基于领域本体的创业板公司年报风险信息抽取方法研究

发布时间：2018-06-12 04:13

本文选题：领域本体 + 创业板　；参考：《湖南大学》2013年硕士论文

【摘要】：年度报告是投资者等利益相关者了解上市公司经营情况的重要信息渠道，近年来报表之外的各种文字性补充说明和解释愈来愈多，年度报告全文的厚度在不断增加，其中文字性内容的比例增加非常迅速。很多学者研究证明上市公司年报描述性信息具有较大的决策相关性，如何从大量的文字内容中抽取价值相关信息是一个值得研究的问题。创业板市场作为我国的新生事物，为高新技术企业提供了很好的融资平台，也为我国科技发展和技术创新提供了极大的支持。由于创业板自身的特点，其给投资者带来高回报的同时也蕴藏着极大的风险，创业板公司在其年报中披露了何种风险、以何种方式披露风险对投资者能否正确认识公司风险有着重要的影响。本文设计了一种基于领域本体的创业板公司年报风险信息抽取方法，首先梳理了以往学者的相关研究和理论，其次通过分析创业板公司年报风险信息的文本特征和大量阅读风险领域相关材料，收集获取风险领域知识，设计出创业板公司年报风险领域本体；并利用本体开发工具protégé实现风险领域本体的构建，设置本体概念间的关联属性，将生成的本体OWL文件存储于本地磁盘中。然后在制定信息抽取规则的基础上，使用基于Java的可扩展系统开发平台Eclipse实现创业板公司年报中披露的风险类型的抽取、风险量化信息以及风险对策信息的识别，并对信息抽取结果进行检验，通过计算准确率和召回率综合指数，，证明本文设计的信息抽取方法具有较好的性能。本文将会计学科与计算机技术、自然语言处理技术结合起来，为传统会计研究提供了一个高效率的基础数据处理工具。
[Abstract]:The annual report is an important information channel for investors and other stakeholders to understand the operating situation of listed companies. In recent years, there are more and more written supplementary explanations and explanations outside the financial statements, and the thickness of the full text of the annual report is increasing. Among them, the proportion of textual content increases very rapidly. Many scholars have proved that the descriptive information of annual report of listed companies has greater decision relevance. How to extract value related information from a large number of text content is a problem worth studying. As a new thing in China, the gem market provides a good financing platform for high-tech enterprises, and also provides a great support for the development of science and technology and technological innovation in China. Because of the characteristics of gem itself, it brings high returns to investors and also contains great risks. What kind of risks do gem companies disclose in their annual reports? How to disclose the risk has an important influence on the investors' understanding of the company's risk. This paper designs a method of extracting risk information of gem annual report based on domain ontology. First of all, it combs the relevant research and theory of previous scholars, secondly, through analyzing the text characteristics of risk information of gem annual report and reading a lot of relevant materials in risk domain, it collects and acquires risk domain knowledge. The risk domain ontology of gem annual report is designed, and the ontology of risk domain is constructed by using ontology development tool prot 茅 g 茅, and the relational attributes between ontology concepts are set up, and the generated ontology owl file is stored in the local disk. Then, based on the rules of information extraction, Eclipse, an extensible system development platform based on Java, is used to realize the risk type extraction, risk quantification information and risk countermeasure information identification, which are disclosed in the gem annual report. The results of information extraction are tested and the accuracy and recall index are calculated. It is proved that the information extraction method designed in this paper has good performance. This paper combines accounting science with computer technology and natural language processing technology, which provides a high efficiency basic data processing tool for traditional accounting research.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F232;F832.51

【参考文献】