
Research Article
Aspect Level Sentiment Analysis to Extract Valuable Insight for Airline’s Customer Feedback and Reviews
@INPROCEEDINGS{10.1007/978-3-031-48888-7_10, author={Bharat Singh and Nidhi Kushwaha}, title={Aspect Level Sentiment Analysis to Extract Valuable Insight for Airline’s Customer Feedback and Reviews}, proceedings={Cognitive Computing and Cyber Physical Systems. 4th EAI International Conference, IC4S 2023, Bhimavaram, Andhra Pradesh, India, August 4-6, 2023, Proceedings, Part I}, proceedings_a={IC4S}, year={2024}, month={1}, keywords={Sentiment Analysis Decision Making Airline Data Social Media BERT Machine Learning}, doi={10.1007/978-3-031-48888-7_10} }
- Bharat Singh
Nidhi Kushwaha
Year: 2024
Aspect Level Sentiment Analysis to Extract Valuable Insight for Airline’s Customer Feedback and Reviews
IC4S
Springer
DOI: 10.1007/978-3-031-48888-7_10
Abstract
In the realm of decision-making, the internet plays a vital and pervasive role, serving as a conduit for individuals worldwide to express their perspectives and viewpoints through various online platforms such as blogs and social media. Consequently, the internet has become inundated with a vast array of both pertinent and extraneous information, presenting a formidable challenge in sifting through the abundance of content to extract the desired information. Sentiment analysis emerges as a valuable tool for addressing this issue, enabling the systematic analysis of each document to discern the prevailing sentiment expressed within. This holds particular relevance in the realm of customer decision-making, as it empowers individuals to make informed choices when selecting the most suitable US airline by evaluating the opinions shared by other customers on online review platforms like Skytrax and micro-blogging sites such as Twitter. We can use these kinds of datasets to provides the aspect level sentiment analysis. Therefore, we have explored, in this article, a language model built upon a pretrained deep neural networks capable of analyzing the sequence of text to classify it as having positive, negative or neutral emotions without explicit human labelling. To analyze and assess these models, data from Twitter’s US airlines sentiment database was used. Experiment on above data set show BERT model to be superior in accuracy while being more significant in less time to train. We observe notable advancements over prior state-of-the-art methods that use supervised feature learning to close the gap.