[extensive reading of papers] a delicious recipe analysis framework for exploring multi modal recipes with variable attributes

Thesis title ： A delicious recipe analysis framework , Explore multimodal recipes with various properties
Time ：2017
source ：ACM
Address of thesis ：https://doi.org/10.1145/3123266.3123272

Abstract
   Multidimensional food analysis based on factors affecting food . Combine various types of continuous and discrete attribute features and multimodal information in the recipe .
   First , We have developed a multi-attribute topic modeling method , This method can model any type of attribute features and text content . Then the multimodal embedding method is used , Build from MATM What I learned Text theme features And the deep learning network Image features The correlation between .
   By learning attributes - Topic relationships and multimodal correlations , We can complete different applications , Include :
(1) Preference analysis and comparison , Better understand the different dimensions of preference patterns , Such as areas and courses ;
(2) Region oriented multimodality 、 Multi attribute information multi-dimensional food summary ;
(3) Multi attribute oriented recipe recommendation .
Besides , The framework is flexible , You can easily combine attributes and modes of any type . In the collection of Yummly On dataset , Qualitative and quantitative evaluation results verify the effectiveness of the method and framework .