Adopting the Mediterranean Diet Score in a Diet Management System

Luca Anselma, Mirko Di Lascio, Antonio Lieto and Alessandro Mazzei

Dipartimento di Informatica, Universit

a di Torino, Corso Svizzera 185, Torino, Italy

Keywords:

Diet Management, Mediterranean Diet, Artiﬁcial Intelligence, Constraints, Ontology.

Abstract:

In this work we want to study the possibility to integrate a Mediterranean Diet Score into an existing diet

management architecture. The main goal is to integrate the quantitative constraints on macronutrients with the

qualitative constraints encoded in the Mediterranean Diet. This paper presents some preliminary results on

this roadmap.

1 INTRODUCTION

Diet formalization can be understood by humans as

the fulﬁllment of some numeric constraints on quan-

tity and time for food consumption. However, for psy-

chological and cognitive reasons, it can be very difﬁ-

cult for a user to understand the numeric constraints

regarding micro and macro nutrients. For this speciﬁc

task, computers can help people both in (1) comput-

ing the values of the macronutrients in the meal ac-

tually eaten, and (2) reminding the best choices for

next meals (Anselma and Mazzei, 2015). However,

it remains open the problem of explanation, that is,

in order to guide an user to improve its dietetic be-

haviour, the computer needs to explain in details of

the diet transgression (Anselma and Mazzei, 2017).

In contrast to numeric constraints, Mediterranean

diet is an example that, at a high level, can be formal-

ized as a set of qualitative constraints on food (Pana-

giotakos et al., 2007; Panagiotakos et al., 2006). In

particular, by assuming a standard quantity for a sin-

gle portion of an ingredient in a recipe (Societ

a Ital-

iana di Nutrizione Umana (SINU), 2014), we can ex-

plain the diet using simple sentences in natural lan-

guage, for instance Use daily olive oil in cooking. In

this way, also diet transgressions and their possible

compensation can be communicated to the user in a

more transparent and efﬁcient way, e.g. You shouldn’t

eat again red meat now because this week you have

eaten it twice. However, it is not trivial to “integrate”

high-level rules of the Mediterranean diet with the

food actually eaten for various reasons. For exam-

ple, if a user eats “caponata”, it is not trivial to in-

fer that (i) the dish contains eggplants, (ii) eggplants

are vegetables (thus the dish contributes to the con-

straint Eat more than 18 servings of vegetables each

month, see Table 1), (iii) it is cooked using olive oil

(thus, it contributes also to the constraint Use daily

olive oil in cooking). Notice that using the recipe of

caponata could be not enough. For example, point

(ii) is a piece of information not usually available in a

recipe (Mazzei, 2014).

In this paper we describe the ﬁrst steps towards

the integration of Mediterranean diet into an exit-

ing framework for the management of the diet called

MADiMan (Multimedia Architecture for Diet Man-

agement) (Anselma et al., 2018). In previous work,

some questions concerning the general architecture of

the MADiMan system (Mazzei et al., 2015; Anselma

and Mazzei, 2015), concerning the capacities of the

reasoning module with a number of computer simu-

lations with virtual agents based on hospital menus

(Anselma et al., 2017) and on Mediterranean menus

(Anselma et al., 2018) and concerning the Natural

Language Generation module (Anselma and Mazzei,

2018) have been addressed.

To the best of our knowledge there is no other

proposal for the automatic computation of a Mediter-

ranean score. However, some researchers proposed

some forms of perfect diet and implemented systems

for their computation. Some authors applied Opera-

tional Research techniques for the planning of a per-

fect diet (see the survey in (Lancaster, 1992) or, more

recently, (Bas, 2014)). Operational Research tech-

niques (often based on the simplex method) plan an

entire diet and they do not support the user in choos-

ing a meal. In (Buisson, 2008) a user can assess the

compatibility of a single meal to a norm and the sys-

tem suggests some actions for balancing the user’s

meal. A different approach was proposed in (Mai-

670

Anselma, L., Di Lascio, M., Lieto, A. and Mazzei, A.

Adopting the Mediterranean Diet Score in a Diet Management System.

