Detecting Sensory Textures with Rheological Characteristics from Recipe Sharing Sites

This study tries to estimate texture of gel related dishes based on both the data from recipe sharing sites, and research results of food science. Since most of recipes are not accompanied by sufficient information about what kind of textures they realize, we propose a method to estimate characteristics of textures for each recipe by applying a joint topic model to bridge sensory texture terms in a recipe sharing site with corresponding quantitative textures resulting from food science research. The result shows that the estimated texture terms for dishes are consistent with rheology, the quantitative textures provided by related food science research.