Sentiment Analysis on Mixed-Languages Customer Reviews: A Hybrid Deep Learning Approach
Ashok Kumar Durairaj, Anandan Chinnalagu · 2021
Machine Learning (ML), Natural Language processing (NLP) techniques are used for Sentiment Analysis (SA) problems. Many study results shown predicting more accurate sentiment in mixed language customer reviews are remains challenge. Customer sentiments are classified as positive negative and neutral. Many businesses providing products and services rely on reviews from customers through social media to improve customer experience and increase revenue. Based on our SA research studies and experiments, deep learning neural networks and fastText sequential models are shown better performance and more accurate sentiment prediction in text datasets compare to traditional classification algorithms. In this experiment we build a novel hybrid deep learning SA model using fastText word embedding library and authors multi-layer SAB-LSTM model called fastText-SA-BLSTM. Compare the hybrid model result with traditional linear SVM (LSVM), fastText and authors SAB-LSTM models. For this experiment, data has been collected from online customer reviews and social media posts. This experiment result shows the proposed hybrid model outperforms traditional models and shows more accurate context-based sentiment results. We conclude this paper with more details of our findings and experimental results as well as a proposed future work.