Research On Multimodal Marketing News Recognition Method Based On Deep Learning

News is an integral part of human cultural life.In the past,news was mainly about events that are closely related to life.However,in recent years,with the explosive increase of e-commerce,marketing news has become more and more intensive and interfered with people's needs to watch the news.How to effectively and accurately News that identifies marketing intent is a key research content of major news sites.With the development of natural language processing technology,the application of deep learning technology to classification and recognition has become one of the research hotspots in recent years.Most of the traditional machine learning algorithms use manual methods to extract features,so there are certain limitations on the text classification problem,and deep neural networks can better express the nature of the data through multi-level structure transformation.This paper fuses models based on deep learning technology,and combines multi-modal classification and recognition with news maps.The main research contents are as follows:(1)According to the characteristics of large amount of news text data and different lengths,and through experimental analysis of convolutional neural network and long-term short-term memory neural network in text classification technology,to solve the problems in these two methods,the character level The way that the representation vector is connected with the word embedding vector.A CNN +LSTMAttention network model structure is proposed.A convolutional neural network can represent words at the character level.Long-term and short-term memory neural networks can better solve the problem of text serialization.Note that The force mechanism can assign weights before the model is output,which is helpful to improve the interpretability of the deep learning model.Through comparative experimental results analysis,the CNN + LSTMAttention model proposed in this paper improvesthe accuracy of classification and recognition of marketing news.(2)In the process of classification and recognition of news text,too little text content of a single news or marketing ambiguity of text information will affect the accuracy of recognition,so the text proposed a multi-mode of news text combined with news map Marketing news identification method.The implementation method uses the OCR technology process proposed in this article for the news map,including CTPN to locate the text area of the map,and design Dense Net combined with CTC technology for text recognition,and extract the text information contained in the news map to supplement the text classification.Identify information to achieve multi-modal effects.Comparative experimental results with news text only show that The multimodal marketing news recognition method used in this paper has higher accuracy and better classification effect.