DOI: 10.5220/0009162506700676

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 5: HEALTHINF, pages 670-676

ISBN: 978-989-758-398-8; ISSN: 2184-4305

DB Recipes

DB Users

Recipe

Code

CheckYourMeal!

app

NLU/IE

service

Reasoner

service

NLG

service

DietManager

service

[2270·4, 2270·4]

Thurs

Sun Mon Tues Weds

Fri Sat

[2690,2690]

[2205,2465]

Mediterranean

reference values

Reasoner service

LARN portions

Figure 1: The current MADiMan architecture (on the left) and a detail of the reasoner module (on the right). Note that the

reasoning is based on STP numeric constraints for macronutrients, on ontological constraints for Mediterranean References

Values (MRVs), and on standard LARN portions deﬁnition.

mone et al., 2018), where “a motivational platform for

supporting the monitoring of users’ behaviors and for

persuading them to follow healthy lifestyles” it has

been presented. In (Maimone et al., 2018), the au-

thors provide a case study based on diet and physical

activity that implements an ontological reasoner.

The speciﬁc research questions that we want to in-

vestigate on in this paper are (1) the use of ontologies

for formalizing the Mediterranean score as deﬁned in

(Panagiotakos et al., 2007; Panagiotakos et al., 2006)

(see Section 3), (2) the design of a procedure, based

on ontological formalization, for the computation of

this Mediterranean score on some human-produced

data (see Section 4).

2 THE MADiMan FRAMEWORK

MADiMan (Multimedia Application for Diet Man-

agement) is an ongoing project

with the aim of build-

ing a virtual assistant that is able to: recover the nu-

tritional information directly from a speciﬁc recipe,

reason over recipes and diets with ﬂexibility, i.e. by

allowing some forms of diet disobedience, and per-

suade the user to minimize such acts of disobedience

(Anselma et al., 2017).

The MADiMan architecture is composed by var-

ious modules (Fig. 1): a mobile app called CheckY-

ourMeal!, a numerical reasoner that decides the com-

patibility of a speciﬁc dish in some point of the diet

http://di.unito.it/madiman

(Anselma et al., 2017), an information extraction

module used to compute the nutrient values of a spe-

ciﬁc recipe, a Natural Language Generation service

that converts the results of the computation to a tex-

tual form (Anselma and Mazzei, 2017).

In Fig. 1 we depict the architecture of the system

implementing the MADiMan virtual dietitian. The in-

formation ﬂow is: (1) A user, by using the CheckY-

ourMeal! app, recovers the speciﬁc recipe of a dish

which she wants to eat. (2) The app, communicating

with the DietManager service, retrieves the user diet

together with the list of the food that the user has eaten

in the last days. (3) The NLU/IE module computes the

salient nutrition information about the speciﬁc meal.

(4) The Reasoner, using the user diet and the list of

the food that she has been eaten in the last days, pro-

duces the ﬁnal recommendation about the dish for the

user. (5) The NLGenerator uses the recommendation

given by the Reasoner, produces an explanation for

the user in simple natural language. (6) The DietMan-

ager sends the result produced by the NLGenerator to

the app: the user will see this ﬁnal result on her smart-

phone. If the user decides to eat the dish, the app will

send this information to the DietManager that will up-

date the list of food eaten.

A crucial point in the proposal of new methods

for helping people with computer is their evaluation.

With this aim, some different paradigms have been

used for evaluating MADiMan.

At ﬁrst, a simulation was performed by using vir-

tual agents, modeling different kinds of users, and real

data regarding hospital menus, for evaluating the nu-

Adopting the Mediterranean Diet Score in a Diet Management System

671

meric reasoning module of MADiMan. In particular,

it has been proved that MADiMan overcomes a rea-

sonable baseline (Anselma et al., 2017). Then, the

same experiment has been replicated by using menus

from the Gedeone database, that is a collection of 500

Mediterranean recipes annotated with their macronu-

trient contents (Anselma et al., 2018).

