No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Towards a fully personalized food recommendation tool
Abstract
We present¹ a personalized ingredient-based Deep Learning recommender on the food domain that exploits ingredients and nutrition information to create recipe representations and propose to every user a more personalized and healthier meal. The recommender will be a critical component in our Meal Prediction Tool (MPT) designed with a focus on the personalization of services, increasing business efficiency and sustainability in the hospitality, restaurant and catering (HoReCa) industry.
... Os Sistemas de Recomendação visam auxiliar seus usuários no processo de tomada de decisão em um determinado contexto, tendo tais sistemas também sido extendidosàárea da recomendação de receitas culinárias [Gorbonos et al. 2018, Mokdara et al. 2018, Nezis et al. 2018, Nirmal et al. 2018]. Os Sistemas de Recomendação de Receitas (SRRs) são propostos como ferramentas para o auxílio aos usuários no encontro de dietas personalizadas e balanceadas nutricionalmente, assim como saborosas, de modo a promover melhores hábitos alimentares entre esses usuários, como uma forma de melhoria de suas saúdes, uma vez que maus hábitos alimentares estão diretamente relacionados ao desenvolvimento de várias doenças crônicas (como problemas cardíacos, diabetes e algumas formas de câncer) [Trattner and Elsweiler 2017]. ...
... A aquisição da base de dados geralmenteé realizada pela recuperação automática de documentos através da aplicação da técnica de web scraping a páginas web especializadas em culinária e gastronomia [Su et al. 2014, Ooi et al. 2015, Mokdara et al. 2018, Nezis et al. 2018, Nirmal et al. 2018]. Tambémé comum o uso de bases de dados presentes em repositórios de dados públicos [Kalajdziski et al. 2018]. ...
... Dentre os principais classificadores empregados em classificação de receitas, podemos citar o classificador Naïve Bayes [Jayaraman et al. 2017, Kalajdziski et al. 2018 [Jayaraman et al. 2017, Nirmal et al. 2018, as Redes Neurais Artificiais [Kalajdziski et al. 2018] e as Redes Neurais Profundas [Mokdara et al. 2018, Nezis et al. 2018]. ...
Neste trabalho, a classificação de receitas culinárias é abordada através da elaboração de uma ferramenta computacional própria para a análise de documentos textuais escritos em Português. A ferramenta proposta será parte fundamental no desenvolvimento de sistemas de recomendação de receitas para os brasileiros, no intuito do incentivo à prática de hábitos alimentares saudáveis por essa população. Uma base de dados nova, obtida através de páginas web brasileiras, é elaborada e testada pelo uso de algoritmos obtidos da literatura de Aprendizagem de Máquina. Experimentos foram efetuados no intuito da seleção dos melhores classificadores para a composição dos módulos de reconhecimento dos sistemas de recomendação a serem desenvolvidos.
... Recent research has also focused on personalized food recommendation systems. Nezis et al. [17] proposed a fully personalized food recommendation system that takes users dietary preferences, allergies, and cultural background into account. A study was conducted with users to assess the precision of the proposed recommendation system and it shows that it outperforms existing food recommendation systems. ...
... They also demonstrated the practical usefulness of RecipeBowl by integrating it into a mobile application for recipe and ingredient recommendation. We propose a recipe recommendation system named RecipeMate, that suggests recipes based on the user's input ingredients, cuisine, and similar to Nezis et al. [17] takes into account of user's dietary preferences such as allergens, user can put in their undesired ingredients, which is optional. The proposed system uses natural language processing techniques to preprocess the recipe dataset, including removing punctuations, converting everything to lowercase, and removing stop words. ...
... Previous works have focused on personalized recommendation of recipes using various features and employing machine learning-based approaches [17]- [20]. While Ge et al. proposed to incorporate users' tags and ratings that indicate food preferences in recommendation [17], we employed a similar approach by utilizing recipe tag information such as main dish, 5-minute-cooking. ...
Countless possibilities of recipe combinations challenge us to determine which additional ingredient goes well with others. In this work, we propose RecipeBowl which is a cooking recommendation system that takes a set of ingredients and cooking tags as input and suggests possible ingredient and recipe choices. We formulate a recipe completion task to train RecipeBowl on our constructed dataset where the model predicts a target ingredient previously eliminated from the original recipe. The RecipeBowl consists of a set encoder and a 2-way decoder for prediction. For the set encoder, we utilize the Set Transformer that builds meaningful set representations. Overall, our model builds a set representation of an leave-one-out recipe and maps it to the ingredient and recipe embedding space. Experimental results demonstrate the effectiveness of our approach. Furthermore, analysis on model predictions and interpretations show interesting insights related to cooking knowledge.
... These systems are able to perform personalized information retrieval based on many factors, like the analysis of user profile and history, sentiment and community analysis concerning the product to be recommended, etc. In literature, some cooking recipe recommendation systems have been presented recently [Mokdara et al. 2018, Nirmal et al. 2018, Nezis et al. 2018, Gorbonos et al. 2018, Nilesh et al. 2019, Rong et al. 2019, Zhang et al. 2019. Recipe recommendation systems could assist users in finding a personalized and balanced diet, promoting good eating habits as a manner to improve their health, once many chronic diseases (such as heart diseases, cancer and diabetes) are associated with bad eating habits [Trattner and Elsweiler 2017]. ...
