Sage Journals: Discover world-class research

Abstract

Unhealthy diets are a major modifiable risk factor for noncommunicable diseases and the leading cause of morbidity and mortality world-wide. Nutrition assessment is crucial for understanding and preventing unhealthy diets, but often relies on self-reporting, which is burdensome, error-prone, and ineffective for long-term tracking. Digital receipts from loyalty cards, enriched with product nutrition information, offer a promising alternative. Current regulations support user-consented access to such receipts and mandates food nutrition information provision, creating a viable legislative foundation for sharing and using digital receipts. Shared ontologies enable consistent management and exchange of digital receipts and food product information across sources and applications. Yet, no current ontology describes enriched digital receipts at product and basket levels with detailed nutrition metrics. We present NutriLink, an ontology connecting digital receipts to product nutrition information and structured dietary recommendations. NutriLink supports evaluating nutritional quality of purchases within and across baskets and delivering structured dietary recommendations. It integrates with the established ontologies, including FoodOn, GoodRelations, and AGROVOC, as well as with schema.org concepts. We demonstrate NutriLink’s value through deployment in a fully automated dietary counseling system with 76 users. NutriLink is freely and openly available for research and practice.

Keywords

food ontology food semantics digital receipts semantic technologies

1. Introduction

The four leading noncommunicable diseases (NCDs)—cardiovascular diseases, cancers, chronic respiratory diseases, and diabetes—are the primary cause of morbidity and mortality world-wide (World Health Organization, 2024b). Diabetes alone led to 1.6 million deaths in 2021 (World Health Organization, 2024a), and it can cause blindness, kidney failure, heart attacks, stroke, as well as nerve damage and reduced circulation, which might lead to limb loss (World Health Organization, 2024a). The number of people living with diabetes has increased rapidly from 200 million in 1990 to 830 million in 2022 (World Health Organization, 2024a). Its global economic burden is projected to reach USD2.5 trillion by 2030—nearly double the cost of USD1.3 trillion in 2015 (Bommer et al., 2018). A major modifiable behavioral risk factor contributing to NCDs is an unhealthy diet, specifically excess salt, sugar, and fats (World Health Organization, 2024b). To understand and prevent unhealthy diets and thereby mitigate NCDs, nutrition assessment and monitoring are pivotal. However, current nutrition assessment tools typically rely on self-reported data such as (digital) food diaries and food records (Chen et al., 2018; Swan et al., 2017). They are prone to recall bias and require substantial manual effort (Bakland et al., 2020; Gemming & Mhurchu, 2016; Naska et al., 2017). These limitations hinder the design of effective and sustainable dietary interventions.

Grocery receipts, particularly digital receipts from loyalty cards, have the potential to revolutionize nutrition assessment and monitoring by integrating food composition data (Mönninghoff et al., 2022; Odunitan-Wayas et al., 2021; Sainz-De-Abajo et al., 2020; Wu et al., 2022). Regulations such as the European Union (EU) General Data Protection Regulation (GDPR; European Union, 2016) (Article 20) grant individuals the right to access and transfer their personal data. This encompasses all personal user data held by a data controller, including historic and up-to-date digital receipts on loyalty cards, in both human- and machine-readable formats. Similar regulations are proliferating world-wide, from the data portability rights that are part of the California Consumer Privacy Rights Act (Justice et al., 2021) to the People’s Republic of China Personal Information Security Specification.¹ These regulations provide a legal basis to obtain data that might be used in fully automated nutrition assessment and monitoring, benefiting both healthy individuals wishing to improve their diet as well as patients with medical needs, such as those who have undergone bariatric surgery (cf. Schönenberger et al., 2022). Compared to nutrition assessment tools based on self-reporting, digital receipts substantially reduce the reporting effort as they do not require manual logging and transcribing. However, they generally lack product nutrition information that is necessary for further analysis. Bridging this gap, EU Regulation No. 1169/2011 (European Union) has mandated the provision of nutrition information for all food products intended for sale to final consumers, including prepackaged and non-prepackaged foods, and foods sold in restaurants. While it does not explicitly require food nutrition information to be provided in digital formats, many food manufacturers, retailers, and crowdsourced food composition databases (FCDs), such as Open Food Facts,² now provide such data. These developments facilitate easier access to digital food nutrition information, increasing the potential of real-time, automated nutrition assessment and monitoring.

Ontologies establish common vocabularies, definitions, and relations between (domain) concepts and thereby facilitate data integration and knowledge management. Food ontologies, such as FoodOn (Dooley et al., 2018), provide a structured framework for organizing and representing food and food-related information, enabling data sharing and integration across different applications and domains. This information can be further linked to standardized models of physical quantities and measurement units (e.g., kilocalories) through ontologies such as QUDT (QUDTorg), while other ontologies such as GoodRelations (Hepp, 2008) offer standard ways to represent digital receipts. However, no existing ontology currently links digital receipts to food nutrition information, highlighting a critical gap in integrating dietary data.

