IITP at SemEval-2017 Task 8 : A Supervised Approach for Rumour Evaluation

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 497–501, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics IITP at SemEval-2017 Task 8 : A Supervised Approach for Rumour Evaluation Vikram Singh, Sunny Narayan, Md Shad Akhtar, Asif Ekbal, Pushpak Bhattacharyya Indian Institute of Technology Patna, India {vikram.mtcs15,sunny.cs13,shad.pcs15,asif,pb}@iitp.ac.in Abstract This paper describes our system partici- pation in the SemEval-2017 Task 8 ‘Ru- mourEval: Determining rumour veracity and support for rumours’. The objec- tive of this task was to predict the stance and veracity of the underlying rumour. We propose a supervised classification ap- proach employing several lexical, content and twitter specific features for learning. Evaluation shows promising results for both the problems. 1 Introduction Twitter along with Facebook is widely used social networking site which generates tons of authen- tic and unauthentic information. The purpose of twitter varies from people to people. Twitter has been greatly used as a communication channel and also as an information source (Zhao and Rosson, 2009). However, Twitter like any other social me- dia platform does not always poses authentic in- formation. It also brings a negative by-product called rumour (Castillo et al., 2011; Derczynski and Bontcheva, 2014; Qazvinian et al., 2011). Ru- mours are the statement which cannot be verified for its correctness. These rumours may confuse people with the unverified information and drive them in poor decision making. In many organiza- tions(political, administration etc.), detection and support for rumour invites great interest from the concerned authorities. Recently, researchers across the globe have started addressing the challenges related to ru- mours. A time sequence classification technique has been proposed for detecting the stance against a rumor (Lukasik et al., 2016). Zubiaga et al. (2016) used sequence of label transitions in tree- structured conversations for classifying stance. A study on speech act classifier for veracity predic- tion is proposed in (Vosoughi, 2015). One of the earlier work reported on rumour detection and classification had used twitter specific and content based features for the prediction (Qazvinian et al., 2011). In this paper we present our proposed system submitted as part of the SemEval-2017 shared task on “RumourEval: Determining rumour veracity and support for rumours”. Our system is super- vised in nature and uses a diverse set of features (c.f. Section 2.3) for training. The task involves Twitter conversation thread where for every source tweet a number of direct and nested reply tweets are present. An example thread is depicted in Ta- ble 1. The task defines two separate sub-problems: A) Support, Deny, Query & Comment (SDQC) classification and B) veracity prediction. The first subtask checks the stance of any tweet(source or reply) w.r.t. the underlying rumour. Reply tweet can be direct or nested. Second subtask predicts the veracity of a rumour i.e. true (rumour), false (not rumour) or unverified (its veracity cannot be verified). Further, there were two variants of the veracity task: closed and open variants. In closed variant, the veracity prediction has to be made solely from the …

Content truncated to 500 words