Event Detection and Summarization in Weblogs with Temporal Collocations

Event Detection and Summarization in Weblogs with Temporal Collocations Chun-Yuan Teng and Hsin-Hsi Chen Department of Computer Science and Information Engineering National Taiwan University Taipei, Taiwan {r93019, hhchen}@csie.ntu.edu.tw Abstract This paper deals with the relationship between weblog content and time. With the proposed temporal mutual information, we analyze the collocations in time dimension, and the interesting collocations related to special events. The temporal mutual information is employed to observe the strength of term-to-term associations over time. An event detection algorithm identifies the collocations that may cause an event in a specific timestamp. An event summarization algorithm retrieves a set of collocations which describe an event. We compare our approach with the approach without considering the time interval. The experimental results demonstrate that the temporal collocations capture the real world semantics and real world events over time. 1. 2. Introduction Compared with traditional media such as online news and enterprise websites, weblogs have several unique characteristics, e.g., containing abundant life experiences and public opinions toward different topics, highly sensitive to the events occurring in the real world, and associated with the personal information of bloggers. Some works have been proposed to leverage these characteristics, e.g., the study of the relationship between the content and bloggers’ profiles (Adamic & Glance, 2005; Burger & Henderson, 2006; Teng & Chen, 2006), and content and real events (Glance, Hurst & Tornkiyo, 2004; Kim, 2005; Thelwall, 2006; Thompson, 2003). In this paper, we will use temporal collocation to model the term-to-term association over time. In the past, some useful collocation models (Manning & Schütze, 1999) have been proposed such as mean and variance, hypothesis test, mutual information, etc. Some works analyze the weblogs from the aspect of time like the dynamics of weblogs in time and location (Mei, et al., 2006), the weblog posting behavior (Doran, Griffith & Henderson, 2006; Hurst, 2006), the topic extraction (Oka, Abe & Kato, 2006), etc. The impacts of events on social media are also discussed, e.g., the change of weblogs after London attack (Thelwall, 2006), the relationship between the warblog and weblogs (Kim, 2005; Thompson, 2003), etc. This paper is organized as follows. Section 2 defines temporal collocation to model the strength of term-to-term associations over time. Section 3 introduces an event detection algorithm to detect the events in weblogs, and an event summarization algorithm to extract the description of an event in a specific time with temporal collocations. Section 4 shows and discusses the experimental results. Section 5 concludes the remarks. Temporal Collocations We derive the temporal collocations from Shannon’s mutual information (Manning & Schütze, 1999) which is defined as follows (Definition 1). Definition 1 (Mutual Information) The mutual information of two terms x and y is defined as: ) ( () ) (, )log (, ) (, PxPy Pxy Pxy Ixy = where P(x,y) is the co-occurrence probability of x and y, and P(x) and P(y) denote the occurrence probability of x and y, respectively. Following the definition of mutual information, we derive the temporal mutual information modeling …

Content truncated to 500 words