[4Yin2] Interactive session 2

[4Yin2-32] An integrated approach for dialogue workshop and graph-based natural language processing

Utilizing Aska that a framework collects opinions beyond linguistic meaning

〇Yuki Ashino¹, Tatsuro Kawamoto², Ken Yasumatsu^3,4 (1.Osaka Kyoiku University Graduate School of Education, 2.Artificial Intelligence Research Center, AIST, 3.Osaka Kyoiku University, 4.Department of Behavior Observation, OGIS RI Co.,Ltd.)

Keywords:matrix reordering, natural language data, graph clustering, free-format texts, synthesis

In this research, we generated natural language data by qualitative approach, analyzed the data by Aska that utilizes machine-learning methods, and evaluated its usefulness by comparing the result of Aska with that of human assessments, all with the aim of proposing an analysis method for natural language data. Concretely, we took the following procedure: firstly collecting opinion data (text data) from each individual through a creative workshop method, secondly letting individuals post and mutually assess the opinions in Aska, and thirdly classifying the data based on a matrix plot obtained through Aska. As a result of comparing the classified data using Aska with that by workshop, it was confirmed that the classification of the opinion data was not dominated by particular words, but by overall meaning of sentence holistically. This study implies that the integrated approach of the creative workshop and Aska enables us to aggregate and interpret opinion data more rationally.