More recently, it has been developed the Check-

YourMeal! app in order to evaluate the performance

and the usability of the whole architecture with hu-

man evaluation into a realistic context. This app has

been used in a ﬁrst human-evaluation experiment to

test the appealing of the automatically generated text

messages (Anselma and Mazzei, 2018). The authors

asked the testers to play a diet game for a simulated

period of two weeks, spending at least 15 minutes

of their time. A tester had to imagine to eat for

two weeks in a restaurant: for each slot of the week

(breakfast, lunch, supper from Monday to Sunday)

she had to choose only among the menus proposed in

the app. In the simulation, the menus were randomly

generated by considering the recipes of the Gedeone

database (Anselma et al., 2018). During the game,

the behaviour of the tester has been logged and at

the end of the week a diet score has been automati-

cally computed. This score measured the proximity

on macronutrients to their ideal values. The logs of

the interaction of the tester with MADiMAn contain

information on which menus – and thus which dishes

of the Gedeone recipe book – have been actually eaten

by the user during the week

(Anselma and Mazzei,

2018).

In this paper, we will use the information con-

tained in the logs of the experiment described in

(Anselma and Mazzei, 2018) as input data for com-

puting the adherence of a tester to the Mediterranean

diet. Indeed, the Mediterranean diet is often described

as a number of prescriptions over one week.

In Section 3 we formalize the Mediterranean score

by using ontological models and in Section 4 we use

data from (Anselma and Mazzei, 2018) in order to

validate the formalization.

3 FORMALIZING THE

MEDITERRANEAN SCORE BY

USING ONTOLOGIES

The Mediterranean diet has been largely investigated

in the medical and cultural heritage literature since

it has a proven relation with longevity and a healthy

Moreover, at the end of the experiment, the testers had

to compile a feedback web form.

life (Keys and Keys, 1975; Soﬁ et al., 2008; Estruch

et al., 2013; Moro, 2016). In particular, the Mediter-

ranean diet is an overall dietary pattern that has long

been associated with lower incidence of cardiovascu-

lar disease and cancer (Keys et al., 1986). This diet

is an example of a lifestyle guide that can be formal-

ized as a set of qualitative constraints about various

categories of foods and their frequency of consump-

tion. Its formalization and application in a diet man-

agement system is particularly interesting since it can

be viewed as a different source of information with

respect to the standard account of calories among the

various (macro and micro) nutrients. Moreover, as

stated in (Panagiotakos et al., 2006): “It has been sug-

gested that overall dietary patterns and not single nu-

trients should be studied, since food items might have

a synergistic and antagonistic effect on health.”.

In (Panagiotakos et al., 2006; Panagiotakos et al.,

2007) the authors proposed a number of Mediter-

ranean food intake reference values (MRVs) for com-

puting a a monthly Mediterranean diet score (MED).

In Table 1 we adapt the MRVs reported in (Panagio-

takos et al., 2006) after changing the frequency of

consumption from month to week (e.g. a frequency of

5-8 servings/month becomes 1-2 servings/week). In

the table there are eleven food categories and for each

category a score in the range 0-5 is given by consid-

ering the number of servings with respect to the MRV

(so the MED score is an integer between 0 and 55).

Most categories – i.e. those that include foods close

to the Mediterranean dietary pattern – have a more is

better score: for instance the Fruits category gives 5

points if the user assumes more than 4.5 portions in

a week. In contrast, red meat, poultry, full fat dairy

products – that are discouraged in the Mediterranean

diet – have a less is better score: for instance the red

meat category gives 5 points if the user does not eat

any red meat at all during the week. An peculiar case

is the consumption of alcoholic beverages, where a

non-monotonic function has been adopted: the low-

est score is given both when more than 700 ml/day

or 0 ml/day is consumed (the motivation is that in the

Mediterranean diet a moderate consumption of alco-

holic beverages is recommended).

In order to compute the MED score, it is necessary

to map the speciﬁc ingredients of a speciﬁc recipe into

the categories of Table 1. In particular, for the exper-

iments described in Section 2, we need to map the

ingredients used for the Gedeone recipes. We decided

to employ an ontology for this speciﬁc task and cast

this problem as a typical ontological problem of in-

stance checking. By starting from an existing ontol-

ogy called PerkApp

(Bailoni et al., 2016; Dragoni

https://horus-ai.fbk.eu/helis/

HEALTHINF 2020 - 13th International Conference on Health Informatics

672

Figure 2: An Extended Branch of the PerkApp ontology showing the connection between Food types and LARN portions.

et al., 2017; Maimone et al., 2018; Donadello et al.,

2019), we deﬁned a number of subclasses needed to

encode the Gedeone ingredients as instances of the

ontology. The PerkApp ontology is built around the

basic concepts of Food (e.g. Pasta alla Carbonara),

Nutrient (e.g. Carbs, Protein, Lipids, Vitamin, Min-

eral), Timespan and Meal (e.g., Lunch, Afternoon).

