情报科学 ›› 2021, Vol. 39 ›› Issue (12): 72-79.

• 理论研究 • 上一篇    下一篇

大数据背景下主流融媒体热点发现机制研究

  

  • 出版日期:2021-12-01 发布日期:2021-12-29

  • Online:2021-12-01 Published:2021-12-29

摘要: 【目的/意义】为提升主流融媒体意识形态建设和舆论引导能力,解决大数据时代背景下主流融媒体多模态
信息资源管理的困境,构建高效的热点发现机制。【方法/过程】笔者着眼于主流融媒体热点发现需求构建需求体
系,然后利用Scrapy-Redis框架、HBase数据库和MapReduce实现了数据的精准采集、有序存储和高效处理,再基于
多模态信息融合的理念,借助 NLP技术对信息资源的特征进行提取,最后利用 LDA2vec模型和 Single-Pass算法实
现了信息归集和热点的发现与更新。【结果/结论】仿真实验结果表明,本研究所使用的方法,能够较好地实现多模
态信息的归集和热点的提取,效果较同类模型有明显提升。【创新/局限】但是在运用NLP技术处理多模态信息时各
处理环节的衔接尚不够流畅,后续仍需进行改进提升。

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.