| 面向共指事件识别的同义表述模式抽取研究 |
| |
| 引用本文： | 王君泽,宋小炯,杜洪涛.面向共指事件识别的同义表述模式抽取研究[J].情报学报,2020(3):297-307. |
| |
| 作者姓名： | 王君泽 宋小炯 杜洪涛 |
| |
| 作者单位： | 华中科技大学公共管理学院;华中科技大学非传统安全研究中心;国家工业信息安全发展研究中心 |
| |
| 基金项目： | 国家自然科学基金项目“面向Web数据的共指事件信息融合模型研究”(61602198)。 |
| |
| 摘 要： | 在共指消解领域,目前已经有大量研究工作围绕实体共指问题展开,而有关事件共指方面的研究则相对较少。由于事件表述的灵活性,共指事件识别的研究重点之一在于如何构建事件表述相似度的计算模型。而在对同一事件的相似表述中,不仅包含词级别的同义表述,还包含语句级别的同义表述。针对该状况,本文基于新闻报道语料的特点,一方面针对词级别同义表述模式的抽取,设计了同义词知识库的自动构建策略,并考虑了缩略语、同位语等情况的处理;另一方面在词级别同义表述模式抽取的基础上,设计了语句级别同义表述实例的识别策略,进而可以抽取同义表述模式并剔除模式中的冗余成分。通过在实际数据集合上的实验,表明了本文策略的有效性。基于抽取到的词级别和语句级别的同义表述模式,可以有效提升共指事件识别的效果;本文工作也可以视为对共指事件识别现有策略的有益补充。
|
| 关 键 词： | 事件共指 共指消解 同义词识别 同义表述模式 |
Research on the Extraction of Synonymous Representation Patterns for Coreference Event Recognition |
| |
| Wang Junze,Song Xiaojiong,Du Hongtao.Research on the Extraction of Synonymous Representation Patterns for Coreference Event Recognition[J].Journal of the China Society for Scientific andTechnical Information,2020(3):297-307. |
| |
| Authors: | Wang Junze Song Xiaojiong Du Hongtao |
| |
| Institution: | (School of Public Administration,Huazhong University of Science and Technology,Wuhan 430074;Non-traditional Security Center,Huazhong University of Science and Technology,Wuhan 430074;China Industrial Control Systems Cyber Emergency Response Team,Beijing 100040) |
| |
| Abstract: | In the field of coreference resolution, many studies have focused on the issue of entity coreference resolution,while papers about event coreference resolution are fewer. The flexibility of event representation means that one of the key points of event coreference resolution tasks is constructing a model for computing the similarity between events representation. Similar representations of the same event include not only synonymous representations at a word level but also synonymous representations at a sentence level. In this paper, based on the characteristics of the news corpus, we designed a strategy which can construct a synonym knowledge base automatically, and account for the processing of abbreviations and appositives. Conversely, on the basis of synonym expression patterns at the word level, we also designed a strategy for identifying synonym expression instances at a sentence level so as to extract synonymous representation patterns at this level and eliminate redundant components in patterns. Experiments on real data sets show the effectiveness of the proposed strategy. Based on the extracted synonymous representation pattern pairs at the word and sentence levels, we can effectively improve the effect of event coreference resolution. Our study can be regarded as a supplement to existing research on event coreference resolution. |
| |
| Keywords: | event coreference coreference resolution synonym recognition synonym representation patterns |
| 本文献已被 维普 等数据库收录！ |
|
