Amrita School of Engineering - CSE at SemEval-2019 Task 6: Manipulating Attention with Temporal Convolutional Neural Network for Offense Identification and Classification

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 540–546 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 540 Amrita School of Engineering - CSE at SemEval-2019 Task 6: Manipulating Attention with Temporal Convolutional Neural Network for Offense Identification and Classification Murali Sridharan, Swapna T R Department of Computer Science and Engineering Amrita School of Engineering, Coimbatore, Amrita Vishwa Vidhyapeetham, India muralisridharan.10@gmail.com,tr swapna@cb.amrita.edu Abstract With the proliferation and ubiquity of smart gadgets and smart devices, across the world, data generated by them has been growing at exponential rates, in particular social media platforms like Facebook, Twitter and Insta- gram have been generating voluminous data on a daily basis. According to Twitter’s us- age statistics, about 500 million tweets are generated each day. While the tweets reflect the users’ opinions on several events across the world, there are tweets which are offen- sive in nature that need to be tagged under the hateful conduct policy of Twitter. Of- fensive tweets have to be identified, captured and processed further, for a variety of reasons, which include i) identifying offensive tweets in order to prevent violent/abusive behaviour in Twitter (or any social media for that mat- ter), ii) creating and maintaining a history of offensive tweets for individual users (would be helpful in creating meta-data for user pro- file), iii) inferring the sentiment of the users on particular event/issue/topic . We (CodaLab Team/User Name: murali sr) have em- ployed neural network models which manip- ulate attention with Temporal Convolutional Neural Network for the three shared sub- tasks i) ATT-TCN (ATTention based Temporal Convolutional Neural Network) employed for shared sub-task A that yielded a best macro- F1 score of 0.46, ii) SAE-ATT-TCN(Self At- tentive Embedding-ATTention based Tempo- ral Convolutional Neural Network) employed for shared sub-task B and sub-task C that yielded best macro-F1 score of 0.61 and 0.51 respectively. Among the two variants ATT- TCN and SAE-ATT-TCN, the latter performed better. 1 Introduction In the prevailing digital era, Deep Learning has penetrated almost all industry verticals and afforded several researchers an effective tool, in handling voluminous data and deriving mean- ingful inferences. Initially, (LeCun et al., 1998) invented Convolutional Neural Network (CNN) model for extraction of local features, which later proved to be the standard choice for Computer Vision tasks. (Hochreiter and Schmidhuber, 1997) the introduced LSTM (Long Short Term Mem- ory) architecture, which went on to become the standard choice for Natural Language Processing (sequence) tasks due to the implicit ordering of the sequence data in words and sentences. Then several architectures, combining LSTM with CNN were introduced that went on to become successful for NLP tasks as well. Deep Learning techniques have leaped forward through multiple NLP tasks such as Modeling, Classification, Translation, Summarization, etc., and have proved to be better compared to traditional techniques. Ever since social media has become ubiquitous there have been individuals who take gratuitous advantage of the anonymous nature of social me- dia platforms, and engage themselves in rude and offensive communications. …

Content truncated to 500 words