学术文本中细粒度知识实体的关联分析

Association Analysis of Fine-Grained Knowledge Entities in Academic Texts

考察特定领域文本中蕴含的细粒度知识实体的使用情况,对知识实体的评估和选择具有重要意义。学术文本中的细粒度知识实体通常具有多个类型、多种关联关系,挖掘知识实体的同质与异质关联关系,有助于深入了解特定领域知识实体的实际使用情况。目前相关研究大多针对学术文本中单一知识实体的抽取和评估,缺乏对知识实体间关系的关注,在一定程度上限制了基于实体抽取进行知识发现的能力。文章以自然语言处理领域为例,对学术论文全文中的细粒度知识实体关联数据进行挖掘,并通过可视化方式揭示关联数据中蕴含的信息。主要是选取全国计算语言学会议2009-2018年间收录的中文论文为原始语料,人工标注论文中使用的知识实体,并针对NLP特点将其细分为“指标实体”“工具实体”“资源实体”“方法实体”4种类型;结合关联规则挖掘算法Apriori和复杂网络分析软件构建知识实体关联网络,揭示该领域常用的知识实体,以及这些知识实体的使用相关性。

The study on fine-grained knowledge entities in domain-specific texts has great significance in the evaluation and selection of knowledge entities.There is a wide variety of fine-grained knowledge entities in academic texts,which usually relate to one another in multiple ways.Mining the homogeneous and heterogeneous association relationship between knowledge entities can help people understand the actual use of knowledge entities in specific fields more deeply.However,most of the current researches focus on the extraction and evaluation of a single knowledge entity in academic texts,lack of attention to the relationship between knowledge entities,which limits the ability of knowledge discovery based on entity extraction to a certain extent.Taking the field of natural language processing(NLP)as an example,this paper mines the data pertaining to the associations between finegrained knowledge entities in full academic articles,and displays the insightful information behind the data through visualization.Specifically,the full texts of Chinese papers published in the National Conference on Computational Linguistics(CCL)from 2009 to 2018 are selected as the original corpus,and the knowledge entities used in the papers are manually annotated.According to the characteristics of NLP,the knowledge entities are divided into four categories:"indicator entity","tool entity","resource entity"and"method entity".In this paper,a knowledge entity association network is established using Apriori algorithm,an association rule mining technique,and complex network analysis software,in an attempt to reveal the common knowledge entities in the field of NLP and their relevance in use.