Automatic Detection and Classification of Social Events

Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 1024–1034, MIT, Massachusetts, USA, 9-11 October 2010. c⃝2010 Association for Computational Linguistics Automatic Detection and Classification of Social Events Apoorv Agarwal Department of Computer Science Columbia University New York, U.S.A. apoorv@cs.columbia.edu Owen Rambow CCLS Columbia University New York, U.S.A. rambow@ccls.columbia.edu Abstract In this paper we introduce the new task of social event extraction from text. We distin- guish two broad types of social events depend- ing on whether only one or both parties are aware of the social contact. We annotate part of Automatic Content Extraction (ACE) data, and perform experiments using Support Vec- tor Machines with Kernel methods. We use a combination of structures derived from phrase structure trees and dependency trees. A char- acteristic of our events (which distinguishes them from ACE events) is that the participat- ing entities can be spread far across the parse trees. We use syntactic and semantic insights to devise a new structure derived from depen- dency trees and show that this plays a role in achieving the best performing system for both social event detection and classification tasks. We also use three data sampling approaches to solve the problem of data skewness. Sam- pling methods improve the F1-measure for the task of relation detection by over 20% abso- lute over the baseline. 1 Introduction This paper introduces a novel natural language pro- cessing (NLP) task, social event extraction. We are interested in this task because it contributes to our overall research goal, which is to extract a social network from written text. The extracted social net- work can be used for various applications such as summarization, question-answering, or the detection of main characters in a story. For example, we man- ually extracted the social network of characters in Alice in Wonderland and ran standard social network analysis algorithms on the network. The most influ- ential characters in the story were correctly detected. Moreover, characters occurring in a scene together were given same social roles and positions. Social network extraction has recently been applied to lit- erary theory (Elson et al., 2010) and has the potential to help organize novels that are becoming machine readable. We take a “social network” to be a network con- sisting of individual human beings and groups of hu- man beings who are connected to each other by the virtue of participating in social events. We define social events to be events that occur between peo- ple where at least one person is aware of the other and of the event taking place. For example, in the sentence John talks to Mary, entities John and Mary are aware of each other and the talking event. In the sentence John thinks Mary is great, only John is aware of Mary and the event is the thinking event. In the sentence Rabbit ran by Alice there is no evi- dence about the cognitive states of Rabbit and Alice (because the Rabbit could have run by Alice without …

Content truncated to 500 words