International Conference on Computer and Knowledge Engineering

Home / 13th International Conference on Computer and Knowledge Engineering

BERT transformers Multitask learning Sarcasm and Sentiment classification (BMSS)

Authors :

Fatemeh Molavi¹ Jamshid Bagherzadeh Mohasefi²

1- Urmia University 2- Supervisor

Keywords :

Sentiment Classification،Sarcasm Detection،Multi Task Learning،Transformers

Abstract :

Numerous researches were reported based on analyzing the sentiment of texts, which are becoming considerably vital to discover the sentiment in various domains of economic, marketing, politic, and etc. However, one of the most challenging aspects of Natural Language Processing (NLP) is discovery of the sarcasm. The sarcasm and the sentiment as similar subjects have been explored in various articles. In this research, two significant achievements were employed. Firstly, the pre-trained BERT transformer model was used to transfer its knowledge to our model. Then, fine tuned BERT model was combined with a machine learning approach, specifically Multitask learning (MTL). By establishing a connection between sarcasm and sentiment, we observed a simultaneous enhancement in both sarcasm detection and sentiment analysis performance. The achieved accuracy results (average of 0.97) were deemed satisfactory and commendable for both tasks.