=Paper=
{{Paper
|id=Vol-3335/MSW_paper3
|storemode=property
|title=A Review on Multilingual Food Recommendation Systems for Critical Medical Conditions In Pregnancy Care
|pdfUrl=https://ceur-ws.org/Vol-3335/MSW_Paper3.pdf
|volume=Vol-3335
|authors=Patience U. Usip,Agnes Udo,Funebi F. Ijebu,Sanju Tiwari
}}
==A Review on Multilingual Food Recommendation Systems for Critical Medical Conditions In Pregnancy Care==
https://ceur-ws.org/Vol-3335/MSW_Paper3.pdf
A Review on Multilingual Food Recommendation
Systems for Critical Medical Conditions in Pregnancy
Care⋆
Patience U. Usip1,∗,† , Agnes Udo1,† , Funebi F. Ijebu1,† and Sanju Tiwari2,†
1
1Computer Science Department, University of Uyo, Uyo, Nigeria
2
Universidad Autonoma de Tamaulipas, Mexico
Abstract
AMedical conditions require proper attention as some of them especially those occurring in women
that are expecting pregnancy require proper care. Although pregnancy exposes them to several critical
health conditions, most of these conditions were actually in existence before pregnancy with their causes
traced back to poor nutrition resulting from food intake. However, there have been a good number of
food recommendation systems in existence with few of them being context-aware. This work is aimed
at performing a critical review of the food recommender systems that considers medical conditions and
design the food ontology for critical medical conditions that adopts the pathology test results of patients
in caring for the existing medical conditions such as pregnancy. Focus of the framework is on deploying
a multilingual ontology-based approach in addition to other required approaches for recommendations.
This work also introduces the addition of uncooked food in the food ontology for recommendations.
The proposed food ontology for critical medical conditions considers the required nutrients based on the
combination of several factors including the pathology test results, location/language, medical conditions
and user preferences. The use of this food ontology framework for food recommender is reliable due to
the additional factors.
Keywords
Critical Health conditions, Ontology-based Representations, Nutrition, Pathological test
1. Introduction
Recommender systems work behind the scenes on many of the world’s most popular websites.
The preceding few decades have shown a tremendous rise in web services like AMAZON, FACE-
BOOK & INSTAGRAM, YOUTUBE, GOOGLE, ETC. Other examples of recommender systems
at work include movies on Netflix, songs on Spotify and profiles on Tinder. And with the rise
of such sites, recommender systems are getting much more important than before. Several
IWMSW-2022: International Workshop on Multilingual Semantic Web, Co-located with the KGSWC-2022, November
21–23, 2022, Madrid, Spain
⋆
You can use this document as the template for preparing your publication. We recommend using the latest version
of the ceurart style.
∗
Corresponding author.
†
These authors contributed equally.
Envelope-Open patienceusip@uniiuyo.edu.ng (P. U. Usip); agnesgudo247@gmail.com (A. Udo); ijebufrancis@uniuyo.edu.ng
(F. F. Ijebu); sanju.tiwari.2007@gmail.com (S. Tiwari)
Orcid 0000-0002-6516-5194 (P. U. Usip)
© 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
recommendation systems use algorithms including the collaborative filtering, content-based
and hybrid approaches. Context-Aware Recommendation (CAR) is a recent approach said to
generate better recommendations based on the specific contextual situation of the user. CARs
use several approaches to incorporate contextual information for the users in the recommen-
dation process and then use such approaches for different applications. The representation
of these knowledge for context-aware recommendations is very necessary. Hence, the choice
of ontologies as representation tools for semantic web reasoning of the proposed food recom-
mender system.
Ontology is best known for defining the concepts and relationships used to describe and repre-
sent an area of concern with the specific role of facilitating data organization and integration in
the semantic web [1]. Ontologies are suitably considered here due to its crucial role in enabling
automatic knowledge processing, sharing, and reuse among various applications. The use of
ontologies and other AI techniques are reported by many potential research studies [2, 3, 4].