In this paper, we introduce the ontology NutriLink. It is designed to promote standardized machine-readable description of digital receipts from loyalty cards, and enriches them with food nutrition information, from the point of purchase to the final stage of (nutrition) data analysis and recommendation. To quantitatively evaluate the nutritional quality of food purchases, this ontology also includes fine-grained product and basket Nutri-Score (Julia & Hercberg, 2017), a front-of-package nutritional labeling system in Europe. NutriLink further links to the established food ontologies for enhanced data coverage and interoperability. To verify the utility of NutriLink, we demonstrate its usage in a fully automated dietary counseling system—this scenario and system also provide the foundation for the development of NutriLink, which was based on the Simplified Agile Methodology for Ontology Development (SAMOD; Peroni, 2016).

2. Related Work

In the following, we survey the context of NutriLink, including food nutrition profiling systems (see Section 2.1), sources of food purchase and food composition data (see Section 2.2), and food and grocery purchase ontologies (see Section 2.3).

2.1. Nutri-Score and the British Food Standards Agency Nutrient Profiling System Dietary Index

Nutri-Score (Julia & Hercberg, 2017; Merz et al., 2024) is a front-of-package nutritional labeling system that categorizes products on a color-coded scale from A (green, healthiest) to E (red, least healthy), allowing for quick visual assessment and practical decision-making assistance. It was first introduced in France, and then adopted by six other European countries,³ including Switzerland (Swiss Confederation, 2022). Based on the Food Standards Agency Nutrient Profiling System (FSA-NPS), Nutri-Score evaluates food products using a scoring mechanism that considers four negative (energy; sugars; saturated fat; sodium) and three positive nutritional components (fiber; protein; percentages of fruits, vegetables, legumes, nuts, and specific oils [FVLNO; Hercberg et al., 2021], or fruits, vegetables, and legumes [FVL] in the most recent 2023 update; Merz et al., 2024). A product’s FSA score is computed from subscores in these seven dimensions and is then mapped to a final Nutri-Score (A–E).

In December 2023, the Nutri-Score algorithm was updated to allow for better discrimination between products and closer alignment with food-based dietary guidelines (Merz et al., 2024). Key changes include recategorizing some products, such as nuts and seeds; increasing maximum (negative) points for sugar (from 10 to 15) and salt (from 10 to 20); excluding oils and nuts from the FVLNO component, and considering non-nutritive sweeteners as an unfavorable factor in beverages.

Beyond individual food products, a comprehensive nutritional quality metric—the British Food Standards Agency Nutrient Profiling System Dietary Index (FSA-NPS DI; Julia & Hercberg, 2017)—is derived from products’ energy-weighted FSA scores. FSA-NPS DI simplifies implementation and reduces controversy around food categorization. Among several nutritional quality indicators of food purchases, FSA-NPS DI has demonstrated stronger calibration and discrimination performance in classifying participants’ consumption of nutrients and food categories (Wu et al., 2022).

2.2. Provenance of Food Purchase and Food Composition Data

Food purchase data has been attracting researchers’ attention because of its potential to objectively indicate dietary patterns (Green et al., 2020; Jenneson et al., 2022; Mayer et al., 2021; Wu et al., 2022). Objectively documented household food purchases, even partial food purchases, show moderate agreement with overall diet quality as measured through 24-h diet recalls (Appelhans et al., 2017; Parker et al., 2021). However, the manual collection and transcription of paper receipts is time-consuming and burdensome. Digital receipts from loyalty programs offer several advantages over paper receipts: Given a robust data pipeline to retrieve digital receipts in a machine-readable form, collecting and analyzing them require little manual effort, and transcribing them is also less error-prone. Additionally, digital receipts provide both up-to-date and historic⁴ shopping records.

Given the growth of the user base of retailers’ loyalty program in many countries, such digital receipts today may enable population-level nutrition assessment and monitoring (Erkkola et al., 2022; Green et al., 2020; Jenneson et al., 2022; Lee et al., 2021; Mönninghoff et al., 2022; Wu et al., 2022). This is particularly pronounced in Switzerland (Wu et al., 2022), where the two largest retailers—Coop and Migros—have a market share of 43% and 37.4%, respectively, in food and food-related products (swissinfo.ch, 2025). Their loyalty programs (3 million Migros Cumulus cards (Migros, 2025) and 3.3 million active Coop Supercard accounts (Coop, 2025)) capture about 75% of in-store sales (Blick, 2021), ensuring high data completeness. From a legislative perspective, the EU GDPR and similar regulations in other jurisdictions enable loyalty card users to access their digital receipts and transfer this information from data controllers, typically retailers.

