Abstract
The wealth of electronically generated communication combined with increased computing power and sophisticated algorithms provides the opportunity for destination managers to listen to travellers. Identification of sentiment with a domain-oriented lexicon is beneficial for natural language processing to analyse public opinion. Indeed, in the context of travel, sentiment analysis enables tourism decision makers to devise marketing and development strategies that address the information learned. This study presents a lexical dictionary approach for sentiment extraction and opinion mining of travel related messages posted using the Twitter microblogging service. In this study, we propose a human coded sentiment dictionary specific to the travel context. Terms were identified from a pool of more than 1.38 million travel related tweets collected over a nine-month period. Human coders assigned sentiment scores to these terms and the travelMT 1.0 dictionary was produced to enhance the existing labMT 1.0 dictionary. The quality of the travelMT 1.0 dictionary was tested against the original labMT 1.0 dictionary and human judges. We found that, with a larger number of travel terms in a tweet, the enhanced dictionary, travelMT 1.0, produces a more accurate sentiment score than the labMT 1.0 dictionary.