Research On Named Entity Recognition In Land And Resources Based On Weibo

In recent years,as an important production factor,land and resources have gradually become the focus of social attention.Therefore,the construction of land and resources information has also become the key to promoting the development of land and resources in the new era.However,at present,there is a problem of insufficient sharing and openness of land and resources data in China.At the same time,the rise of government Weibo has caused more and more land and resources management departments to publish the latest progress in the field of land and resources to the public through Weibo.Therefore,Weibo provides a large amount of real and reliable,latest research data for land and resources research.Named entity recognition is the basis for natural language processing tasks such as information extraction,intelligent question answering,and public opinion monitoring.Named entity recognition in land and resources can extract key information from massive land and resources data and construct structured data,which is of great significance to the construction of land and resources information.In the existing research,there is no named entity recognition for the land and resources field.Due to the lack of a systematic entity system and the complexity of the entity type,using general domain named entity recognition system and the existing models are not applicable to the land and resources field.Therefore,in the context of increasing the spread and influence of land and resources informatization construction and land and resources government microblogs,this paper focuses on three problems.First,based on the deficiencies in the openness and sharig of land and resources data in research,this paper takes land and resources data in Weibo as the research object and constructs a corpus in the field of land and resources based on Weibo.Second,based on the lack of named entity system in land and resources field,this paper builds a recognition system of named entities in land and resources field by the corpus data characteristics.Third,aiming at the complex type of named entity system in land and resources field and the inability to directly apply the existing model methods,this paper based on the rules and constitutional characteristics of different named entities,combined with the idea of integrated learning and deep learning models,proposing a hybrid strategydriven identification model for named entity in the land and resources field.After the model construction is completed,the effectiveness of the hybrid strategy-driven naming entity recognition model proposed in this paper is experimentally studied.The experimental data source is the Weibo-based corpus of land and resources field constructed in this paper,with a total of 10526 data.The experimental results are divided into two dimensions.First,the total number and proportion of seven types of named entities is obtained at the overall level.It is found that the geographical and organization named entities are the most,accounting for about 2/3 of the total number of entities.The names with the highest frequency of various types of named entities are geographical entity "China",organization entity "Ministry of Natural Resources",time entity "2019",geological disaster entity "earthquake".Otherwise,"land","cultivated land","mines" are the most of land and resources names.Policy entity and name entity due to the variety of forms and the repetition rate is low,so there is no prominent performance in the word frequency word cloud of all entities.Second,by getting randomly sample from the corpus,this paper use the accuracy rate,recall rate,and F1 value to evaluate the model.The final accuracy rate is 90.93%,the recall rate is 87.84%,and the F1 value is 89.36%.The model proposed by this paper has a good recognition effect on the named entity recognition in land and resources field.