Service performance evaluation using fuzzy semantic extraction of on-line reviews

We are presenting a methodological approach to transform customer online reviews into quantitative metrics with the purpose of incorporating them into a service performance benchmarking methodology. We build on existing work in different fields, namely sentiment analysis and semantic orientation to extract the online opinions, and fuzzy logic to quantify the results into representative metrics of service quality. Fuzzy numbers were used to represent the linguistic scale behind customer online comments that reflect their evaluation of the service. We then use Data Envelopment Analysis (DEA) to quantify the performance of comparable service units. We test our methodology using published tourist destination data for the Caribbean Islands in addition to online reviews left by visitors to each island in a well-known customer review website.