Document Type : Research Paper

Authors

• Ghazaleh Soltani ¹
• ameneh khadivar ²
• neda paravian ¹

¹ Department of management, Faculty of Social Sciences and Economics Alzahra University, Tehran, Iran

² Alzahra University

Abstract

With the expansion of the internet and the emergence of social media, airlines are increasingly using these platforms to engage with customers. Twitter, as one of the most popular microblogging platforms, plays a key role in communication between airlines and passengers. This study aims to analyze sentiments and topics related to Iranian airlines on Twitter, based on 2,876 English tweets collected between April 2008 and March 2024 using Playwright. After preprocessing and interpreting 230 emojis with artificial intelligence, sentiment labeling was performed using both lexicon-based and machine learning approaches (RoBERTa model). Among the five evaluated algorithms, logistic regression with the lexicon-based method achieved the best performance, with an F-score of 0/81. Incorporating emoji information significantly improved model accuracy. Topic modeling with LDA revealed that “safety” was the primary concern of users. These findings can contribute to optimizing the communication and service strategies of Iranian airlines.

Keywords

• Sentiment analysis
• topic modeling
• Twitter
• deep learning
• lexicon-based approach

Main Subjects

• business management