All these classes, and a number of derived subclasses,

are necessary to encode the elements of a (healthy)

diet.

We have extended the PerkApp ontology by in-

cluding a class LarnPortion encoding a conversion

table deﬁned in (Societ

a Italiana di Nutrizione Umana

(SINU), 2014) that maps, for each Food type avail-

able in the ontology, the corresponding quantity in

standard servings assumed in the Mediterranean Diet

(more detailed information about the LARN are avail-

able in the next section). Fig. 2 presents a branch

of the extended ontology connecting Food types and

LARNPortion. Such expansion allowed us to model,

for example, the standard Mediterranean Diet portion

for the Food type Fish by explicitly including the ax-

iom that any instance x ∈ Fish has a corresponding

FishPortion (a reiﬁed instance of the class LARN

portion).

Such reiﬁed instance is then additionally linked,

via the datatype property hasLARNvalue, to a cor-

responding integer representing the effective LARN

value, in grams, for all the instances of the Fish cate-

gory (that is 150 grams in such case). The same mod-

elling procedure has been followed for each Food type

available in the ontology. Such ontological extension,

as a ﬁrst consequence, directly helped us to deal with

point (ii) described in the introduction (e.g., eggplants

are vegetables) concerning the completion, with taxo-

nomical information, about single instances of Food.

In addition, such encoding allowed us to query the ex-

tended ontology in order to ask, for each instance of

food acquired from the log coming from the CheckY-

ourMeal! app, the associated portions assumed in the

Mediterranean Diet, as indicated in the LARN guide-

lines (Societ

a Italiana di Nutrizione Umana (SINU),

2014). Note that the mapping between the instances

of food available in the log and the instances in the

ontology is currently obtained with a simple string-

matching procedure. An example of SPARQL query

extracting the corresponding Food category for an in-

stance labelled as Swordﬁsh, and additionally asking

for the associated LARN portion is the following:

SELECT DISTINCT ?food ?type ?portion

WHERE {

?food a ?type .

?type rdfs:subClassOf* ?x .

?food rdfs:label "Swordfish" @en.

?food pka:hasLARNportion ?p .

?p pka:hasLARNvalue ?portion .

}

We have additionally started the population of the

ontology with instances of Food belonging to the

Gedeone ingredients that were not originally avail-

able in the PerkApp ontology. The current version

of the ontology contains 150 out of 700 instances of

the Gedeone ingredients.

4 COMPUTING MED FROM

REAL DATA

The process used for the computation of the MED

score starting from the simulation log produced in

the experiment with CheckYourMeal! (Anselma and

Mazzei, 2018) (see Section 2) can be understood as

composed by four basic steps. Assuming to have

the list and quantity of the food eaten during a week

thanks to the respective recipes, we have to:

1. Convert the quantity of ingredients into grams,

2. Convert the quantity of ingredients in grams into

standard servings deﬁned in (Societ

a Italiana di

Nutrizione Umana (SINU), 2014),

3. Convert the ingredients into the corresponding

MED category deﬁned in (Panagiotakos et al.,

2006) by using the food ontology described in

Section 3,

Adopting the Mediterranean Diet Score in a Diet Management System

673

Table 1: The Mediterranean food intake reference values (MRVs) which are used to compute the MED diet score (adapted

from (Panagiotakos et al., 2007; Panagiotakos et al., 2006)).

Frequency of consumption

How often do you consume (servings/week) Never 0-1 1-2 2-3 3-4.5 >4.5

Non-reﬁned cereals (whole grain bread, pasta, rice, etc.) 0 1 2 3 4 5

Potatoes 0 1 2 3 4 5

Fruits 0 1 2 3 4 5

Vegetables 0 1 2 3 4 5

Legumes 0 1 2 3 4 5

Fish 0 1 2 3 4 5

Red meat and products 5 4 3 2 1 0

Poultry 5 4 3 2 1 0

Full fat dairy products (cheese, yogurt, and milk) 5 4 3 2 1 0

Use of olive oil in cooking (times/week)

Never Rare 0-1 1-3 3-5 Daily

0 1 2 3 4 5

Alcoholic beverages (ml/day, 100 ml=12 g ethanol)

<300 300 400 500 600 >700 or 0

5 4 3 2 1 0

Table 2: The weekly MED scores computed for three dis-

tinct users participating to the CheckYourMeal! experi-

ment. The maximum score is 55.

