Jointly Event Extraction and Visualization on Twitter via Probabilistic Modelling

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, pages 269–278, Berlin, Germany, August 7-12, 2016. c⃝2016 Association for Computational Linguistics Jointly Event Extraction and Visualization on Twitter via Probabilistic Modelling Deyu Zhou†‡ Tianmeng Gao† Yulan He§ † School of Computer Science and Engineering, Key Laboratory of Computer Network and Information Integration, Ministry of Education, Southeast University, China ‡ State Key Laboratory for Novel Software Technology, Nanjing University, China § School of Engineering and Applied Science, Aston University, UK d.zhou@seu.edu.cn, gaotianmeng@seu.edu.cn, y.he@cantab.net Abstract Event extraction from texts aims to de- tect structured information such as what has happened, to whom, where and when. Event extraction and visualization are typ- ically considered as two different tasks. In this paper, we propose a novel approach based on probabilistic modelling to joint- ly extract and visualize events from tweet- s where both tasks benefit from each oth- er. We model each event as a joint dis- tribution over named entities, a date, a lo- cation and event-related keywords. More- over, both tweets and event instances are associated with coordinates in the visual- ization space. The manifold assumption that the intrinsic geometry of tweets is a low-rank, non-linear manifold within the high-dimensional space is incorporated in- to the learning framework using a regu- larization. Experimental results show that the proposed approach can effectively deal with both event extraction and visualiza- tion and performs remarkably better than both the state-of-the-art event extraction method and a pipeline approach for event extraction and visualization. 1 Introduction Event extraction, one of the important and chal- lenging tasks in information extraction, aims to detect structured information such as what has happened, to whom, where and when. The out- puts of event extraction could be beneficial for downstream applications such as summarization and personalized news systems. Data visualiza- tion, an important exploratory data analysis task, provides a simple way to reveal the relationships among data (Nakaji and Yanai, 2012). Although event extraction and visualization are two different tasks and typically studied separately in the literature, these two tasks are highly related. Documents which are close to each other in the low-dimensional visualization space are likely to describe the same event. Events in nearby loca- tions in the visualization space are likely to share similar event elements. Therefore, jointly learn- ing the two tasks could potentially bring benefits to each other. However, it is not straightforward to learn event extraction and visualization jointly since event extraction usually relies on semantic parsing results (McClosky et al., 2011) while vi- sualization is accomplished by dimensionality re- duction (Iwata et al., 2007; L´opez-Rubio et al., 2002). In this paper, we propose a novel probabilistic model, called Latent Event Extraction & Visual- ization (LEEV) model, for joint event extraction and visualization on Twitter. It is partly inspired by the Latent Event Model (LEM) (Zhou et al., 2015) where each tweet is assigned to one even- t instance and each event is modeled as a join- t distribution over named entities, a date/time, …

Content truncated to 500 words