学术文本词汇功能识别--基于标题生成策略和注意力机制的问题方法抽取  

Recognition of Lexical Functions in Academic Texts:Problem Method Extraction Based on Title Generation Strategy and Attention Mechanism

作　　者：程齐凯[1,2] 李鹏程 张国标[1,2] 陆伟 Cheng Qikai;Li Pengcheng;Zhang Guobiao;Lu Wei(School of Information Management,Wuhan University,Wuhan 430072;Institute for Information Retrieval and Knowledge Mining,Wuhan University,Wuhan 430072)

机构地区：[1]武汉大学信息管理学院,武汉430072 [2]武汉大学信息检索与知识挖掘研究所,武汉430072

出　　处：《情报学报》2021年第1期43-52,共10页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金项目“基于多语义信息融合的学术文献引文推荐研究”(71673211);国家自然科学基金青年科学基金项目“基于深度语义挖掘的引文推荐多样化研究”(71704137)。

摘　　要：学术文本词汇功能识别的目的是实现学术文本中表征问题、方法和对象等词汇的抽取。针对传统识别方法中训练难以获取所导致的识别准确率低、召回率有限和泛化能力差等问题,本研究提出了一种基于深度学习和标题生成策略的学术文本词汇功能识别方法,将任务形式由信息抽取转化为特定形式的标题生成问题。本研究采用构建seq2seq模型和引入注意力机制的方式捕获词汇多层语义信息,最终实现学术文本中问题和方法指代词的生成和获取。实验结果表明,通过应用深度学习方法和标题生成策略,本研究提出的模型能够从摘要中有效识别学术文献的主要研究问题和主要研究方法,并较已有方法在识别效果上有明显提升。The purpose of academic text problem and method identification is to extract research questions and methods from academic text.Aimed at solving the problems of low recognition accuracy,limited recall rate,and poor generalization ability caused by the difficulty of obtaining the training set in traditional recognition methods,this study proposes an academic text problem recognition method based on a deep learning and title generation strategy.The method converts the extraction and recognition of the problem method into the form of title generation in a specific form.By constructing a seq2seq model and introducing an attention mechanism,multi-layer semantic word information was captured to generate and obtain the problem and method pronouns in academic texts.The experimental results showed that through the application of deep learning methods and title generation strategies,this study effectively identified core research problems and core research methods in academic literature.

关 键 词：词汇功能识别 深度学习 自动文摘 学术文本 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]