User 1 User 2 User 3

Non-reﬁned cereals 5 5 5

Potatoes 2 2 2

Fruits 3 5 4

Vegetables 5 5 5

Legumes 3 1 2

Fish 2 2 3

Red meat and products 4 5 4

Poultry 5 5 5

Full fat dairy products 1 1 3

Use of olive oil 5 5 5

Alcoholic beverages 5 5 5

MED score

40 41 43

(73%) (75%) (78%)

4. Sum and normalize the score of the categories

over the week by using the scoring of Table 1.

There are two challenging issues arising from step

(1) of this process, that are: (i) in order to normalize

all quantities in grams, we have to know the speciﬁc

weight of each liquid mentioned in the recipe, and (ii)

often some quantities occur under vague expressions

(“a glass of water”, “a handful of almonds”, etc.).

These problems can be solved assuming standard

quantities for the various expressions (e.g. a glass

of water contains approximately 200 milliliters and

weighs approximately 200 grams) since conversions

between expressions and volumes are quite standard-

ized

A key point of step (2) regards the encoding of

See http://www.onlineconversion.com/weight volume

cooking.htm for a possible reference.

standard quantities for servings. We decided to use

the standard quantities suggested by SINU (Societ

Italiana di Nutrizione Umana – Italian Society for Hu-

man Nutrition) (Societ

a Italiana di Nutrizione Umana

(SINU), 2014). SINU publishes the “LARN” (Liv-

elli di Assunzione di Riferimento di Nutrienti ed en-

ergia per la popolazione italiana – Nutrient Refer-

ence Intake and Energy Levels for the Italian Popula-

tion) (Societ

a Italiana di Nutrizione Umana (SINU),

2014), which are the standard reference for the nu-

trition of Italian people (analogous documents exist

for other countries); in particular in LARN it is possi-

ble to ﬁnd a quantitative standard for deﬁning a serv-

ing amount. For example, a serving of dried legumes

is deﬁned as 50 grams, corresponding to 3-4 table-

spoons, and a serving of fresh legumes is deﬁned as

150 grams, corresponding to half a plate. Note that

the food categories described in LARN (Societ

a Ital-

iana di Nutrizione Umana (SINU), 2014) are distinct

from the MED categories described in Table 1. Thus,

we need to enrich the ontological model, as described

in Section 3, with a number of subclasses connecting

the two knowledge sources. For instance, we need to

implement the concept of dried legume, which was

not originally present in the PerkApp ontology.

In order to give an example of computation of the

MED score, let us suppose that we found from the

log that during the week the user ate 3 cups of whole

grain rice. In step (1), knowing that a cup of whole

grain rice is equivalent to 190 grams, this quantity is

converted into 570 grams of whole-grain rice. In step

(2), we ﬁnd from LARN that a serving of whole grain

rice is 80 grams, so that the user ate 7 servings of

whole grain rice. In step (3), using ontological rea-

soning (i.e., SPARQL queries) we ﬁnd that 7 servings

HEALTHINF 2020 - 13th International Conference on Health Informatics

674

Table 3: Amounts of macronutrients in kcal for the food

ingested by the three users participating to the CheckY-

ourMeal! experiment.

User 1 User 2 User 3

Carbohydrates taken 9993 6810 15128

Proteins taken 1819 1200 1937

Lipids taken 4906 3845 7078

Total energy taken 16718 11855 24143

of whole grain rice correspond to 7 servings of non-

reﬁned cereals. Finally, in step (4), from Table 1 we

ﬁnd that this quantity contributes to the weekly MED

score with 5 points.

In Table 2 we report the MED scores for three dis-

tinct users participating to the CheckYourMeal! ex-

periment. The three users have assumed consider-

ably different amounts of food and of macronutrients,

as shown by Table 3, where User 3 has consumed a

high amount of food during the week (in average 3442

kcal/day), User 2 a low amount (1694 kcal/day) and

User 1 an intermediate value (2388 kcal/day). The

MED scores for the three users are similar (Table 2).

