Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 01, 2022 - June 04, 2022
Annotation work is necessary for natural language processing based on machine learning for robots. For the purpose to realize the automation of this task, we studied some automatic labeling methods for product review evaluation, and proposed two methods using the analysis of adjectives and learning of human ratings. The proposed two methods were evaluated by experiments and their results were compared. Based on these results, we also discussed the usefulness of the proposed methods when applying to other datasets.
