on the one hand, it is difficult to retrieve high quality results for keyword queries with fuzzy semantics or inaccurate representation; On the other hand, unstructured keywords are difficult to describe structured query requirements. Aiming at these problems, this paper proposes query recommendation technology on text data and structured data, which can help users express their query intention accurately on the premise of ease of use. The main research work and contributions are as follows: 1. Topic-related query word recommendation on text data: in order to assist users to generate high-quality keyword queries, a topic-related query recommendation method is proposed in this paper. This method takes into account some queries input by users, analyzes the implicit query topics, and recommends the keywords related to the topic, so as to assist users to generate high quality complete queries and accurately express the query intention. In addition, the query word recommendation supports real-time automatic completion. As the user types the prefix of the query word alphabetically, it can respond in real time, and the query word containing the prefix is recommended. Therefore, it is convenient for users to select or modify query terms quickly and improve the efficiency of query. 2. Recommendation of interactive SQL query statement on traditional relational data: aiming at the problem that traditional keyword query can not accurately express the requirement of structured query in relational database, an interactive SQL query statement recommendation method is proposed in this paper. The relevant SQL statements are recommended in real time according to the keywords entered by the user, and the SQL statements are sorted according to the correlation. This method proposes an effective recommendation model to measure the correlation between keywords and SQL statements, and designs a fast algorithm to support the real-time recommendation of SQL statements, so as to reduce the query burden at the same time. In order to solve the problem of information gap between unstructured keyword query and structured data, the user can express the query intention of structured data accurately, and solve the problem of information gap between unstructured keyword query and structured data. This paper provides a query method of relational database which is easy to use and has strong expressive ability. 3. Deep Web Database query recommendation: under the premise of restricted access to Deep World wide Web database, this paper proposes a keyword based deep Web database query recommendation method. In this method, query log mining and database sampling technology are used to analyze the user's query intention in the case of restricted access, and map the keywords to the structured form query on the database and obtain the related results online. It provides an effective method for the seamless integration of deep web query and existing search engines.
【学位授予单位】：清华大学
【学位级别】：博士
【学位授予年份】：2012
【分类号】：TP391.3
本文编号：2417966
[Abstract]:Keyword query is a classical query method for text data (such as the World wide Web). It only needs users to input simple keywords to get the results, which saves the burden of learning query language and understanding the underlying data. Because of its good usability, keyword query is also widely used in structured data (such as relational database and deep Web database). However, as the underlying data become more and more complex, the limitations of this simple and easy-to-use query in terms of expressive ability are increasingly exposed: on the one hand, it is difficult to retrieve high quality results for keyword queries with fuzzy semantics or inaccurate representation; On the other hand, unstructured keywords are difficult to describe structured query requirements. Aiming at these problems, this paper proposes query recommendation technology on text data and structured data, which can help users express their query intention accurately on the premise of ease of use. The main research work and contributions are as follows: 1. Topic-related query word recommendation on text data: in order to assist users to generate high-quality keyword queries, a topic-related query recommendation method is proposed in this paper. This method takes into account some queries input by users, analyzes the implicit query topics, and recommends the keywords related to the topic, so as to assist users to generate high quality complete queries and accurately express the query intention. In addition, the query word recommendation supports real-time automatic completion. As the user types the prefix of the query word alphabetically, it can respond in real time, and the query word containing the prefix is recommended. Therefore, it is convenient for users to select or modify query terms quickly and improve the efficiency of query. 2. Recommendation of interactive SQL query statement on traditional relational data: aiming at the problem that traditional keyword query can not accurately express the requirement of structured query in relational database, an interactive SQL query statement recommendation method is proposed in this paper. The relevant SQL statements are recommended in real time according to the keywords entered by the user, and the SQL statements are sorted according to the correlation. This method proposes an effective recommendation model to measure the correlation between keywords and SQL statements, and designs a fast algorithm to support the real-time recommendation of SQL statements, so as to reduce the query burden at the same time. In order to solve the problem of information gap between unstructured keyword query and structured data, the user can express the query intention of structured data accurately, and solve the problem of information gap between unstructured keyword query and structured data. This paper provides a query method of relational database which is easy to use and has strong expressive ability. 3. Deep Web Database query recommendation: under the premise of restricted access to Deep World wide Web database, this paper proposes a keyword based deep Web database query recommendation method. In this method, query log mining and database sampling technology are used to analyze the user's query intention in the case of restricted access, and map the keywords to the structured form query on the database and obtain the related results online. It provides an effective method for the seamless integration of deep web query and existing search engines.
【学位授予单位】：清华大学
【学位级别】：博士
【学位授予年份】：2012
【分类号】：TP391.3
【参考文献】
相关期刊论文 前2条
1 刘玉奎;周立柱;范举;;中文深度万维网数据库的现状研究[J];计算机学报;2011年02期
2 范举;周立柱;;基于关键词的深度万维网数据库选择[J];计算机学报;2011年10期
，本文编号：2417966
本文链接：https://www.wllwen.com/kejilunwen/sousuoyinqinglunwen/2417966.html
