Research on Improvement of Information Platform for Local Tourism by Paragraph Vector

In this paper, we discuss and propose the analysis and search method of various information on tourism in the Suwa area of Nagano Prefecture on the Internet. These pieces of information include not only long sentences such as web pages and blogs, but also a lot of content of SNS composed of short sentences of about several words. Therefore, by the conventional search method, based on the occurrence probability of words in sentences, sufficient accuracy cannot be expected for the search of SNS information composed of several words. In this research, we examined a method using Paragraph Vector for expressing relationships of words included in sentences. By doing this, we aim to acquire the same level of search performance even for SNS content composed of several words.