Another application domain with high demand for context-based recommendation is the food
and health domain with a particular task of providing care for expectant mother and the unborn
baby (pregnancy) regarding what should be eaten. Making decisions about what to eat, when
to eat and even the proportion of food intake has become major problem in our everyday lives,
particularly the expectant mothers, due to a wide variety of ingredients, cultures, personal
tastes and individual medical conditions has made choosing the right food at the right time to
be as a very difficult task. Today, many diseases that were previously thought as hereditary are
now seen to be connected to biological dys-function related to nutrition.
Already existing food recommender systems mostly focused on meals (cooked food) and users’
preferences without considering uncooked food and pathology test of users for accurate break-
down of the nutritional state of the user. Among many factors for healthy living, right food
consumption is one of the major factors, specifically, for people suffering from critical health
ailments such as over-weight, under-weight, heart disease, cancer, diabetes, allergies, obesity,
which may significantly lead to many other complications in pregnancy such as birth defects,
premature birth, miscarriage, stillbirth and infertility. Complications in pregnancy can be life
threatening (critical), most common, minor, rare, early and chronic. Although, certain medical
conditions may complicate a pregnancy, with proper care, most women can enjoy a healthy
pregnancy even with her health challenges. Various studies portray that inappropriate and
insufficient intake of food are the major reasons of various health issues and diseases (WHO/U-
NISEF, 2001).
Most of the critical health conditions occur due to the excess and deficiency of food element.
Some of the key elements of food include carbohydrate, calcium, iron, protein and vitamins and
the consumption of balanced diet can normalize some of the medical conditions. Addressing
the problem of selecting what and how to eat can promote healthy living and decrease usage of
unhealthy food ingredients or compounds such as oxidants. Healthy eating is possible only
when you ensure that your body is getting the right food elements in the right quantity that
it requires. Therefore, it is essential to know about the food nutrition to be sure of what is
consumed into the body. These food elements categorically include carbonaceous (starch, sugar,
fats), nitrogenous(albumen) and inorganic substances (minerals).
The requirement of food changes over time, with respect to the stage of life that we are in, for
example while a child requires food to grow and develop, adults need it more for the energy
required to function daily. In addition to that, food is also required so as to repair the various
parts of the body that are worn away, or those that have to be discarded due to wear and tear
caused by critical health conditions. Therefore, we need to continuously replenish our body
with a variety of elements present in food and drink. Pathology test is the major mode of
examination that offers the status of these key elements.
To get appropriate food intake to improve overall well-being, smart food recommender systems
will offer prevalent recommendation for healthier food choices [5, 6]. This work is aimed at
carrying out a critical review of food recommendation systems that consider medical conditions
and giving the design of a context aware food recommender framework for medical conditions
and pregnancy care. Its focus is on providing an ontology-based model for classification and
recommendation of food (cooked and uncooked) for users with critical health conditions in the
right proportion for consumption. This work will bring an improvement to the challenge of
an existing system by directly recommending right food (cooked and uncooked), nutritional
elements and the proportion of food consumption especially for critical health conditions based
on user’s pathological reports
The rest of the paper is organized as follows. Section 2 presents the critical review on related
works while the methodologies and the data collection adopted for the design of the system
are described in section 3. Section 4 presents a detailed food ontology to include the uncooked
brand of food with the required proportion of the nutritional elements for pregnant women
with medical conditions.
2. Related Literature about Recommendation Algorithms
2.1. Recommendation Algorithms
The major and most common algorithms used in recommendation systems are discussed as
follows. Collaborative filtering is basically an algorithm used in the recommendation system
that basically makes the use of similarities between the items and users in order to provide the
right recommendations. This means this type of algorithm can provide a recommendation to
user A depending on the interest of a similar user B. What makes collaborative filtering different
is the fact the past user-item interaction is enough for the predictions to be made for similar
users.
Content-based recommendation, on the other hand, solely depends on the choice of the users
themselves, and the recommendations are made based on the items or products the users like
based on their previous feedback or actions. The content-based recommendations shown in
Figure 3 are less problematic for the new users as the items can be described by their character-
istics or content, and thus relevant recommendations can be made for the new entities.
The hybrid approach of recommendations combines collaborative and content-based recom-
mendations. In a system, first the content recommender takes place as no user data is present,
then after using the system the user preferences with similar users are established. These kinds
of approaches are said to have provided better and accurate recommendations. Netflix uses a
hybrid approach of recommendations because it compares the watching habits of similar users
and also offer movies that share the characteristics with all those films which the user has rated
high.
