Many people reserve and purchase travel products via internet. Travelers refer to them to gain more accurate information about specific places, the amount of time they will spend there, and so on. This paper focuses on such worldwide travel site, “A,” and selects 16 hotels, which span two approaches: one from luxurious to modest, and the other, on the basis of geography, from Central Tokyo to local areas. These hotels are classified into four quadrants, and data from site A is collected on the basis of 19,224 reviews. Each data point is composed of each contributor’s review text, assessment scores, and the date on which the review was published. Data points are used to analyze three things: (1) Words extracted from reviews, which are grouped according to a topic model to extract characteristics about a hotel; (2) Hotel rankings, topics, and appearance words per region; and (3) the relationships between topics and evaluations, which are quantified to consider future guidelines and services. Through these analyses, we discuss the relationship between the evaluation of a hotel and its reviews on the basis of regions and hotel rankings.
抄録全体を表示