Deep News Event Ranker Based On User Relevant Query

CASIA OpenIR > 多模态人工智能系统全国重点实验室 > 互联网大数据与信息安全

	Deep News Event Ranker Based On User Relevant Query
	Kong XF(孔祥飞)1,2 ; Kong QC(孔庆超)1 ; Mao WJ(毛文吉)1,2 ; Tang SQ(唐少强)3; Kong QC(孔庆超)
	2018-08
会议名称	International Conference on Cloud Computing and Big Data Analysis
页码	363-367
会议日期	April 20-22, 2018
会议地点	Chengdu, China
摘要	News Event Ranking(NER), which takes eventrelated news documents as the ranking unit, has been addressed in many research work and implemented in securityoriented applications(e.g. public event monitoring, mining and retrieval). Previous work solely rank news event based on event relevant information, while user relevant information equally important for characterizing news event is totally neglected. In this paper, we depict news event with extra user comments sentiment polarity information, and address news event ranking problem by incorporating user relevant information into the input query. Given an input query, which contains event related objective aspects(e.g. actors, locations, date) and user related subjective aspects(e.g. public attention and opinion polarity), we develop a Deep News Event Ranker model to integrate objective event information and subjective user information. Firstly, a semantic similarity interaction module transforms query keywords, news document and news comments to their semantic vector representation and calculates queryndocument similarity and queryncomment similarity. Then a Feature Extraction Based On CNNs and LSTM module extract query term importance features, query term frequency features and BM25-like relevance features for ranking. Finally, a Feature Aggregation module merges the extracted features with some auxiliary relevance features and produces a global relevance score. Experiments on a large news dataset demonstrate the eﬀectiveness of our proposed model compared to several baseline models.
关键词	News Event Ranker User Relevant Query User Related Subjective Aspects Deep News Event Ranker
收录类别	EI
语种	英语
文献类型	会议论文
条目标识符	http://ir.ia.ac.cn/handle/173211/21031
专题	多模态人工智能系统全国重点实验室_互联网大数据与信息安全
通讯作者	Kong QC(孔庆超)
作者单位	1.中国科学院大学 2.中科院自动化所 3.北京大学
第一作者单位	中国科学院自动化研究所
推荐引用方式 GB/T 7714	Kong XF,Kong QC,Mao WJ,et al. Deep News Event Ranker Based On User Relevant Query[C],2018:363-367.

条目包含的文件		下载所有文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
dnn_ranker_icccbda.p（87KB）	会议论文		开放获取	CC BY-NC-SA	浏览下载