Investigation of Preprocessing of Multilingual Online Reviews for Automatic Classification
M. Okada, Kenji Hashimoto · 2012
Online reviews of commercial sites are important sources for customers to obtain information and opinions by other customers. However, generally, the reviews contain several mixed information such as purpose and sentiments of reviewers. Therefore, when the reviews are classified automatically using some machine learning methods, it is necessary to construct adequate data preliminary for machine learning. In this paper, we obtained multilingual reviews from a travel information portal site "TripAdvisor", and investigated whether preprocessing is effective or not in order to classify reviews appropriately using a machine learning method "Support Vector Machine (SVM)." We also investigated difference of effectiveness when the reviews are written by different languages.