Research On Automatic Extraction Of Enterprise Supply Relationship Based On NLP

With the deepening of economic globalization,the competition between enterprises has become the competition between supply chains.A good supply chain is indispensable for enterprises to improve competitiveness,reduce costs and increase profit margins.For a supply chain,the most important part is the supply relationship between enterprises.In addition,the supply relationship of enterprises is also of great significance to entrepreneurs,consumers and investors.Enterprise supply relationship enables entrepreneurs to understand the industry situation and help them choose the right suppliers at the beginning of entering the industry.For consumers,when choosing a product,the component supplier information of the product manufacturer can provide a reference for their purchase decision.For investors,the supply relationship between enterprises can help them assess the trend of the investment market correctly and obtain profits.Therefore,the extraction of supply relationship is of great significance.This dissertation mainly focuses on the research of the extraction of supply relationship between enterprises.The supply relationship between enterprises is of great significance in understanding and analyzing industries,making investment decisions and choosing business partners.It is a problem of entity relationship extraction,and named entity recognition is the basis of entity relationship extraction.According to the current research situation of named entity recognition and relation extraction,the methods based on rules have a high accuracy,but it needs the participation of domain experts and is not transplantable;while the methods based on statistical do not require the participation of domain experts,but it often do not achieve the desired results in the specific application scenarios and relay on the training data badly.Applying only one method will not solve the problem,this dissertation combines the method based on statistical with the method based on rules,and puts forward my own ideas for solving the problem of enterprise supply relationship extraction.This dissertation has three aspects of enterprise supply relationship extractionresearches.Firstly,thesis studies the method of Chinese company entity identification.The characteristics of corporate names in financial texts are analyzed.In view of the existing problems,on the basis of the existing named entity recognition tools,this dissertation combines dependency syntax analysis,dictionary and rule method to deal with the juxtaposition of multiple company names,and achieves good results in the text of the annual report of listed companies.Secondly,in terms of product name identification,this dissertation integrates boundary words,product keywords and other features into the conditional random field,uses the dependency syntax to analyze the recognition results and identify the boundary of complex product names.The method has made good results of the accuracy,recall rate and F value in the annual report of listed companies.Finally,in the aspect of enterprise supply relationship judgment and extraction,this dissertation uses the recent syntactic dependent verbs to judge the semantic relationship between entities,constructs a relation word library by combining artificial construction with automatic construction,and uses the relation word to judge the theme of the text.And experiments on the annual report of listed companies has meet the expected requirement.The main contributions of the dissertation contain: firstly,according to the characteristics of product names in financial and economic texts,this dissertation classifies them into product category words,and effectively recognizes them by combining the features of boundary words and dictionaries.In addition,this dissertation proposes the dependency syntax rules to modify the recognition results of company name and product name,effectively improving the effect of entity recognition.Finally,in the aspect of relational judgment,in order to accurately extract the enterprise supply relationship,this dissertation proposes to use the method of dependency syntax analysis combined with relation words to judge the relationship between multiple entities.