面向科学研究主题的文献隐含时间信息分析与挖掘

摘要

本研究针对缺乏检索对象时间范围和精确语义表达式的用户弱信息检索需求,为了更好地体现出文献之间的时间关联性,提出了一种文本内容时间信息的分析和挖掘方法.本文归纳总结了学术文献常用时间词,并基于图书情报学题录信息完成文本时间词抽取.在此基础上利用标签主题模型labeled-LDA实现文本隐含时间标签的生成,以及基于时间维度的文本语义相似度计算.实验结果表明,本研究能较好地生成文献隐含时间信息,也能更有效地帮助用户提高弱信息检索结果的相关性.%Taking account of the weak information requirement which is lack of representation in the time scope and precise semantic expression description,this research proposes a new data mining and analysis method for text content in order to better reflect the temporal similarity among literature.The common temporal expressions in research papers are summarized and temporal expression extraction is finished on bibliographic data in the paper.Then we generate the implied temporal profile of each paper using labeled-LDA model,and calculate text similarity computing between papers based on time dimension of text representation.The experiment proves that our method works well in creating implied temporal profile of each document,and also improves the relevance of search results in users' weak information searching scenarios.