[1I2-J-5] Web Intelligence

[1I2-J-5-03] Proposal of Similar Area Discovery Based on Impression Extracted from Social Media

〇Yasufumi Takama¹, Yosuke Sakamoto¹, Kenichiro Kobayashi², Takehiko Hashimoto² (1. Tokyo Metropolitan University, 2. GA technologies)

Keywords:impression extraction, information recommendation, ReTech

This paper proposes a method for finding similar areas in terms of impression. Nowadays, social media is one of our important resources, from which we can obtain subjective information, such as the impression of some products and places. Such information is expected to be useful when looking for the places to live. The proposed method constructs an emotional word dictionary and uses it to extract the impression of the area around a station from reviews. The similarity between stations is calculated based on the extracted impression. This paper compares the quality of dictionaries constructed with different methods and shows the accuracy of the similarity judgment.