Context-Aware Lexicon-Based Semantic Similarity of Text

基于上下文感知词典的文本语义相似度

• 批准号: 530281-2018
• 负责人: Mercer, Robert
• 金额: $ 1.82万
• 依托单位: University of Western Ontario
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=659595
• 关键词: Context; Aware; Lexicon; Based; Semantic

The company, Messagepoint, has a cloud-based software platform that provides services for**businesses to enable easy communication with their customers. Examples of this type of communication include simple text messages on customer billing information and complete letter inserts that accompany communication between a business and its customers. The company-specific problem is the comparison of text messages in multiple languages for meaning similarity. Lexicon-based similarity algorithms have shown to work well in the context of word matching when exploring the semantic similarity space along word-relation hierarchies. However, these word-focused algorithms, are not very effective when computing semantic similarity for a larger set of words (such as sentences or paragraphs). Natural Language Processing methods will be integrated into the company's platform in order to generate a shallow semantic representation of the text messages generated by the company's software. This shallow semantic analysis will provide representations of the messages that can be compared for similarity of meaning.

该公司MessagePoint拥有一个基于云的软件平台，可为**企业提供服务，以使其与客户轻松沟通。此类通信的示例包括有关客户计费信息的简单短信以及企业与其客户之间交流的完整信件插入。特定于公司的问题是用多种语言比较文本消息的含义相似性。基于词典的相似性算法在探索沿单词关联层次结构的语义相似性空间时已显示在单词匹配的上下文中可以很好地工作。但是，这些以单词为中心的算法在计算较大单词（例如句子或段落）的语义相似性时并不是很有效。自然语言处理方法将集成到公司的平台中，以生成公司软件生成的文本消息的浅色语义表示。这种浅的语义分析将提供可以比较的信息的表示形式。