基于模板学习的智能侨情问句生成方法  

Intelligent Question Generation Method Based on Template Learning for Overseas Chinese Situation

作　　者：方昱龙 王泽锦 王华珍[1] 何霆 FANG Yulong;WANG Zejin;WANG Huazhen;HE Ting(College of Computer Science and Technology,Huaqiao University,Xiamen 361021,China)

机构地区：[1]华侨大学计算机科学与技术学院,福建厦门361021

出　　处：《华侨大学学报（自然科学版）》2023年第6期735-742,共8页Journal of Huaqiao University(Natural Science)

基　　金：国家重点研发计划项目(2018YFB1402500);教育部中外语言交流合作中心国际中文教育研究课题(21YH30B);福建省社会科学基金资助项目(FJ2021B110);中央高校基本科研业务费自主项目(TZYB-202005);华侨大学“华侨华人研究”专项经费资助一般项目(HQHRYB2019-01)。

摘　　要：为解决侨情问句甚少导致训练的侨情语料较少的问题,提出一种基于模板学习的智能侨情问句生成方法。首先,对侨情篇章文本进行包含主题、关系、对象的三元组抽取;其次,构建训练数据集,输入数据由主题和关系构成,输出数据为问句模板;随后,采用以BERT+LSTM+Attention为核心算法的seq2seq框架,实现问句模板生成;最后,对模板问句进行主题文本替换,从而得到最终的实例化问句。采用BLEU,ROUGE-N,公开问答系统评测及人工评价方式对文中方法进行评价。结果表明:BLEU,ROUGE-N,公开问答系统评测及人工评价方式对文中方法的评测结果分别为0.77,0.67,81%,88%,较基线模型有较大的提升。To address the issue of limited training for overseas Chinese language materials due to the scarcity of overseas Chinese question sentences,a template learning based intelligent overseas Chinese situation generation method is proposed.Firstly,the text of overseas Chinese situation is extracted by triplet including theme,relationship and object.Secondly,the training data set is constructed,its input data are composed of themes and relationships,and its output data is question template.Then,BERT+LSTM+Attention as the core algorithm of seq2seq framework is applied to generate question template.Finally,the template question is replaced by the theme text to get the final instantiated question.BLEU,ROUGE-N,public question answering system evaluation and human evaluation method were used to evaluate the proposed method.The results show that the evaluation results of the BLEU,ROUGE-N,public question answering system evaluation and human evaluation method are 0.77,0.67,81%and 88%,respectively,with significant improvements compared to the baseline model.

关 键 词：侨情 问句生成 模板学习 seq2seq 注意力机制 

分 类 号：TP394.4[自动化与计算机技术—计算机应用技术]