Nowadays, online reviews (User Generated Content – UGC) written by people on review web platforms, e-commerce website etc. offer the opportunity for business to deeply analyse Customer Satisfaction (CS) with products or services. The availability of such free textual data allows practitioners to study customer behaviour exploiting huge amounts of data. To help processing and analysing such data, sentiment analysis and emotion analysis have been proposed and extensively adopted in literature. Sentiment analysis represents the process of automatic identification and categorization of opinions expressed in a piece of text, especially focusing in determining whether user’s attitude towards a particular item (i.e. topic, product or service) is positive, neutral or negative. On the other hand, emotional analysis aims to find the hidden emotions behind texts.

With the aim to provide businesses with insights into trends concerning their product or services, advanced methods, including text mining and sentiment analysis, have been used to transform the unstructured social media data into structured data series. In the recent literature, researchers have devoted efforts to obtain structured time series from texts and images.

After a comprehensive literature review, an application of sentiment analysis for time series data is presented. Through webscraping we extract reviews on Tripadvisor activities concerning movie-set tourism. Being some scenes of “The Lord of the Rings” (LOTR) and “the Hobbit” movies shot in Hinuera, New Zealand, we scraped all comments reviews of LOTR-related activities available there. Linking the sentiment extracted from the reviews to the date in which the reviews have been written allows to obtain customer satisfaction time-series that reflect the trend in the customers’ opinion concerning the product/service. Further insights concerning the level of appreciation toward different aspects can be obtained by relating the reviews to the topics they deal with, using LDA topic modelling to extract such topic information for each review. Concluding, the main issues related to sentiment analysis for time series are highlighted offering some suggestions and recommendations for future analysis.