| The named entities in crop text records contain a large number of text information closely related to ornamental plants.It is an important condition to quickly and accurately identify the named entities of ornamental plants in plant text data,which can promote the development of knowledge mapping in agricultural intelligence.As an important sub task of natural language processing,named entity recognition is the key of text information recognition.At present,in the field of ornamental plants,text named entity recognition is still a blank.The research on named entity recognition of ornamental plants has a promoting significance for the establishment of agricultural question answering system and the construction of agricultural knowledge map.In this paper,based on the related named entity recognition technology,the main research work is as follows:(1)This paper explores the entity recognition models CRF,bilstm and bilstm-crf related to the research process of this subject,analyzes the results of applying the three algorithms to the entity recognition of ornamental plants,and compares the advantages and disadvantages of several algorithms;(2)Construct the named entity recognition data set of ornamental plants,including the collection,cleaning,label setting and labeling of the original data;(3)The accuracy rate,recall rate and F1 value of CRF,bilstm and bilstm-crf were compared and analyzed in the constructed named entity recognition data set of ornamental plants,and the conclusion that the performance of bilstm-crf model was the most outstanding was obtained,in which the accuracy rate was 94.09%,recall rate was 94.18% and F1 value was 94.00%.;(4)A named entity recognition system of ornamental plants based on bilstm CRF model was constructed,and the main function modules were set up. |