One can notice that the use of a Mediterranean recipe

book has resulted in a high MED score for all three

users. In particular, there are high scores for the con-

sumption of non-reﬁned cereals, olive oil and vegeta-

bles for all the users, and a moderate score for the

consumption of potatoes. This example conﬁrms the

intuition that quantitative constraints arising from di-

etary reference values on macronutrients and qualita-

tive constraints arising from the principles of Mediter-

ranean diet are two different notions that needed to be

treated with different mechanisms.

5 CONCLUSIONS

In this paper we have described a ﬁrst proposal for the

ontological formalization of the Mediterranean score

MED as proposed in (Panagiotakos et al., 2007; Pana-

giotakos et al., 2006) as an extension of the food on-

tology PerkApp (Maimone et al., 2018). Moreover,

we have discussed how to effectively implement the

computation of MED in the framework of MADiMan,

a system for diet management. By using data from the

CheckYourMeal! experiment, we have computed in

MADiMan the MED score for three distinct users and

so we have empirically showed that macronutrients

and MED score are independent measures for rating

the quality of meals for a healthy diet, and as a conse-

quence it necessary to compute both independently.

To the best of our knowledge, this is the ﬁrst work

that formalizes the MED score in terms of ontolog-

ical constraints and proposes an effective procedure

for computing it.

In future work we intend to use the computation

of the MED score to augment the answers returned

by MADiMan with an explanation about the compat-

ibility of a speciﬁc dish with the Mediterranean diet.

REFERENCES

Anselma, L. and Mazzei, A. (2015). Towards Diet Manage-

ment with Automatic Reasoning and Persuasive Natu-

ral Language Generation. In Progress in Artiﬁcial In-

telligence - 17th Portuguese Conference on Artiﬁcial

Intelligence, EPIA 2015, Coimbra, Portugal, Septem-

ber 8-11, 2015. Proceedings, pages 79–90.

Anselma, L. and Mazzei, A. (2017). An approach for ex-

plaining reasoning on the diet domain. In Basile, P.,

Croce, D., and Guerini, M., editors, Proc. of the 1st

Workshop on Natural Language for Artiﬁcial Intelli-

gence co-located with AI*IA 2017, Bari, Italy., vol-

ume 1983 of CEUR Workshop Proceedings, pages 4–

17. CEUR-WS.org.

Anselma, L. and Mazzei, A. (2018). Designing and testing

the messages produced by a virtual dietitian. In Krah-

mer, E., Gatt, A., and Goudbeek, M., editors, Pro-

ceedings of the 11th International Conference on Nat-

ural Language Generation, Tilburg University, The

Netherlands, November 5-8, 2018, pages 244–253.

Association for Computational Linguistics.

Anselma, L., Mazzei, A., and De Michieli, F. (2017).

An artiﬁcial intelligence framework for compensating

transgressions and its application to diet management.

Journal of Biomedical Informatics, 68:58–70.

Anselma, L., Mazzei, A., and Pirone, A. (2018). Automatic

reasoning evaluation in diet management based on an

Italian cookbook. In Proceedings of the Joint Work-

shop on Multimedia for Cooking and Eating Activities

and Multimedia Assisted Dietary Management, CEA/-

MADiMa ’18, pages 59–62, New York, NY, USA.

ACM.

Bailoni, T., Dragoni, M., Eccher, C., Guerini, M.,

and Maimone, R. (2016). Healthy lifestyle sup-

port: the perkapp ontology. In OWL: Experiences

and Directions–Reasoner Evaluation, pages 15–23.

Springer.

Bas, E. (2014). A robust optimization approach to diet prob-

lem with overall glycemic load as objective function.

Applied Mathematical Modelling, 38(19–20):4926 –

4940.

Buisson, J.-C. (2008). Nutri-educ, a nutrition software ap-

plication for balancing meals, using fuzzy arithmetic

and heuristic search algorithms. Artif. Intell. Med.,

42(3):213–227.

