Research And Application Of Named Entity Recognition Method For The Bidding Data

Named entity recognition is one of the hot research topics in the field of data mining and natural language processing.With the explosive growth of network data,the demand of people increases on how to quickly and accurately obtain the significant information in the data.Named entity recognition is a key technology to extract key information,and plays a key role in the fields of natural language processing,such as information extraction,information retrieval,text classification and so on.Researchers pay much attention to it.In this paper,according to the theoretical study of Chinese named entity recognition,aiming at the urgent requirements of the extraction and recognition of named entities of the bidding data,the research focuses on the design of rules about named entity extraction and the method of named entity recognition based on hybrid model.Our research builds a data set obtaining the national bidding data in network platforms,and proves the validity of the method through experiments.This method can meet the requirement of entity extraction include the expert names,project contact names,contact address,bidding organization names,agency names and winning organization names in bidding data.The main work and achievements of this paper include:(1)This paper analyses the constitution rules of entities and the features of texts in the bidding data.And according to the characteristics,this paper makes boundary rule base and entity rule base suitable for named entity recognition of bidding data.Then make the research on named entities with rule-based method.(2)The paper proposes a method of named entity recognition based on hybrid model.The method uses two order hidden Markov model as the statistical model,and improves the viterbi algorithm,then makes full use of context information to recognize entity.At the same time,according to the characteristics of bidding data,the rules based pre-processing and post correction are added to the statistical model to improve the effect of named entity recognition.(3)According to the constructed data set obtaining the national bidding data in network platforms,our research uses three sets of experiments to verify the recognition effect of six kinds of entities with the rule-based method,the hybrid model proposed in this paper,the LTP system of Harbin Institute of Technology and the NLPIR system of Chinese Academy of Sciences.The results of experiments show that the proposed hybrid model for named entity recognition in bidding data has a better recognition effect.The results of above research show that for the named entity recognition in bidding data,the proposed method based on hybrid model has good recognition effect,and can be used as a basic method for constructing the named entity recognition system of the bidding data.The method can effectively improve the searching efficiency and accuracy when the relevant organizations obtain bidding information.