The earliest recommendation system, called Tapestry filters, use information by collaborative
filtering system. Collaborative recommendation system is the most famous and commonly
used one. The system analyses the preferences from the set of users within the system. It finds
out the set of users with similar characteristics and takes this relevance as evidence to induct
the potential preferences of the users. Besides recommending the interested information to
the users, this research is expected to recommend the information that may arouse the users’
potential demands based on users’ culture and pathology test.
Various studies portray that inappropriate and insufficient intake of food are the major reasons
of various health issues and diseases. A study conducted by World Health Organization (WHO)
estimate that about 30% of the total population of the world is suffering from various diseases,
60% deaths each year in children are related to malnutrition and about 9% of heart attack
deaths. Moreover, many children are suffering from Vitamin-A deficiency, 200 billion people are
suffering from iron deficiency (anemia) and a lot of people are suffering from iodine deficiency
(WHO/UNISEF, 2001).
Several works proposed different recommendation systems related to food. These systems can
be categorized as:
(a) food recommendation systems [7, 8]
(b) menu recommendations [9]
(c) diet plan recommendations[10]
(d) health recommendations for different diseases like diabetes and cardiovascular [11, 12]
(e) recipe recommendations [13, 14]
The above mentioned systems provide recommendations to either some specific disease or
balance the diet without considering information about any disease or nutrition deficiency in the
body. For instance, in[7] a food recommendation system is proposed for diabetic patients that
used K-mean clustering and Self-Organizing Map for clustering analysis of food. The system
recommends various foods for diabetic patients without considering the disease level that may
fluctuate frequently in different situations of the patient and the food recommendations may
also vary accordingly.
Similarly, the authors in [8], do not consider the nutrition factors that have significant impor-
tance for a balanced diet recommendation. Tags and latent factor are used for android-based
food recommender system [8]. The system recommends personalized recipe to the user based
on tags and ratings provided in user preferences. The proposed system used latent feature
vectors and matrix factorization in their algorithm. Prediction accuracy is achieved by use of
tags which closely match the recommendations with users’ preferences. However, the authors
do not consider the nutrition factors in order to balance the diet of the user according to his
needs.
Content based food recommender system [13] is proposed which recommend food recipes
according to the preferences already given by the user. The preferred recipes of the user are
fragmented into ingredients which are assigned ratings according to the stored users’ prefer-
ences. The recipes with the matching ingredient are recommended. The authors do not consider
the nutrition factors and the balance in the diet. Moreover, chances of identical recommendation
are also present because the preference of the user may not change on daily basis.
In [15], knowledge based dietary nutrition recommendation system is proposed for obesity. The
recommendations include dietary nutrition and diet menus for individuals using collaborative
filtering technique. An application for mobile users is also developed in order to recommend
the dietary and menus to the users. Similarly, a food recommender system is proposed in [16]
for patients in care facilities. The application is designed for caregivers in the care facilities in
order to offer the food according to the patient preferences.
Majority of these recommendation systems extract users’ preferences from different sources like
users’ ratings [17, 5], recipe choices [18, 19] and browsing history1 [20] ,[21, 22] . For instance,
in [19] a recipe recommendation system is proposed using social navigation system. The social
navigation system extracts users’ choices of recipes and in return recommends the recipes. Sim-
ilarly, in[22], a recipe recommendation system is proposed that is capable of learning similarity
measure of recipes using crowd card-sorting. The above-mentioned recommendation systems
lack in solving a common problem known as cold start problem. All these systems must wait
for the users to enter enough data for the effective recommendations[23].
Some of the commercial applications offer users for a quick survey in order to get users prefer-
ences in a short time. For instance, the survey used by [24] is specifically designed to match the
lifestyle of the user i.e., healthy, sportsman, pregnant, etc. The survey also attempts to avoid
various foods which do not match the user’s lifestyle. Similarly, a questionnaire is used by
[25] through which a user answers different questions about his/her lifestyle, food preferences,
nutrient intake, and habits. The system once extracts all the basic information is then able to
recommend different meals for daily and weekly basis.