Digital receipts typically do not contain product nutrition information. Nonetheless, they can be enriched leveraging information from FCDs that meet the following criteria (Pilz et al., 2022; Wu et al., 2025): (i) be able to map ambiguous (abbreviated, e.g., “FBud App.Brae,” or broad, e.g., “Apple”) item names on receipts to unique product identifiers; (ii) contain product size information, typically based on local stereotypes (e.g., a standard cucumber in France weighs 400 g); (iii) contain product nutrition information, such as energy and nutrient content per 100 g; and (iv) contain nutritionally meaningful food categorization. For Nutri-Score calculation, FCDs must also contain product FVLNO shares (Hercberg et al., 2021), or FVL shares in the 2023 Nutri-Score update (Merz et al., 2024). Extracting and calculating exact percentages from ingredient lists requires significant manual effort and domain experience.

2.3. Food and Grocery Shopping Ontologies

Researchers have been designing food ontologies to facilitate the representation and integration of food-related data. Boulos et al. (2015) provide a detailed comparison of several food ontologies: FoodWiki (Çelik, 2015), AGROVOC (Caracciolo et al., 2012), Open Food Facts (Open Food Facts), the Food Product Ontology (Kolchin & Zamula, 2013), and FOODS (Snae & Bruckner, 2008). In addition, the FoodOn ontology (Dooley et al., 2018) fills gaps in food product terminology and supports food traceability, including information such as animal and plant food sources, food categories and products, preservation processes, contact surfaces, and packaging.

These ontologies have been designed to serve specific use cases and are hence with limitations. First, they focus on individual food products, not considering the combined nutritional quality of all consumed food products. Second, they are not suitable for integrating the full Nutri-Score framework. For instance, while Open Food Facts² does include information about products’ final Nutri-Score values, it misses information about product Nutri-Score category and the product FVLNO (or FVL) percentages, both of which are required for computing product and basket FSA scores.

Apart from ontologies that focus on food products themselves, several ontologies could be leveraged to represent grocery items. One example is the GoodRelations ontology (Hepp, 2008) for e-commerce. This ontology provides a standard set of terms and relationships to describe products, services, and businesses on the Web. GoodRelations is also integrated with the Open Food Facts ontology (Open Food Facts) to represent core product data such as barcodes and brands (Hepp, 2008).

Furthermore, schema.org provides several concepts that can be used to structure grocery purchases, such as Product,⁵ Offer,⁶ and Order.⁷ Concepts from schema.org are commonly used in the established ontologies. For instance, 5 of 315 schema.org classes and 27 GoodRelations classes cover the same entities.⁸ One example is that gr:ProductOrService⁹ is a subclass of schema:Product.5 Moreover, gr:hasGTIN-14 is a subproperty of schema:productID.¹⁰ Similarly, the FoodOn ontology employs the image¹¹ property from schema.org.

None of the surveyed ontologies were directly usable for our purposes, which involve integrating digital receipts with FSA-based dietary recommendations. Nevertheless, inspired by these ontologies, we propose the NutriLink ontology and integrate it with the several established ontologies (see Section 3.2).

3. The NutriLink Ontology

The NutriLink ontology was created following SAMOD (Peroni, 2016), based on three key components: (i) digital receipts from loyalty cards; (ii) food nutrition information from an existing FCD; and (iii) aggregated information about products from the same basket. We aim to develop an ontology that effectively represents digital receipts enriched with food nutrition information and links them with dietary recommendations.

3.1. Motivating Scenario and Competency Questions

Our motivating scenario is to provide structured dietary recommendations to users given their most recent food purchases. Based on this scenario, we derived a set of competency questions (CQs) for the NutriLink ontology: NutriLink should enable a system that collects information about purchased food items, such as product quantity (CQ1). From such information, it should be able to aggregate nutrition information across baskets (CQ2), and provide the possibility to compute and display the Nutri-Score values of shopped products and baskets to users (CQ3). To further enhance user feedback, the system should be able to provide nutrition and expense analyses, where bought food items are categorized into different food groups (CQ4). Finally, the system should be able to access necessary information for generating structured dietary recommendations (CQ5). We discuss each CQ in more detail in the following.

CQ1. What is the quantity of a specific product in a basket?

Digital receipts frequently provide limited information about item quantities and the specific quantity unit. Many items on digital receipts report a quantity of simply 1 without a unit. Hence, the specific product size and corresponding measurement unit may only be deduced when integrating with an FCD. Additionally, various measurement units could be used for nutrient measurements (e.g., g and mg) and product weights and volume (e.g., g, kg, ml, and l). Unit standardization is crucial for further data analysis.

