动态语义网的高价值热点主题识别与演化路径分析

High-Value Hot Topic Identification and Evolutionary Path Analysis Based on Dynamic Semantic Network

[目的/意义]高效准确地识别社会诉求主题、把握社会诉求转变节点、追踪主题演化趋势,进而为政务服务和社会治理的和谐有序发展提供支撑。[方法/过程]提出一套基于语义网的高价值主题识别和演化路径分析方法。首先,基于本地上下文语义解析思想,利用词汇共现构建动态语义关系网;其次,利用社区发现算法识别子社区,采用RFM模型对关键词进行价值划分,依据高价值层次关键词识别主题标签;接着,通过计算相邻时间区间的主题相似度来反映主题演化关系;最后,利用上海市的社会诉求数据进行模型验证,与K-means方法进行主题识别效果的比较,并利用精确率、召回率和F1值进行方法效果评测。[结果/结论]研究结果发现,该方法应用效果的提升差额均大于0.3,具有明显的优化效果。研究能够为政府网站领导信箱模块反映的公众关切事项构建全景视图,也能为探索其他社交文本挖掘方法以及支撑国家治理大数据分析实践提供新的思路。

[Purpose/Significance]Effectively and accurately identify the theme of social demands,grasp the transformation nodes of social demands,and track the evolution trend of the theme,so as to provide support for the harmonious and orderly development of government services and social governance.[Method/Process]A set of high-value topic identification and evolution path analysis methods based on semantic web was proposed.Firstly,based on the idea of local contextual semantic parsing,this study used lexical co-occurrence networks to construct a dynamic semantic relational network;Secondly,this study used community discovery algorithm to identify sub communities,used RFM model was applied to divide the value of keywords and topic tags was identified according to high value keywords,and then the topic evolution relationship was reflected by measuring the topic similarity in adjacent time intervals;Finally,the model was verified with the social appeal data of Shanghai,and the theme recognition effect was compared with the K-means method.The accuracy rate,recall rate and F1 value were used to evaluate the effect of the method.[Result/Conclusion]All the three indicators are above the 0.3 cut-off values,proving significant optimization effects have been achieved.This study can build a panoramic view of public concerns reflected in the mailbox module of government website leaders,and also provide new ideas for exploring other social text mining methods and supporting the big data analysis practice of national governance.