Generating Personalized Recipes from Historical User Preferences

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 5976–5982, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 5976 Generating Personalized Recipes from Historical User Preferences Bodhisattwa Prasad Majumder∗, Shuyang Li∗, Jianmo Ni, Julian McAuley Computer Science and Engineering University of California, San Diego {bmajumde, shl008, jin018, jmcauley}@ucsd.edu Abstract Existing approaches to recipe generation are unable to create recipes for users with culinary preferences but incomplete knowledge of in- gredients in specific dishes. We propose a new task of personalized recipe generation to help these users: expanding a name and incom- plete ingredient details into complete natural- text instructions aligned with the user’s histor- ical preferences. We attend on technique- and recipe-level representations of a user’s previ- ously consumed recipes, fusing these ‘user- aware’ representations in an attention fusion layer to control recipe text generation. Exper- iments on a new dataset of 180K recipes and 700K interactions show our model’s ability to generate plausible and personalized recipes compared to non-personalized baselines. 1 Introduction In the kitchen, we increasingly rely on instructions from cooking websites: recipes. A cook with a predilection for Asian cuisine may wish to pre- pare chicken curry, but may not know all neces- sary ingredients apart from a few basics. These users with limited knowledge cannot rely on ex- isting recipe generation approaches that focus on creating coherent recipes given all ingredients and a recipe name (Kiddon et al., 2016). Such mod- els do not address issues of personal preference (e.g. culinary tastes, garnish choices) and incom- plete recipe details. We propose to approach both problems via personalized generation of plausi- ble, user-specific recipes using user preferences extracted from previously consumed recipes. Our work combines two important tasks from natural language processing and recommender systems: data-to-text generation (Gatt and Krah- mer, 2018) and personalized recommendation ∗ denotes equal contribution (Rashid et al., 2002). Our model takes as user in- put the name of a specific dish, a few key ingre- dients, and a calorie level. We pass these loose input specifications to an encoder-decoder frame- work and attend on user profiles—learned latent representations of recipes previously consumed by a user—to generate a recipe personalized to the user’s tastes. We fuse these ‘user-aware’ represen- tations with decoder output in an attention fusion layer to jointly determine text generation. Quan- titative (perplexity, user-ranking) and qualitative analysis on user-aware model outputs confirm that personalization indeed assists in generating plau- sible recipes from incomplete ingredients. While personalized text generation has seen success in conveying user writing styles in the product review (Ni et al., 2017; Ni and McAuley, 2018) and dialogue (Zhang et al., 2018) spaces, we are the first to consider it for the problem of recipe generation, where output quality is heavily depen- dent on the content of the instructions—such as ingredients and cooking techniques. To summarize, our main contributions are as follows: 1. We explore a new task of generating plausi- …

Content truncated to 500 words