This work presents an ontology-based multi-lingual food recommendation system that will
specifically deal with the pathological tests results. Our system considers diseases related to
pathological reports, most common nutrition factors in recommending the food items to the
users and the proportion of the food intake. For this purpose, we will use a database of 400
pathological test reports to categorize various diseases that occur due to the changes from the
normal ranges of compounds/parameters.
1
urlhttps://www.nutrinohealth.com/
Table 1
Summary of research contributions related to food recommendation system
Ref Aims Data Method Key Findings Critique
Food recommendation Manual rating of food Health and standard TF-IDF (Term Manual rating is
[26] system to users with a items food database Frequency-Inverse prone to error and
simple scenario Document Fre- reliable information
quency) extraction could be ignored
and Similarity be-
tween the food items
and user profile.
[5] Corresponding ingredi- Content-Based (CB) al- Medical condi-
ents in recipe gorithm and Recipes pre- tions or nutrition
diction and recommen- deficiency not
dation considered
[27] summary and highlight Recommendation of Simulation using his- Performance algo- Mimic users’ profile
of approaches in recent recipes, meal plans, torical rithms measurement and feedback for rec-
state-of -art system groceries and menu in ommendation
25 recent papers
[28] Recommend dishes Food ontology construc- K-Means clustering Generates a dietetic Limited to only one
(breakfast, lunch and tion from nutrient food algorithms and Self plan based on acces- medical conditions
dinner) for diabetes dataset Organizing Map sibility and filtering
patients (SOM) unsuitable food.
[29] Determine and identify Queuing database from Semantic matching makes use of seman- System does not
side effects of nutrients Republic of Turkey Min- (concept matching tic rules, consider recipes for
in packaged food prod- istry of Food Agriculture engine (CME)) Infer- healthy eating habit
ucts on market and rec- and Livestock ence mechanism
ommend based on medi-
cal condition
[20] Provide dietary assis- Composition of foods in- A cloud-based so- Recommends various No food recommen-
tance to people with tegrated dataset (CoFID) lution using Ant foods and nutrition dations breakdown
medical conditions - 3,400 food items with 26 Colony Optimization to the people based for different timings
entries and 345 patholog- (ACO) algorithm on their pathological of the day, such as
ical test reports test reports as well breakfast, lunch, and
as manages and up- dinner and the pro-
dates the heuristic in- portion of food items
formation was not considered
3. Domain Data Collection towards a Context-Aware Food
Recommender Framework
Addressing the problem of selecting what and how to eat can promote healthy living and de-
crease usage of unhealthy food ingredients or compounds such as oxidants. To get appropriate
food intake to improve overall well-being, smart food recommender systems will offer prevalent
recommendation for healthier food choices [5, 6]. The mode of examination to offer the key
element is pathology test.
The following categories of knowledge (concepts and relations) were obtained from literatures
and adopted in this work. 1. Nutritional Elements: The key nutritional elements of food such
as carbohydrate, calcium, iron, protein and vitamins. 2. Categories of Food Elements: They
are classified as carbonaceous (starch, sugar, fats), nitrogenous(albumen) and inorganic sub-
stances (minerals). 3. Complications in Pregnancy: Birth defects, Premature death, miscarriage,
stillbirth, infertility, Preeclampsia, Preterm Labor, Gestational Diabetes, 4. Classes of Compli-
cations in Pregnancy: life threatening (critical), most common, minor, rare, early, chronic 5.
Critical medical conditions: over-weight, under-weight, heart disease, cancer, diabetes, allergies,
obesity, pregnancy 6. Food: food can be cooked(meal) and uncooked 7. Pathology test: Can
be blood tests, urine tests, stools (faeces) and bodily tissues 8. Specialisations in pathology:
chemical, haematology, anatomical, Cytopathology, medical microbiology (bacteria, viruses,
fungi, parasites), immunopathology, genetic, forensic pathology, general, clinical In addition
to that, food is also required so as to repair the various parts of the body that are worn away,
or those that have to be discarded due to wear and tear caused by critical health conditions.
Therefore, we need to continuously replenish our body with a variety of elements present in
food and drink. Healthy eating is possible only when you ensure that your body is getting the
right food elements in the right quantity that it requires. Therefore, it is essential to know about
the food nutrition to be sure of what you consumed into the body.
