Comparative analysis of ML and DL techniques on aspect-based sentiment analysis

The user generated data on the web from blogs, web forums and social networking sites has become so enormous and extracting useful information from them has become an interesting research topic in recent years. Aspect Based Sentiment Analysis (ABSA) detect the feature which is nothing but aspect from the information and then apply sentiment analysis on that feature. The subtask of ABSA is Aspect Extraction, assigning Polarity to Aspects and Aspect and Aspect Sentiment Classification. Initially various Machine learning algorithms like Conditional Random Forest (CRF), Naive bayes, Decision Trees and Support Vector Machine (SVM) were used to implement the subtask of ABSA. Then with the introduction of deep learning algorithms like RNN, LSTM and transformers the efficacy of Aspect based Sentiment analysis were improved. In this paper we provide a comparative study between ML and DL approaches to improve overall performance of Aspect based Sentiment analysis.