基于叙词表的农业舆情话题发现算法研究

Study on Agriculture Public Opinion Topic Detection Algorithm Based on Thesaurus

针对如何高效地发现农业舆情话题,提出了一种基于叙词表的舆情话题发现算法。该算法首先基于《农业叙词表》和综合性词表及网络新词构建叙词词典,作为中文分词软件的词典;然后运用TF-IDF计算特征词的权值,选取前P个特征词表示文本,并基于叙词间的关系计算词语相似度;最后,以叙词为节点构建无向图,通过对无向图聚类实现网络热点话题的发现。分析结果表明,该算法的最小识别代价为0.3534,算法运行效率相比传统算法较高。

For efficient detection of agriculture public opinion topic from massive information,a network topic detection algorithm based on thesaurus was given in this paper.Firstly,based on Agriculture Comprehen-sive Thesauruses and network new words,a thesaurus dictionary was built as the dictionary of Chinese word software;then the weight of feature words were caculated by TF -IDF,and several feature words were selected as represent text,the words similarity was computed combining with the relationship of thesaurus;finally,the thesaurus were taken as nodes to build an undirected graph,and the network hot topic detection was realized through the undirected graph clustering.The analysis results showed that the minimum algorithm cost of this method was 0.3534,its algorithm efficiency was higher than that of traditional algorithm.