4. Food Ontology for Critical Medical Conditions
The Meal Recommender System will recommend list of desired meals based on the health
condition of the User. The Ontology classifies the meal with concentration on those suitable for
patients with critical health conditions in both Veg and Non-Veg classes and at specific times
for the meal (Breakfast, Lunch and Dinner). The ontology design is as described in Figure 1.
The presence of the location or language which every patient has and every menu for all
meals are based on the location/language introduce the concept of multilinguality in the food
recommendation process. The use of the pathology test results for each patients unfolds the
nutrition deficiency of these patients, their susceptibility to some critical medical conditions
and the need to recommend food based on their nutritional values.
The concepts and relationships used to describe and represent the domain of discourse are shown
in the ontology diagram with all the linkages to all necessary domain data/information. The role
of ontologies in Semantic Web is to facilitate data organization and integration. The concepts
include: patients, pathology test results, critical medical conditions, balanced diet informed by
nutrition contained in food. The recommendations will be based on the language/location of
Figure 1: Food Ontology for Critical Medical Conditions.
the patient. Other sub-concepts are as contained in the food ontology without the linked data.
5. Conclusion
This integrated data (known as Linked Data) can be used for reasoning or simply querying is
the main strength of the Semantic Web as it introduces various levels of complexities to the
food ontology. Ontologies can play a crucial role in enabling automatic knowledge processing,
sharing, and reuse among applications. Reasoning with the food ontology in multilinguality
is therefore an interesting research task that needs to be carried out and doing this requires a
context-aware recommender framework.
References
[1] M. R. Saeed, C. Chelmis, V. K. Prasanna, Automatic integration and querying of semantic
rich heterogeneous data: Laying the foundations for semantic web of things, in: Managing
the Web of Things, Elsevier, 2017, pp. 251–273.
[2] F. Ortiz-Rodriguez, J. M. Medina-Quintero, S. Tiwari, V. Villanueva, Egodo ontology: Shar-
ing, retrieving, and exchanging legal documentation across e-government, in: Futuristic
Trends for Sustainable Development and Sustainable Ecosystems, IGI Global, 2022, pp.
261–276.
[3] F. Ortiz-Rodriguez, S. Tiwari, R. Panchal, J. M. Medina-Quintero, R. Barrera, Mexin:
Multidialectal ontology supporting nlp approach to improve government electronic com-
munication with the mexican ethnic groups, in: DG. O 2022: The 23rd Annual International
Conference on Digital Government Research, 2022, pp. 461–463.
[4] R. Panchal, P. Swaminarayan, S. Tiwari, F. Ortiz-Rodriguez, Aishe-onto: a semantic model
for public higher education universities, in: DG. O2021: The 22nd Annual International
Conference on Digital Government Research, 2021, pp. 545–547.
[5] J. Freyne, S. Berkovsky, Intelligent food planning: personalized recipe recommendation,
in: Proceedings of the 15th international conference on Intelligent user interfaces, 2010,
pp. 321–324.
[6] M. Harvey, B. Ludwig, D. Elsweiler, You are what you eat: Learning user tastes for rating
prediction, in: International symposium on string processing and information retrieval,
Springer, 2013, pp. 153–164.
[7] M. Phanich, P. Pholkul, S. Phimoltares, Food recommendation system using clustering
analysis for diabetic patients, in: 2010 International Conference on Information Science
and Applications, IEEE, 2010, pp. 1–8.
[8] M. Ge, M. Elahi, I. Fernaández-Tobías, F. Ricci, D. Massimo, Using tags and latent factors
in a food recommender system, in: Proceedings of the 5th International Conference on
Digital Health 2015, 2015, pp. 105–112.
[9] M. Runo, Foodroid: a food recommendation app for university canteens, Unpublished
semester thesis, Swiss Federal Institute of Theology, Zurich (2011).
[10] C.-J. Su, Y.-A. Chen, C.-W. Chih, Personalized ubiquitous diet plan service based on ontology
and web services, International Journal of Information and Education Technology 3 (2013)
522.
