[目的/意义] 关键词作为一种能够揭示学术文本主题及核心内容的词汇或术语，对其进行功能识别可为知识和文献的快速、精确获取提供底层索引支持。[方法/过程] 针对现有研究在关键词上下文建模中多局限于文本层面的符号语义表征，在深入挖掘文献行书规律的基础上，提出一种基于多特征融合的词汇功能识别模型。模型在采用BERT模型捕获关键词上下文依赖特征的同时，融合关键词在关键词列表和全文中的位置信息以及词汇功能先验知识信息，继而采用注意力机制和前馈神经网络对关键词进行问题方法的语义功能判别。[结果/结论] 实验结果显示，关键词的位置信息和先验知识均能有效提升关键词语义功能识别性能，其中先验知识对识别效果的提升有较大贡献。
[Purpose/significance] Keywords, as a kind of vocabulary or term that can reveal the subject and core content of a text, can identify the functions and provide the underlying index support for fast and accurate acquisition of knowledge and documents.[Method/process] Aiming at the existing studies that are mostly limited to the semantic representation of symbols at the text level in vocabulary context modeling, this paper proposes a lexical function recognition model based on multi-feature fusion. On the basis of capturing the context-dependent features of keywords using the BERT model, the position information of keywords in the keyword list and the full text and prior knowledge of vocabulary functions are fused, and then the attention mechanism and feed-forward neural network are used for the identification of key words by problem-solving method.[Result/conclusion] The experimental results show that both the location information and priori knowledge of the keywords can improve their word function recognition effect, and the prior knowledge has a greater contribution to the recognition effect.
[1] 杨涛.中文智能搜索引擎浅析[J].图书情报工作,2002,56(1):62-65.
[2] 万华林, CHOWDHURY M U.基于支持向量机的图像语义分类[J].软件学报,2003,14(11):1891-1899.
[3] LU W, LI X, LIU Z F, et al. How do author-selected keywords function semantically in scientific manuscripts[J]. Knowledge organization,2019,46(6):403-418.
[4] DALRYMPLE M, KAPLAN R M, MAXWELL J T, et al. Formal issues in lexical-functional grammar[M]. Stanford:CSLI Publications, 1995.
[5] MOON T, ERK K, BALDRIDGE J. Crouching dirichlet, hidden markov model:unsupervised POS tagging with context local tag generation[C]//Proceeding of the 2010 conference on empirical methods in natural language processing.Cambridge:Association for Computational Linguistics, 2010:196-206.
[6] 孙静,李军辉,周国栋.基于条件随机场的无监督中文词性标注[J].计算机应用于软件,2011,28(4):21-23.
[7] KONDO T, NANBA H, TAKEZAWA T, et al. Technical trend analysis by analyzing research papers' titles[M]. Berlin:Springer, 2009.
[8] NANBA H, KONDO T, TAKEZAWA T. Automatic creation of a technical trend map from research papers and patents[C]//Proceedings of the 3rd international workshop on patent information retrieval.New York:Association for Computing Machinery,2010:11-16.
[9] GUPTA S, MANNING C. Analyzing the dynamics of research by extracting key aspects of scientific papers[C]//Proceedings of 5th international joint conference on natural language processing.Stroudsburg:Association for Computational Linguistics, 2011:1-9.
[10] TSAI C T, KUNDU G, DAN R. Concept-based analysis of scientific literature[C]//Proceeding of the 22nd acm international conference on conference on information & knowledge management.New York:Association for Computing Machinery,2013:1733-1738.
[11] 程齐凯,李信.面向语义出版的学术文本词汇语义功能自动识别[J].数字图书馆论坛, 2017,159(8):24-31.
[12] HEFFERNAN K, TEUFEL S. Identifying problems and solutions in scientific text[J]. Scientometrics,2018, 116(1):1-16.
[13] 陆伟,李鹏程,张国标,等.学术文本词汇功能识别——基于BERT向量化表示的关键词自动分类研究[J].情报学报,2020,39(12):1320-1329.
[14] 唐明,朱磊,邹显春.基于Word2Vec的一种文档向量表示[J].计算机科学,2016,43(6):214-217,269.
[15] VASWANI A, SHAZEER N, PARMAR N. Attention is all you need[C]//Proceedings of the 31th international conference on neural information processing systems.New York:Curran Associates Inc, 2017:6000-6010.
[16] DEVLIN J, CHANG M W, LEE K, et al. Bert:pre-training of deep bidirectonal transformers for language understanding[C]//Proceeding of the 2019 conference of the North American chapter of the Association for Computational Linguistics:human language technologies. Minneapolis:Association for Computational Linguistics,2019:4171-4186.
[17] 王佳敏,陆伟,刘家伟,等.多层次融合的学术文本结构功能识别研究[J].图书情报工作, 2019,63(13):95-104.
[18] 黄永,陆伟,程齐凯,等.学术文本的结构功能识别——基于段落的识别[J].情报学报, 2016,35(5):530-538.
