联合句法与位置信息的方面情感三元组抽取  

Aspect sentiment triplet extraction based on syntactic and position information

作　　者：王浩畅[1] 黄嘉婷 赵铁军[2] WANG Hao-chang;HUANG Jia-ting;ZHAO Tie-jun(School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China;School of Computer Science and Technology,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]东北石油大学计算机与信息技术学院,黑龙江大庆163318 [2]哈尔滨工业大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《计算机工程与设计》2024年第10期3096-3102,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61402099、61702093)。

摘　　要：为提高方面级情感三元组抽取任务的准确率,提出一种联合依存句法关系和位置偏移信息的抽取模型。在模型上下文编码中添加句法关系,结合图卷积网络捕获结构和结点属性信息,增强三元组要素之间的交互能力;在多任务学习部分加入相对位置偏移信息,充分挖掘方面-观点词对的关系,提高三元组要素抽取的精度。在4个基准英文数据集上的实验结果表明,该方法效果显著且优于其它基线模型。To improve the accuracy of extraction task of aspect sentiment triplet,an extraction model based on joint dependency syntactic relations and location deviation information was proposed.The syntactic relation was added to the model context coding,and the structure and node attribute information were captured through the graph convolutional network to enhance the interaction ability between triplet elements.The relative position offset information was added to the multi-task learning part,and the relationship between aspect-opinion word pairs was fully mined,so as to improve the accuracy of triplet element extraction.Experimental results on four standard English data sets show that the proposed method is effective and superior to other baseline models.

关 键 词：方面级情感分析 三元组抽取 多任务学习 图卷积网络 依存句法 双向长短时记忆网络 深度学习 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]