面向词汇功能的学术文本语义识别与知识图谱构建

Nowadays，scientists take use of scholar search engine (Google Scholar, SCI, etc.) and informatics tools to find scholar information. However, these tools and the methods behind them cannot address some basic but also important questions such as: "Is term 'Data Mining' a kind of technique the paper takes or topic the paper focus on?", "What methods were developed to solve a particular problem according to these papers?", "Who is the one proposed this research topic for the first time?", "What does the term 'state of the art' in a paper refer to?". This subject aims to develop a suit of theories, methods and tools to help scientists to find the answers of the questions listed above in a fast and convenient way. The core component of this subject is to identify the functionality of mentions in research paper. We try to build a semantic framework to define the functions of term and the relations between terms, and propose some methods to address problems such as term function identification, relation extraction, coreference resolution, concept evolution identification. Thus this subject chooses some research domain to build a functionality oriented scientific research knowledge graph. This subject is of significant academic and practical value. The methods and knowledge graph can be used to improve the performace of scholar search engine and enhance the semantic leve of academic text analysis. This subject have also potential value in automatic summarization, knowledge management, et.al.

当前，科研人员难以通过学术搜索引擎和科学计量工具快速回答一些基本但重要的问题，如："数据挖掘在某文献中是方法还是主题？"、"某问题有哪些技术可解决"、"某问题最早由哪篇文献提出"、"文献中所指的state of the art具体是什么？" 本课题着眼于学术文本词汇功能识别这一核心任务，构建一套理论、方法和工具，帮助科研人员快速回答上述问题。本课题试图对学术文本中词汇功能及词汇间语义关系进行建模，探索自动化方法以识别词汇的功能（技术、主题、领域等）；建立词汇语义关联；实现学术词汇的指代消解，分析概念的演化脉络；在此基础上，构建面向词汇功能的知识图谱。 本研究具有较大的理论与应用价值。提出的技术方法及构建的知识图谱可以用于提升学术搜素引擎搜索结果的质量，改进学术文本分析效果，提升相关应用的语义化水平，还可以广泛应用于自动摘要、知识管理等各个领域。

本课题着眼于学术文本词汇功能识别这一核心任务，构建一套理论、方法和工具，帮助科研人员快速回答上述问题。本课题试图对学术文本中词汇功能及词汇间语义关系进行建模，探索自动化方法以识别词汇的功能（技术、主题、领域等）；建立词汇语义关联；实现学术词汇的指代消解，分析概念的演化脉络；在此基础上，构建面向词汇功能的知识图谱。本研究提出的技术方法及构建的知识图谱可以用于提升学术搜索引擎搜索结果的质量，改进学术文本分析效果，提升相关应用的语义化水平，还可以广泛应用于自动摘要、知识管理等各个领域。. 围绕学术文本的词汇功能语义、多粒度概念抽取、词汇功能识别和词汇功能维度的知识图谱等专题，发表论文被SSCI/SCIE索引1篇，EI索引1篇；领域权威期刊论文、核心期刊论文18篇左右；完成了专著1本；构建了计算机领域词汇语料库1个，开发了学术文本语义分析工具集4个；申请相关专利4项；获得国家级奖励1项，省部级奖励1项。