Recipe sharing websites have become very popular in the past years, allowing individuals to use such systems in an attempt to find a desired recipe. But sometimes finding recipes which best fit the user's wishes, while still satisfying his food restrictions, may become a very time consuming and difficult task. In this work, we propose a recipe multi-label classification approach as part of a recipe recommendation system for people with food restrictions, in an attempt to automatically identify whether an input recipe or list of ingredients fits into one or more food restrictions, satisfying both user's expectations and needs. The experimental evaluation includes two approaches for feature selection, as a manner to reduce the computational costs for the proposed system.
... Os Sistemas de Recomendação [Isinkaye et al. 2015] são técnicas de filtragem automática que reduzem a quantidade de dados recuperada por uma busca realizada por um usuário, tendo tais técnicas sido adaptadas em várias aplicações no contexto de culinária [Mokdara et al. 2018, Nirmal et al. 2018, Nezis et al. 2018, Gorbonos et al. 2018]. Os Sistemas de Recuperação de Receitas (SRRs) visam auxiliar seus usuários a encontrarem receitas e dietas que sejam balanceadas nutricionalmente e que atendam seus gostos pessoais. ...
Atualmente, mesmo com o aumento no número de páginas web e sistemas de compartilhamento de receitas, usuários podem ter dificuldade na busca por pratos específicos através da enorme quantidade de dados contidas nesses repositórios. Encontrar receitas que se adequem a um conjunto de ingredintes em mãos, contemplando as vontades e restrições desses usuários, pode ser uma tarefa demorada ou mesmo impossível. Neste trabalho, um sistema de recomendação e geração de receitas é proposto, baseado na substituição de ingredientes das receitas e em uma abordagem focada nos dados, em uma tentativa de ajudar os usuários a encontrarem receitas que contemplem tanto seus desejos, quanto suas restrições alimentares, evitando desperdícios.
The enormous variations in food choices and lifestyle in today’s world have given rise to the demand of using recommender system as a suitable tool in making appropriate food choices. Need for choosing nutritious food items is becoming important in today’s modern lifestyle as people are getting indulged in eating unhealthy food and thus leading to miserable health status. Many lifestyle-related diseases, such as diabetes and obesity, can also be reversed by proper diet. Filtering techniques in recommender system use a dataset of items and user’s preferences as input to discover a list of well-judged items as suggestions. In light of the above said, this paper presents an exploratory study of the recommendation approaches and methods utilized for food and recipe recommendations in the past decade. For this purpose, several relevant papers published in the food domain from 2006 to 2023 have been extensively studied. Further, the research articles are categorized based on the filtering techniques, methods, functionalities and data sources used by the researchers. A comparative study of the recommendation approaches revealed the advantages and disadvantages of different approaches. Also, the paper emphasizes the importance of food recommendation techniques in health and nutrition management of an individual. The review highlights the need to further explore the implementation of recommender systems in the food industry. Furthermore, the findings of this survey provide researchers with insights and future directions on recommender systems in the domain.
The use of 3-Dimensional (3D) printing,
known as Digital fabrication (DF) or additive
manufacturing (AM), technology in the food sector has
countless potential to fabricate 3D constructs with
complex geometries, customization, and on-demand
production. For this reason, 3D technology is driving
major innovations in the food industry. This paper
presents the construction of a chocolate 3D printer by
applying the pressure pump technique using chocolate
as a printing material. Here the conventional 3D
printer’s design was developed as a chocolate 3D
printer. As an improvement, a new extruder
mechanism was introduced. The extruder was
developed to print the chocolate materials. In the
working mechanism, the 3D printer reads the design
instruction and chocolate material is extruding
accordingly, through the nozzle of the pump to the bed
of the 3D printer followed by the design (layer by layer).
The special part of this chocolate 3D printer is the
pressure pump in the extruder part. That pressure
pump provides pressure on melted chocolate from the
chocolate container to the nozzle point. The usability
and efficiency of the 3D printer were tested with sample
designs. The obtained results were presented and
discussed. Together with these advances this 3D printer
can be used to produce complex food models and design
unique patterns in chocolate-based sweets by satisfying
customers.
Poor nutrition is one of the major causes of ill-health and death in the western world and is caused by a variety of factors including lack of nutritional understanding and preponderance towards eating convenience foods. We wish to build systems which can recommend nutritious meal plans to users, however a crucial pre-requisite is to be able to recommend recipes that people will like. In this work we investigate key factors contributing to how recipes are rated by analysing the results of a longitudinal study (n=124) in order to understand how best to approach the recommendation problem. We identify a number of important contextual factors which can influence the choice of rating. Based on this analysis, we construct several recipe recommendation models that are able to leverage understanding of user's likes and dislikes in terms of ingredients and combinations of ingredients and in terms of nutritional content. Via experiment over our dataset we are able to show that these models can significantly outperform a number of competitive baselines.
