A method for estimation similarity of recipe sentences employing distributed representation and Word Mover’s Distance

Abstract

This paper proposes a method for estimating similarity of recipe sentences employing distributed representation and Word Mover’s Distance (WMD). Recipe sharing service on internet is increasingly popular and have huge amount of recipe. Shared recipes have tags and other information, and we can easily categorize recipes by using these tag information. But tags attached to recipes are sometimes incorrect or not enough. So estimating distance (reverse of similarity) of recipes by using recipe sentence is useful to categorization and bring benefits to user. In this research, we investigated effectiveness of our method.