Topical Stance Detection for Twitter: A Two-Phase LSTM Model Using Attention

Abstract

The topical stance detection problem addresses detecting the stance of the text content with respect to a given topic: whether the sentiment of the given text content is in FAVOR of (positive), is AGAINST (negative), or is NONE (neutral) towards the given topic. Using the concept of attention, we develop a two-phase solution. In the first phase, we classify subjectivity -whether a given tweet is neutral or subjective with respect to the given topic. In the second phase, we classify sentiment of the subjective tweets (ignoring the neutral tweets) -whether a given subjective tweet has a FAVOR or AGAINST stance towards the topic. We propose a Long Short-Term memory (LSTM) based deep neural network for each phase, and embed attention at each of the phases. On the SemEval 2016 stance detection Twitter task dataset [7], we obtain a best-case macro F-score of 68.84% and a best-case accuracy of 60.2%, outperforming the existing deep learning based solutions. Our framework, T-PAN, is the first in the topical stance detection literature, that uses deep learning within a two-phase architecture.

Your Stance is Exposed! Analysing Possible Factors for Stance Detection on Social Media

“…Studying stance needs to cover the intersection dimensions of stance taking process, which are mainly influenced by linguistic forms and social interactions frames [37]. Most of the previous studies define stance as a textual entailment task where the main processing depends on the raw text only [5,16,40,42,45]. In this form of stance detection, a given text entails a stance towards a premise (target).…”

“…[19] proposed attention-based neural network by using a target-specific information which produced an overall F-score of 68.79%. Another marginal improvement introduced by [16] by using attention based LSTM model, which achieved 68.84% F-score. Whereas in [57] the usage of bi-directional GRU-CNN yielded an F-score of 69.42%.…”

Early Detection of Rumours on Twitter via Stance Transfer Learning

“…Towards answering these research questions, we made several contributions. To address the issue of lack of attitude annotations for user comments, we propose CNN-and BERT-based deep neural models to learn attitude representation from user comments via transfer learning from resources for stance prediction [1,5,20,24,29]. We further propose CNN-BiLSTM and BERT neural models to integrate attitude representation and content representation for tweets and their comments for rumour detection.…”

“…Stance detection [1,5,20,24,29] aims to automatically detect user attitudes towards given posts, whether the user is in favour of, against or neutral toward the target post. Some deep neural models are proposed for the task and achieve reasonable performance [1,5,29].…”

“…We propose to formulate the task of rumour detection as a transfer learning problem. To achieve rumour classification for a tweet message s i where user comments C i do not have attitude annotation, we propose to learn representation for attitudes for user comments via transfer learning based on the readily available annotated resources for stance prediction in the literature [1,5,20,24,29]. The idea is to pretrain a model on the stance data source to learn stance representation and then transfer and integrate this knowledge to the neural model for tweet and comment contents for rumour detection.…”