Maki Sakamoto Laboratory of the University of Electro-Communications and Taski Co., Ltd., which handles the real estate tech business, jointly researched "scoring the atmosphere of the city using onomatope and predicting regression from statistical information", and the 2022 Japanese Society for Artificial Intelligence National Convention Announced at.
In marketing reports created by real estate developers when planning buildings in the city, qualitative information about the characteristics and atmosphere of the city tends to be biased toward the experience and subjectivity of the person in charge.Therefore, the research team conducted a joint research on "scoring the atmosphere of the city using onomatope", which enables quantification of qualitative information on real estate information, in order to create marketing reports based on facts.
Onomatopoeia is a word that expresses various states and movements with sound.In the research, a questionnaire survey was conducted on 20 students of the University of Electro-Communications in order to quantify the atmosphere of the city.As a result, 76 onomatopoeia were extracted as parameters used to measure the atmosphere of the city. 76 onomatopoeia were quantified by 43 kinds of adjective pair scales by the patented technology possessed by Sakamoto laboratory, and 10 onomatopoeia were extracted from each cluster by cluster analysis by Ward's method.
In addition, 40 stations in Tokyo were randomly selected and scored by subjective evaluation using the SD method (Semantic Differential Method) in 5 stages for 400 CrowdWorks users.As explanatory variables for predicting onomatopoeia, we used government statistics and 162 types of city statistics on population and households published by the Tokyo Metropolitan Government Statistics Department.As a result of constructing and learning a regression model that predicts the atmosphere of the city by inputting statistics using vector regression, it became clear that there are onomatope with high influence and onomatope with low influence in determining the atmosphere of the city. ..
In the future, we will work on the addition of learning data, the analysis of features that are effective for prediction, and further analysis and improvement of Onomatope, which had low prediction accuracy, and will work on the development of a system that scores the atmosphere of the city.