Purpose
The purpose of this paper is to analyze the correlation between the Twitter activity of two airline companies and their stock performance at the Istanbul Stock Exchange (BIST).
Design/methodology/approach
Overall, 113,018 tweets were divided into 34,152 semantic and 78,866 share tweets. Semantic tweets are tweets mentioning company’s products or services and were labeled manually and with deep learning models. Share tweets were divided into 13,618 relevant and 65,248 irrelevant tweets.
Findings
A positive correlation was found between share tweets and stock performance. Semantic tweets did not display a correlation with stock performance. Relevant share tweets displayed as a strong correlation as all share tweets for one company. Also, the manual labeling of 8,000 tweets led to the discovery of many insights related to service provision in the airway industry, management of digital support channels, management of reputation on social media and using Twitter as a customer support platform.
Practical implications
Relevant share tweets comprise only 20% of all share tweets for one company and show the same level of correlation with stock performance. This means that the efficiency of business intelligence solutions created to monitor Twitter activity can be improved five times by saving computational power, network bandwidth and data storage.
Originality/value
Previous research has analyzed all Twitter activity taken together. By dividing tweets into semantic and share tweets, this paper illustrates that it is, in fact, share tweets that are correlated with stock performance and not semantic tweets.