Research On Named Entity Recognition For Chinese Social Media Base On Unified Model

With the development of information age,people urgently need to quickly understand the information from the vast amount of information texts,Named entity recognition technology is born.Named Entity Recognition(NER)refers to the classification and identification of quantitative phrases such as the name,organization name,place name,and meaningful time and date from the texts.For many text mining tasks,NER system is an important component,the research of NER model to enhance the recognition effect has academic and practical value.This paper focuses on how to improve the effect of NER under the condition of massive unlabeled texts.This paper propose a UNER(unified model of NER)in Chinese Social,The unified model consists of a cross-domain learning model and a semi-supervised learning model with by weight.Cross-domain learning needs to determine the similarity between sentences in the common domain and the target domain.The learning rate function is used to automatically adjust the learning rate of sentences in different fields.Combined with transfer probability algorithm,so as to achieve the adaptation between the auxiliary area and the target area,Improve the generalization of cross-domain learning model.The semi-supervised learning model combines self-train and autonomous learning for Chinese NRE,adopts the confidence function to adjust the learning rate,continuously iteratively extracts samples from the unlabeled corpus in the target domain and adds them to the training set for training.By actively learning the unlabeled information in the target domain,the workload of manual annotation is greatly reduced,making the self-train of large-scale data feasible.The experimental results show that the unified model improves the effectiveness of NER in Chinese social media.Unified model combines cross-domain learning model and semi-supervised learning model.The unified model greatly reduces the work of annotated corpus and improves the recognition effect of NER in Chinese social media.