A full description of the indexing processes which combines statistics, languages features and keyword influence features was given in this study, with 4 indexing experiments of 20 threads, showing the effective way to index public opinion thread with precision and recall up to 0.45 and 0.75.%本研究设计了一种基于统计、语言特征和主题词影响力的综合加权评估的舆情主题标引方案及其关键算法,并通过四个实验较为全面地分析该方案的标引效果和优缺点。
展开▼