[11] A. B. Evert, J. L. Boucher, M. Cypress, S. A. Dunbar, M. J. Franz, E. J. Mayer-Davis, J. J.
Neumiller, R. Nwankwo, C. L. Verdi, P. Urbanski, et al., Nutrition therapy recommendations
for the management of adults with diabetes, Diabetes care 37 (2014) S120–S143.
[12] M. L. LeFevre, U. P. S. T. Force*, Behavioral counseling to promote a healthful diet and
physical activity for cardiovascular disease prevention in adults with cardiovascular risk
factors: Us preventive services task force recommendation statement, Annals of internal
medicine 161 (2014) 587–593.
[13] J. Freyne, S. Berkovsky, Evaluating recommender systems for supportive technologies, in:
User Modeling and Adaptation for Daily Routines, Springer, 2013, pp. 195–217.
[14] C.-Y. Teng, Y.-R. Lin, L. A. Adamic, Recipe recommendation using ingredient networks, in:
Proceedings of the 4th annual ACM web science conference, 2012, pp. 298–307.
[15] H. Jung, K. Chung, Knowledge-based dietary nutrition recommendation for obese man-
agement, Information Technology and Management 17 (2016) 29–42.
[16] T. De Pessemier, S. Dooms, L. Martens, A food recommender for patients in a care facility,
in: Proceedings of the 7th ACM conference on Recommender systems, 2013, pp. 209–212.
[17] P. Forbes, M. Zhu, Content-boosted matrix factorization for recommender systems: ex-
periments with recipe recommendation, in: Proceedings of the fifth ACM conference on
Recommender systems, 2011, pp. 261–264.
[18] G. Geleijnse, P. Nachtigall, P. van Kaam, L. Wijgergangs, A personalized recipe advice
system to promote healthful choices, in: Proceedings of the 16th international conference
on Intelligent user interfaces, 2011, pp. 437–438.
[19] M. Svensson, K. Höök, R. Cöster, Designing and evaluating kalas: A social navigation
system for food recipes, ACM Transactions on Computer-Human Interaction (TOCHI) 12
(2005) 374–400.
[20] F. Rehman, O. Khalid, K. Bilal, S. A. Madani, et al., Diet-right: A smart food recommendation
system, KSII Transactions on Internet and Information Systems (TIIS) 11 (2017) 2910–2925.
[21] M. Ueda, S. Asanuma, Y. Miyawaki, S. Nakajima, Recipe recommendation method by
considering the users preference and ingredient quantity of target recipe, in: Proceedings
of the international multiconference of engineers and computer scientists, volume 1, 2014,
pp. 12–14.
[22] Y. Van Pinxteren, G. Geleijnse, P. Kamsteeg, Deriving a recipe similarity measure for
recommending healthful meals, in: Proceedings of the 16th international conference on
Intelligent user interfaces, 2011, pp. 105–114.
[23] L. Yang, C.-K. Hsieh, H. Yang, J. P. Pollak, N. Dell, S. Belongie, C. Cole, D. Estrin, Yum-me: a
personalized nutrient-based meal recommender system, ACM Transactions on Information
Systems (TOIS) 36 (2017) 1–31.
[24] K. TONBALAK, D. ONGAN, Yaşlılıkta beslenme durumunu etkileyen engellilik sorunlarına
435–440karşı uygulanan beslenme çözümleri, İzmir Katip Çelebi Üniversitesi Sağlık
Bilimleri Fakültesi Dergisi 7 (????).
[25] S. A. Madani, Diet-right: A smart food recommendation system (2017).
[26] M. El-Dosuky, M. Z. Rashad, T. Hamza, A. El-Bassiouny, Food recommendation using
ontology and heuristics, in: International conference on advanced machine learning
technologies and applications, Springer, 2012, pp. 423–429.
[27] C. Trattner, D. Elsweiler, Food recommender systems: important contributions, challenges
and future research directions, arXiv preprint arXiv:1711.02760 (2017).
[28] B. Raj Kumar, K. Latha, Dfrs: Diet food recommendation system for diabetic patients
based on ontology, Int J Appl Eng Res 10 (2015) 2765–70.
[29] D. Çelik, Foodwiki: Ontology-driven mobile safe food consumption system, The scientific
World journal 2015 (2015).