摘要
机器善于处理言语中的客观理性信息,却拙于理解言语中的主观情感信息。情感计算是语言人工智能领域有待破解的一道难题,准确识别评价词的语用指向对象是其中的一大瓶颈。文章从计算语言学角度研究了评价形容词状语的语用指向,基于词汇特征、句法结构与语义框架,归纳概括出14种语用指向类型,精准找出帮助计算机自动辨别各种类型的形式化特征,编码成机器可读的形式语言,构建评价形容词状语语用指向识别与抽取规则簇。实验证明,规则簇有良好的领域适应性(F1=0.932)和词汇覆盖性(F1=0.917)。文章将语用指向理论与机器情感计算相融合,证实了基于评价词的语用指向本体研究,构建知识驱动、推理可解释的评价对象抽取系统的可能性,提供了一种评价对象自动抽取的新方案。
The machine is good at dealing with objective and rational information in speech,but poor at understanding the subjective affective information.Affective computing is a difficult problem to be solved in the field of linguistic artificial intelligence.One bottleneck problem is to accurately identify the pragmatic target of evaluative words.From the perspective of computational linguistics,this paper studies the pragmatic orientation of evaluative adjective adverbials.Based on the lexical feature,syntactic structure and semantic framework,it summarizes 14 types of pragmatic orientation,precisely points out the formal features that help the computer automatically identify various types of pragmatic orientation,encodes them into machine-readable formal language,and constructs the rule cluster of pragmatic orientation recognition and extraction of evaluative adjective adverbials.Experiments show that the rule cluster has good domain adaptability(F1=0.932)and vocabulary coverage(F1=0.917).Combining the pragmatic orientation theory with machine affective computing,this paper confirms the possibility of building a knowledge-driven and reasoning-interpretable evaluative object extraction system based on the linguistic research on evaluative words'pragmatic orientation,and provides a new scheme for automatic evaluation object extraction.
作者
周红照
侯敏
滕永林
Zhou Hongzhao;Hou Min;Teng Yonglin
出处
《语言研究集刊》
2023年第2期210-224,394,共16页
Bulletin of Linguistic Studies
基金
国家社会科学基金项目“中文情感计算的形式语义逻辑研究”(项目编号22CYY023)的资助。
关键词
评价形容词状语
语用指向
评价对象抽取
形式特征
规则簇
evaluative adjective adverbial
pragmatic orientation
evaluative object extraction
formal feature
rule cluster
作者简介
周红照,zhzwin2011@163.com;侯敏,houmin@cuc.edu.cn;滕永林,tengyonglin@cuc.edu.cn
