Interpreting Randomly Wired Graph Models for Chinese NER
作者机构:The School of Electronic and Information EngineeringSuzhouUniversity of Science and TechnologySuzhouChina The School of Computer ScienceTexas Tech UniversityTexasUSA
出 版 物:《Computer Modeling in Engineering & Sciences》 (工程与科学中的计算机建模(英文))
年 卷 期:2023年第134卷第1期
页 面:747-761页
核心收录:
学科分类:0710[理学-生物学] 07[理学] 0701[理学-数学] 0812[工学-计算机科学与技术(可授工学、理学学位)] 070101[理学-基础数学]
基 金:supported by the National Science Foundation of China(NSFC)underGrants 61876217 and 62176175 the Innovative Team of Jiangsu Province under Grant XYDXX-086 Jiangsu Postgraduate Research and Innovation Plan(KYCX20_2762)
主 题:Named entity recognition graph neural network saliency map random graph network interpretation
摘 要:Interpreting deep neural networks is of great importance to understand and verify deep models for natural language processing(NLP)***,most existing approaches only focus on improving the performance of models but ignore their *** this work,we propose a Randomly Wired Graph Neural Network(RWGNN)by using graph to model the structure of Neural Network,which could solve two major problems(word-boundary ambiguity and polysemy)of ***,we develop a pipeline to explain the RWGNNby using Saliency Map and Adversarial *** results demonstrate that our approach can identify meaningful and reasonable interpretations for hidden states of RWGNN.