Research On Hot Spot Discovery In The Automotive Field Based On Improved BiLSTM-CRF

In the era of big data,data in various fields has shown explosive growth.Obtaining valuable information from massive data is the subject of continuous research by scholars in various fields in recent years.At the same time,with the improvement of people’s living standards,cars have gradually entered thousands of households.The increasing number of automobile users has prompted major domestic automobile manufacturers to improve the performance of their own automobile products to improve their core competitiveness in the automobile market.Therefore,it is very important for automobile manufacturers to collect the discussion directions of automobile users and grasp the focus of relevant researchers on automobile,so as to improve product development programs and capture the optimization direction of automobile products.Based on this,this paper mainly focuses on the hot topic discovery in the automotive field.The innovation work includes the following three aspects.(1)The modeling based on automobile patent text is proposed.Collect the data set of automobile patent text,and model the patent title,patent abstract and other data according to the characteristics of automobile patent text,retain the professional terminology and technical information in the patent text,and prepare the data for the subsequent entity recognition and hot spot discovery;(2)An improved entity recognition framework for BiLSTM-CRF is proposed.Combined with the large-scale semantic representation ability of the BERT pre-trained language model,the context information modeling ability of the BiLSTM deep learning model and the output sequence annotation ability of the CRF machine learning model,the entity recognition in the automotive field is performed;(3)An improved weighted word similarity calculation method is proposed.The TFIDF model is used to extract keywords from the document set,and the improved weighted word similarity algorithm is used to calculate the similarity between the entity recognition result and the extracted keywords to achieve hot spot discovery.Through the entity recognition comparison experiment on the data set and the analysis of the hot spot discovery results,the effectiveness of the proposed algorithm is effectively verified,which provides some reference for related enterprises to improve product performance and grasp further research and development directions.