Analysis of Sentiment Index for Water Quality Satisfaction using Web Crawling

Abstract

Owing to recent water accidents, consumers' distrust of tap water and their reluctance to use it have continued. Accordingly, in this study, we analyzed tap water-related data mentioned on social media before and after water quality accidents (red-colored water and larvae issues) to identify consumers' perceptions. To achieve this, a database was built for the data collected through web crawling, and a three-stage customized sentiment lexicon was created: positive, negative, and neutral. Sentiment analysis was then performed using the manufactured sentiment lexicon, and the emotional index was evaluated to predict satisfaction with tap water by calculating the ratio of positive, negative, and neutral words from the analysis results. In the case of red water, the average values of the emotional index before and after the occurrence were calculated as 77.6 and 73.5, respectively, and the lowest value at the time of the event of red water was 59.76, which was lower than the average. For the case of larvae, it was confirmed that the average value before and after the accident was 82.2, and the lowest value during the occurrence period was 58.74. According to these results, it would be possible to monitor any changes in the emotional index value and identify events early in the case of an outlier. In this respect, a method of using the analyzed emotional index as an evaluation index for satisfaction was proposed. If this evaluation technology is reflected in the survey methodology, it is expected that it could be used as an effective countermeasure to respond quickly in an accident and improve citizens' satisfaction with tap water.