时态信息检索研究综述

Reviews on Temporal Information Retrieval

【目的】总结国内外时态信息检索研究现状,以期为相关学者更好地把握时态信息检索研究问题提供理论基础。【文献范围】在Google Scholar中分别以检索式"Temporal Information"与"时态信息"且不限定时间范围地进行文献检索,获得部分相关文献后,再结合追溯法最终得到92篇相关文献。【方法】基于文献调研与归纳总结方法,分别从文档中时态信息抽取、查询中时态信息识别和时间感知排序三方面对时态信息检索的相关研究进行综述与评述。【结果】研究发现时态信息检索研究存在着如下问题和挑战:国外对时态检索研究比较多,而国内的相关研究甚少;利用表征时间信息的实体与事件演化信息识别文档关注时间的相关研究不足;缺乏对非周期变化查询的意图预测;时态信息检索模型实验的可重复性有待提高。【局限】未对该领域的文档采集、文档索引以及相关应用进行文献综述。【结论】构建标准化的评测数据集以及无参数时态信息检索模型将是时态信息检索领域的未来方向研究。

[Objective] This study aims to summarize the research status of temporal information retrieval （T-IR） and to provide theoretical basis for the study of the relevant scholars to better grasp the T-IR problems. [Coverage] We first used Google Scholar to search related literatures by typing the keywords ＂termporal information retirevar＇ in Chinese and English repectively, without time limit. After getting some related literatures, we further used the retrospective method to get more related literatures. Finally, we get 92 literatures totally. [Methods] Based on method of literature survey and methods of inducting and summarizing, a survey of the existing literature on temporal information retrieval was presented from the following three aspects： extracting temporal information from document, identifying temporal information in queries and temporal ranking model. [Results] The problems and challenges existing in temporal information retrieval are as follows： little related work existing in China while most of related work existing in foreign countries; lack of methods of data collection and data indexing reflecting dynamic characteristics of real network; ignorance of the important role of the entity and event represent time information when identify the focus time of document; lack of the predicting intent for non-periodic queries and the improvement of reproducibility of temporal information retrieval model experiment to be needed. [Limitations] This paper did not review the document crawling, document index and corresponding application of temporal information retrieval. [Conclusions] The construction of standardized evaluation datasets and non-parameter temporal information retrieval models will be the future research trends of T-IR.