Abstract: 【Purpose/significance】In order to improve the ideological construction and public opinion guidance ability of mainstream
media convergence and solve the dilemma of multimodal information resource management of mainstream media convergence in the
era of big data,build an efficient hot spot discovery mechanism.【Method/process】Focusing on the hot spot discovery needs of main?
stream media convergence,the author constructs the demand system,and then realizes the accurate collection,orderly storage and effi-
cient processing of data by using the Scrapy-Redis framework,HBase database and MapReduce.Based on the concept of multimodal in-
formation fusion, the author extracts the characteristics of information resources with the help of NLP technology, Finally, LDA2vec
model and single-pass algorithm are used to realize information collection and hot topics discovery and update.【Result/conclusion】
The simulation results show that the method used in this study can better realize the collection of multimodal information and the ex-
traction of hot spots,and the effect is significantly improved compared with similar models【. Innovation/limitation】However,when using NLP technology to process multi-modal information,the connection of each processing link is not smooth enough,and it still needs to be improved in the follow-up.