基于层次结构和共现关系的在线评论主题图谱构建方法研究

Research on the Method of Constructing Topic Map for Online Reviews Based on the Hierarchy and Co-occurrence Relationship

[目的/意义]产品在线评论中既包含用户对产品的整体体验评估,又包含对产品的特定属性评估,通常超负荷、碎片化且价值密度低。文章旨在提出一种在线评论的主题图谱构建方法,试图将在线评论进行更有效的信息组织。[方法/过程]文章首先设计模式层三元组,然后提出数据层实例化的方法,该方法主要是结合线索利用理论明确主题结构,基于依存句法分析等自然语言处理技术抽取主题词、计算主题情感强度并发现主题的共现关系,而后利用统计分析方法计算主题共现的权重。将该方法应用于实际的在线评论中进行主题图谱的构建实验,最后将生成的主题图谱存储于图数据库中。[结果/结论]构建的主题图谱清晰展现主题之间的层次关系和共现关系以及主题的情感强度,具有可行性与有效性,可以为主题图谱在在线评论信息组织方面的研究提供新的思路,有助于消费者、企业和平台利用在线评论信息资源从微观的主题描述关系和宏观的情感表达上全面掌握产品及其相关情况,提升在线评论信息的商业价值。

[Purpose/significance]Online product reviews include not only the users’overall experience evaluation of the product,but also the evaluation of the specific attributes of the product,which are usually overloaded,fragmented,and have a low-value density.The purpose of the paper is to propose a method for constructing a topic map of online reviews,trying to organize online reviews more effectively.[Method/process]This research first designs the triples of the pattern layer,and then proposes the instantiation method of the data layer.This method mainly combines clues utilization theory to clarify the topic structure,and then extracts the topic words,calculates the emotional intensity of topics,and discovers the co-occurrence relationship of the topics based on natural language processing techniques such as dependency parsing analysis.The statistical analysis method is used to cal-culate the weight of the topic co-occurrence.This method is applied to actual online reviews to construct the topic map,and finally which is stored in the graph database.[Result/conclusion]The constructed topic map can clearly show the hierarchical relation-ship and co-occurrence relationship between topics as well as the emotional intensity of the topics,which is feasible and effective,and can provide new ideas for the research of the topic maps in the organization of online review information.It helps consumers,enterprises and platforms to use the information resources of online review to fully grasp the products and related conditions from both micro topic description relationship and macro emotional expression,enhancing the business value of online review information.