SSN_MLRG1 at SemEval-2022 Task 10: Structured Sentiment Analysis using 2-layer BiLSTM

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 1324 - 1328 July 14-15, 2022 ©2022 Association for Computational Linguistics SSN_MLRG1 at SemEval-2022 Task 10: Structured Sentiment Analysis using 2-layer BiLSTM Karun Anantharaman, Divyasri K, Jayannthan PT Ms. S.Angel Deborah , Ms. S. Rajalakshmi, Dr. R.S. Milton, Dr. T. T. Mirnalinee Department of Computer Science and Engineering SSN College of Engineering Chennai 603 110, Tamil Nadu, India karun19049@cse.ssn.edu.in, {divyasri2011037,jayannthan2010606}@ssn.edu.in, {rajalakshmis, angeldeborahs}@ssn.edu.in, {miltonrs, mirnalineett}@ssn.edu.in Abstract Task 10 in SemEval 2022 is a composite task which entails analysis of opinion tuples, and recognition and demarcation of their nature. In this paper, we will elaborate on how such a methodology is implemented, how it is un- dertaken for a Structured Sentiment Analysis, and the results obtained thereof. To achieve this objective, we have adopted a bi-layered BiLSTM approach. In our research, a varia- tion on the norm has been effected towards enhancement of accuracy, by basing the cate- gorization meted out to an individual member as a by-product of its adjacent members, using specialized algorithms to ensure the veracity of the output, which has been modelled to be the holistically most accurate label for the entire sequence.Such a strategy is superior in terms of its parsing accuracy and requires less time. This manner of action has yielded an SF1 of 0.33 in the highest-performing configuration. 1 Introduction Sentiment Analysis is a specialized area under Natural Language Processing which deals with the extraction of opinions and emotions from text which may include reviews, social media posts, forums or news. Sentiment Analysis has become a powerful tool for detecting the pub- lic opinion on any topic. This research article evinces how the senti- ments and opinions expressed by people in various environments could be understood by means of a structured sentiment analysis (SSA) approach. There are conventional means to this end, in transition-based and graph-based techniques, but a potent alternative is the novel way to envision dependency parsing tasks as algorithmic pattern-recognition by way of se- quence labelling. Sequence labeling is essen- tially an agglomeration of multiple discrete categorizations, extended on to each element of the entire sequence. The model is trained on 7 seven data sets with 5 different languages En- glish, Spanish, Basque, Norwegian and Cata- lan. In Subtask-A we try to separately identify the holder and target in the text, along with the sentiment expression used using a Sequence Labelling Approach. In Subtask-B we predict whether the ex- tracted tokens have a relation or the lack of one. Subtask-C involves the concatenation of the results of Subtask-A and Subtask-B to provide the final predictions. 2 Background 2.1 Definitions The following contains descriptions of the models made use of, and related terminology. Namely we will define LSTM, BiLSTM, Sigmoid, Linear, Max Pooling Layers and Word Embeddings and the BIO Sequence Labelling approach. Long Short-Term Memory(LSTM) 1324 Speech-recognition language models, among numerous other activities, have seen encour- aging levels of success by availing Recurrent Neural Networks (RNN) (Mikolov et al., 2010, 2011) …

Content truncated to 500 words