We present a prototype of a personalized recipe advice system, which facilitates its users to make health-aware meal choices based on past selections. To stimulate the adoption of a healthier lifestyle, a goal setting mechanism is applied in combination with personalized recipe suggestions.
Front-of-pack 'traffic-light' nutrition labelling has been widely proposed as a tool to improve public health nutrition. This study examined changes to consumer food purchases after the introduction of traffic-light labels with the aim of assessing the impact of the labels on the 'healthiness' of foods purchased. The study examined sales data from a major UK retailer in 2007. We analysed products in two categories ('ready meals' and sandwiches), investigating the percentage change in sales 4 weeks before and after traffic-light labels were introduced, and taking into account seasonality, product promotions and product life-cycle. We investigated whether changes in sales were related to the healthiness of products. All products that were not new and not on promotion immediately before or after the introduction of traffic-light labels were selected for the analysis (n = 6 for ready meals and n = 12 for sandwiches). For the selected ready-meals, sales increased (by 2.4% of category sales) in the 4 weeks after the introduction of traffic-light labels, whereas sales of the selected sandwiches did not change significantly. Critically, there was no association between changes in product sales and the healthiness of the products. This short-term study based on a small number of ready meals and sandwiches found that the introduction of a system of four traffic-light labels had no discernable effect on the relative healthiness of consumer purchases. Further research on the influence of nutrition signposting will be needed before this labelling format can be considered a promising public health intervention.
Food recommenders have the potential to positively influence the eating habits of users. To achieve this, however, we need to understand how healthy recommendations are and the factors which influence this. Focusing on two approaches from the literature (single item and daily meal plan recommendation) and utilizing a large Internet sourced dataset from Allrecipes.com, we show how algorithmic solutions relate to the healthiness of the underlying recipe collection. First, we analyze the healthiness of Allrecipes.com recipes using nutritional standards from the World Health Organisation and the United Kingdom Food Standards Agency. Second, we investigate user interaction patterns and how these relate to the healthiness of recipes. Third, we experiment with both recommendation approaches. Our results indicate that overall the recipes in the collection are quite unhealthy, but this varies across categories on the website. Users in general tend to interact most often with the least healthy recipes. Recommender algorithms tend to score popular items highly and thus on average promote unhealthy items. This can be tempered, however, with simple post-filtering approaches, which we show by experiment are better suited to some algorithms than others. Similarly, we show that the generation of meal plans can dramatically increase the number of healthy options open to users. One of the main findings is, nevertheless, that the utility of both approaches is strongly restricted by the recipe collection. Based on our findings we draw conclusions how researchers should attempt to make food recommendation systems promote healthy nutrition.
Many machine learning algorithms require the input to be represented as a
fixed-length feature vector. When it comes to texts, one of the most common
fixed-length features is bag-of-words. Despite their popularity, bag-of-words
features have two major weaknesses: they lose the ordering of the words and
they also ignore semantics of the words. For example, "powerful," "strong" and
"Paris" are equally distant. In this paper, we propose Paragraph Vector, an
unsupervised algorithm that learns fixed-length feature representations from
variable-length pieces of texts, such as sentences, paragraphs, and documents.
Our algorithm represents each document by a dense vector which is trained to
predict words in the document. Its construction gives our algorithm the
potential to overcome the weaknesses of bag-of-words models. Empirical results
show that Paragraph Vectors outperforms bag-of-words models as well as other
techniques for text representations. Finally, we achieve new state-of-the-art
results on several text classification and sentiment analysis tasks.
As the obesity epidemic takes hold across the world many medical professionals are referring users to online systems aimed at educating and persuading users to alter their lifestyle. The challenge for many of these systems is to increase initial adoption and sustain participation for sufficient time to have real impact on the life of its users. In this work we present some preliminary investigation into the design of a recipe recommender, aimed at educating and sustaining user participation, which makes tailored recommendations of healthy recipes. We concentrate on the two initial dimensions of food recommendations: data capture and food-recipe relationships and present a study into the suitability of varying recommender algorithms for the recommendation of recipes.
The recording and sharing of cooking recipes, a human activity dating back
thousands of years, naturally became an early and prominent social use of the
web. The resulting online recipe collections are repositories of ingredient
combinations and cooking methods whose large-scale and variety yield
interesting insights about both the fundamentals of cooking and user
preferences. At the level of an individual ingredient we measure whether it
tends to be essential or can be dropped or added, and whether its quantity can
be modified. We also construct two types of networks to capture the
relationships between ingredients. The complement network captures which
ingredients tend to co-occur frequently, and is composed of two large
communities: one savory, the other sweet. The substitute network, derived from
user-generated suggestions for modifications, can be decomposed into many
communities of functionally equivalent ingredients, and captures users'
preference for healthier variants of a recipe. Our experiments reveal that
recipe ratings can be well predicted with features derived from combinations of
ingredient networks and nutrition information.
Gijs Geleijnse Peggy Nachtigall Pim van Kaam and Luciënne Wijgergangs
- Peggy Gijs Geleijnse
- Luciënne Nachtigall Pim Van Kaam
- Wijgergangs