基于混合策略的藏文人称代词指代消解研究  

Research of tibetan personal pronouns anaphora resolution based on mixed strategy

作　　者：夏吾吉 华却才让 XIA Wuji;HUAQUE Cairang(Tibetan Information Processing Key Laboratory of Ministry of Education,Qinghai Normal University,Xining 810008,China;Normal College for Nationalities,Qinghai Normal University,Xining 810008,China)

机构地区：[1]青海师范大学藏文信息处理教育部重点实验室,西宁810008 [2]青海师范大学民族师范学院,西宁810008

出　　处：《计算机工程与应用》2018年第7期66-69,113,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61363055);青海师范大学中青年科研基金(No.17ZR11);青海省科技计划项目(no.2017-GX-146)

摘　　要：指代消解是文本理解和信息抽取的一项重要任务。针对这一任务,提出了基于混合策略的藏文人称代词指代消解方法,通过对藏文人名、人称代词的形态特征和构词规律的研究,制定了三类消解规则和有效统计特征,采用基于规则、最大熵模型以及规则与最大熵模型相结合的三种方法实现了藏文人称代词的指代消解系统。在包含2 306个待消解对的藏文句子集上,经测试分别获得76.02%、86.21%和88.16%的F值。Anaphora resolution is a vital task in text information processing and information extraction.In view of this task,this paper presents an approach on Tibetan personal pronouns anaphora resolution based on mixed strategy.By researching on morphological features and word-formation patterns of Tibetan personal names and personal pronouns,four rules and four features are established.A method of rules,maximum entropy model and a mixed method of the rules and maximum entropy model are utilized for addressing the task.In the experiments,2 306 Tibetan sentences containing the digestion pairs are tested,and the F values of the above three methods are 76.02%,86.21%and 88.16%respectively.

关 键 词：藏文人称代词 最大熵模型 混合策略 指代消解 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]