Bi-Lingual (English, Punjabi) Sarcastic Sentiment Analysis by using Classification Methods

Abstract

Sentiment analysis is one of the heated topic in the field of text mining. As the social media data is increased day by day the main need of the data scientists is to classify the data so that it can be further used for decision making or knowledge discovery. Now –a-days everything and everyone available online so to check the latest trends in business or in daily life one must consider the online data. The main focus of sentiment analysis is to focus on positive or negative comments so that a well define picture is created that what is trending or not but the sarcasm manipulates the data as in sarcastic comment negative comment consider as positive because of the presence of positive words in the comment or data so it is necessary to detect the sarcasm in online data . The data on social media is available in various languages so sentiment analysis in regional languages is also a main step . In the proposed work we focus on two languages i.e Punjabi and English. Here we use deep learning based neural networks for the sarcasm detection in English as well as Punjabi language. In the proposed work we consider three datasets i.e. balanced English dataset, Balanced Punjabi Dataset and unbalanced Punjabi dataset. We used six different models to check the accuracy of the classified data the models we used are LSTM with word embedding layer, BiLSTM with , LSTM+LSTM, BiLSTM+BiLSTM, LSTM+BiLSTM, CNN respectively. LSTM provide better accuracy for balanced Punjabi and English dataset i.e. 95.63% and 94.17% respectively. The accuracy for unbalanced Punjabi dataset is provided by BiLSTM i.e.96.31%.