Routing Enforced Generative Model for Recipe Generation

Abstract

One of the most challenging part of recipe generation is to deal with the complex restrictions among the input ingredients. Previous researches simplify the problem by treating the inputs independently and generating recipes containing as much information as possible.In this work, we propose a routing method to dive into the content selection under the internal restrictions. The routing enforced generative model (RGM) can generate appropriate recipes according to the given ingredients and user preferences. Our model yields new stateof-the-art results on the recipe generation task with significant improvements on BLEU, F1 and human evaluation.

Ratatouille: A tool for Novel Recipe Generation

“…Several advances in the field of computation and dataset availability allows to generate recipes from food images [11]- [13] and generate personalized recipes based on user demands [2], [8]. A rule-based generation algorithm has been implemented to develop a recipe generation system, EPICURE [14].…”

“…Authors used bidirectional encoder-decoder to generate recipes and evaluate the generated recipes. In a similar spirit, a routing enforced generative model [2] was proposed to generate recipes considering the ingredients and user preference. Transformer-based models such as GPT [18], BERT [19], Roberta [20] outperform the LSTM [21] and RNN-based models to generate the novel recipes.…”

“…Deep learning models are trained with the anticipation of novel recipe generation through the list of ingredients. Several models were propounded Transformer-based model [1], routing enforced generation model [2], language model [3]- [5], knowledge-based models [6], [7] and GRU (Gated Recurrent Unit) based encoder-decoder model [8], [9]. These models were able to generate novel recipes along with recipe instructions and ingredients list but most of them could not generate quantity and units of ingredient.…”