基于同义实体识别的web信息集成 web information integration based on synonymous entities recognition.pdfVIP

2015 年 第24 卷 第9 期 计 算 机 系 统 应 用 基于同义实体识别的Web 信息集成① 徐喆昊, 吴共庆, 胡学钢 (合肥工业大学 计算机系, 合肥 230009) 摘 要: 准确有效地集成海量Web 信息, 是Web 信息动态聚合、市场情报分析、舆情分析、商业智能等分析型 应用的重要基础. 针对数据集成过程中不同实体指代同一实体的问题, 利用搜索引擎返回的页面摘要信息, 设计 并实现了一种基于搜索引擎的同义实体识别算法FSE, 并提出了一种基于同义实体识别的Web 信息集成框架. 在 医院信息集成测试数据集上的实验结果表明, FSE 算法效果优于基于VarientDice、VarientCosine 、VarientJaccard 、 VarientOverlap 相似度计算的同义实体识别算法. 关键词: Web 信息集成; 同义实体识别; 相似度计算; 搜索引擎 Web Information Integration Based on Synonymous Entities Recognition XU Zhe-Hao, WU Gong-Qing, HU Xue-Gang (Department of Computer Science, Hefei University of Technology, Hefei 230009, China) Abstract: Integrating massive information on the Web accurately and effectively is the important basis of developing analytic applications, such as Web information dynamic aggregation tools, market information analysis tools, public opinion analysis tools, and business intelligence tools, etc. To solve the problem that different presentations refer to the same entity during the integrating process, this paper proposes an algorithm to recognize the synonymous entities by using the snippets from the search engine and a frame of Web information integration based on synonymous entities recognition. The experimental results on hospital information integration testing data sets show that the proposed method outperforms the synonymous entities recognition based on VarientDice, VarientCosine, VarientJaccard and VarientOverlap. Key words: Web information integration; synonymous entities recognition; similarity computation; search engine 1 引言 联网上的Web 数据资源, 提取其中用户真正需要的信 随着互联网的飞速发展, 互联网已逐渐成为信息 息, 并以友好的方式展现给用户, 成为 Web 集成领域 生产和消费的平台, 丰富的数据资源和方便的查询手 新的挑战. 段也促使其成为获取信息的重要途径.