Donadello, I., Dragoni, M., and Eccher, C. (2019). Per-

suasive explanation of reasoning inferences on dietary

data. In Demidova, E., Dietze, S., Breslin, J. G.,

Gottschalk, S., Cimiano, P., Ell, B., Lawrynowicz,

A., Moss, L., and Ngomo, A. N., editors, Joint Pro-

ceedings of the 6th International Workshop on Dataset

Adopting the Mediterranean Diet Score in a Diet Management System

675

PROFlLing and Search & the 1st Workshop on Se-

mantic Explainability co-located with the 18th In-

ternational Semantic Web Conference (ISWC 2019),

Auckland, New Zealand, October 27, 2019, volume

2465 of CEUR Workshop Proceedings, pages 46–61.

CEUR-WS.org.

Dragoni, M., Bailoni, T., Eccher, C., Guerini, M., and Mai-

mone, R. (2017). A semantic-enabled platform for

supporting healthy lifestyles. In Proceedings of the

Symposium on Applied Computing, SAC ’17, pages

315–322, New York, NY, USA. ACM.

Estruch, R., Ros, E., Salas-Salvad

o, J., Covas, M.-I.,

Corella, D., Ar

os, F., G

omez-Gracia, E., Ruiz-

Guti

errez, V., Fiol, M., Lapetra, J., et al. (2013).

Primary prevention of cardiovascular disease with

a mediterranean diet. New England Journal of

Medicine, 368(14):1279–1290.

Keys, A. and Keys, M. (1975). How to Eat Well and Stay

Well the Mediterranean Way. Doubleday.

Keys, A., Mienotti, A., Karvonen, M. J., Aravanis, C.,

Blackburn, H., Buzina, R., Djordjevic, B., Dontas, A.,

Fidanza, F., Keys, M. H., et al. (1986). The diet and

15-year death rate in the seven countries study. Amer-

ican journal of epidemiology, 124(6):903–915.

Lancaster, L. M. (1992). The history of the application of

mathematical programming to menu planning. Euro-

pean Journal of Operational Research, 57(3):339 –

347.

Maimone, R., Guerini, M., Dragoni, M., Bailoni, T., and

Eccher, C. (2018). Perkapp: A general purpose per-

suasion architecture for healthy lifestyles. Journal of

Biomedical Informatics, 82:70–87.

Mazzei, A. (2014). On the lexical coverage of some re-

sources on Italian cooking recipes. In Proc. of CLiC-it

2014, First Italian Conference on Computational Lin-

guistics, pages 254–259.

Mazzei, A., Anselma, L., De Michieli, F., Bolioli, A., Casu,

M., Gerbrandy, J., and Lunardi, I. (2015). Mobile

computing and artiﬁcial intelligence for diet manage-

ment. Lecture Notes in Computer Science, 9281:342–

349.

Moro, E. (2016). The Mediterranean diet from Ancel Keys

to the UNESCO cultural heritage. A pattern of sustain-

able development between myth and reality. Procedia

- Social and Behavioral Sciences, 223:655 – 661. 2nd

International Symposium NEW METROPOLITAN

PERSPECTIVES - Strategic planning, spatial plan-

ning, economic programs and decision support tools,

through the implementation of Horizon/Europe2020.

ISTH2020, Reggio Calabria (Italy), 18-20 May 2016.

Panagiotakos, D. B., Pitsavos, C., Arvaniti, F., and Ste-

fanadis, C. (2007). Adherence to the Mediterranean

food pattern predicts the prevalence of hypertension,

hypercholesterolemia, diabetes and obesity, among

healthy adults; the accuracy of the meddietscore. Pre-

ventive medicine, 44(4):335–340.

Panagiotakos, D. B., Pitsavos, C., and Stefanadis, C. (2006).

Dietary patterns: a Mediterranean diet score and its

relation to clinical and biological markers of cardio-

vascular disease risk. Nutrition, Metabolism and Car-

diovascular Diseases, 16(8):559–568.

Societ

a Italiana di Nutrizione Umana (SINU) (2014).

LARN: Livelli di Assunzione di Riferimento di Nu-

trienti ed Energia per la Popolazione Italiana.

Soﬁ, F., Cesari, F., Abbate, R., Gensini, G. F., and Casini,

A. (2008). Adherence to mediterranean diet and health

status: meta-analysis. Bmj, 337:a1344.

HEALTHINF 2020 - 13th International Conference on Health Informatics

676