Automatic Stance Detection Using End-to-End Memory Networks

Proceedings of NAACL-HLT 2018, pages 767–776 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Automatic Stance Detection Using End-to-End Memory Networks Mitra Mohtarami1, Ramy Baly1, James Glass1, Preslav Nakov2 Lluís Màrquez3∗, Alessandro Moschitti3∗ 1MIT Computer Science and Artificial Intelligence Laboratory, Cambridge, MA, USA 2Qatar Computing Research Institute, HBKU, Doha, Qatar; 3Amazon {mitra,baly,glass}@csail.mit.edu pnakov@hbku.edu.qa; {lluismv,amosch}@amazon.com Abstract We present an effective end-to-end memory network model that jointly (i) predicts whether a given document can be considered as rele- vant evidence for a given claim, and (ii) ex- tracts snippets of evidence that can be used to reason about the factuality of the target claim. Our model combines the advantages of con- volutional and recurrent neural networks as part of a memory network. We further in- troduce a similarity matrix at the inference level of the memory network in order to ex- tract snippets of evidence for input claims more accurately. Our experiments on a pub- lic benchmark dataset, FakeNewsChallenge, demonstrate the effectiveness of our approach. 1 Introduction Recently, an unprecedented amount of false infor- mation has been flooding the Internet with aims ranging from affecting individual people’s beliefs and decisions (Mihaylov et al., 2015; Mihaylov and Nakov, 2016) to influencing major events such as political elections (Vosoughi et al., 2018). Con- sequently, manual fact checking has emerged with the promise to support accurate and unbiased anal- ysis of rumors spreading in social medias, as well as of claims made by public figures or news media. As manual fact checking is a very tedious task, automatic fact checking has been proposed as a possible alternative. This is often broken into in- termediate steps in order to alleviate the task com- plexity. One such step is stance detection, which is also useful for human experts as a stand-alone task. The task aims to identify the relative per- spective of a piece of text with respect to a claim, typically modeled using labels such as agree, dis- agree, discuss, and unrelated; Figure 1 gives some examples. ∗Work conducted while these authors were at QCRI. Claim: Robert Plant Ripped up $800M Led Zeppelin Re- union Contract. Stance Snippet agree Led Zeppelin’s Robert Plant turned down £500m to reform supergroup... disagree Robert Plant’s publicist has described as “rub- bish” a Daily Mirror report that he rejected a £500m Led Zeppelin reunion... discuss Robert Plant reportedly tore up an $800 million Led Zeppelin reunion deal... unrelated Richard Branson’s Virgin Galactic is set to launch SpaceShipTwo today... Figure 1: Examples of snippets of text and their stance with respect to the target claim. Here, we address the problem of stance detection using a novel model based on end-to-end memory networks (Sukhbaatar et al., 2015), which incorpo- rates convolutional and recurrent neural networks, as well as a similarity matrix. Our model jointly addresses the problems of predicting the stance of a text with respect to a given claim, and of extract- ing relevant text snippets as support for the predic- tion of the model. We …

Content truncated to 500 words