CQ2. What is the aggregated quantity and nutrition information across all products in a basket?

It should be possible to retrieve aggregated information across all items in a basket, including total price, discount (if applicable), and basket nutritional metrics. Storing aggregated basket information eliminates repetitive calculations of static data (e.g., total price of a basket) and significantly simplifies subsequent data analysis. Nevertheless, the biggest challenge to aggregate basket-level information lies in maintaining a high-quality FCD as described in Section 2.2.

CQ3. What are the product- and basket-level Nutri-Score values in an individual’s shopping history?

A clear and intuitive metric, such as the widely recognized Nutri-Score label (Julia & Hercberg, 2017), could effectively provide individuals with feedback on their overall food purchases. In Switzerland, adopting Nutri-Score is particularly promising, since many consumers are already familiar with it. If calculated across multiple shopping baskets and displayed effectively (Wu et al., 2025), Nutri-Score could help users better understand their dietary patterns and make more informed food choices aligned with their nutrition goals. The basket-level Nutri-Score can be derived from the basket FSA-NPS DI (see Section 2.1) and by applying the FSA score thresholds in the Nutri-Score definition to convert to letters A–E.

CQ4. What are the monthly energy and expense contributions of each food category for a shopper?

To enhance participants’ understanding and engagement with provided dietary recommendations, the system should be able to inform them about the food categories that contribute most to their calorie intake and spending.

CQ5. What information about the most recent baskets is needed to generate structured dietary recommendations?

To provide relevant and robust dietary recommendations, we consider users’ most recent food purchases and their nutritional quality. The nutritional quality of the recent food purchases is indicated by energy-weighted sub-FSA scores (e.g., sugar score; there are seven sub-FSA scores as introduced in Section 2.1) of all food products in a basket. This energy-weighted approach is adopted by the definition of FSA-NPS DI (Julia & Hercberg, 2017) (see Section 2.1). Using household purchase data, our nutritional analysis is based on the assumption that each individual consumes food proportionally across all categories. For instance, if a person consumes 10% of the vegetables, this person also consumes 10% of other categories, such as sweets. Although simplified, this analysis approach allows for meaningful dietary insights derived from readily available shopping records. Through discussions with dietitians and doctors from our partner organization—University Hospital of Bern/Inselspital, Switzerland—we have defined a structured recommendation format: “Increase/Reduce [FSA-NPS component] from [Food Category]” (e.g., “Reduce sugar from sweets”). More details about the structured dietary recommendations can be found in Wu et al. (2025).

3.2. Semantic Model

The NutriLink ontology has been developed with the objective to answer these CQs while defining, describing, and integrating relevant attributes of digital receipts, food nutrition information, and dietary recommendations. After analyzing the CQs in Section 3.1, we decided to reuse the following ontologies:

the QUDT (Quantities, Units, Dimensions, and Data Types) ontology¹² to work with various quantity and unit standards and for unit conversion;

the FOAF (Friend of a Friend) ontology¹³ to represent users;

the VAEM (Vocabulary for Attaching Essential Metadata) vocabulary¹⁴ to represent names; and

the Dublin Core vocabulary¹⁵ to represent images.

To expand NutriLink’s coverage of food products globally, we integrate NutriLink with the several established ontologies (see Section 2.3). The concrete integration points, manually identified by the authors, focus on nl:Product and nl:gtin, and are outlined in Table 1.

Table 1.
Description of Several Established Ontologies and schema.org, and the Integration Points Between them and NutriLink.

Ontology Description Integration Point(s) $^{a}$

FoodOn (Dooley et al., 2018) FoodOn contains vocabulary for naming food materials and their anatomical and taxonomic origins, from raw harvested food to processed food products, for humans and domesticated animals. foodon:FoodProduct is a subclass of nl:Product

GoodRelations (gr) (Hepp, 2008) GoodRelations is a lightweight ontology for exchanging e-commerce information, namely data about products, offers, points of sale, prices, terms and conditions, on the Web.
$∙$
nl:Product is a subclass of gr:ProductOrServiceModel
$∙$
nl:gtin is an equivalent property gr:hasGTIN-14

AGROVOC (Caracciolo et al., 2012) AGROVOC is a relevant Linked Open Data set about agriculture available for public use and facilitates access and visibility of data across domains and languages. agrovoc:products is an instance of nl:Product

schema.org schema.org is a collaborative, community activity with a mission to create, maintain, and promote schemas for structured data on the Internet, on web pages, in email messages, and beyond. The integration of NutriLink with schema.org concepts is done through its integration with GoodRelations and FoodOn, which already integrate schema.org concepts.

Ontology	Description	Integration Point(s) $^{a}$
FoodOn (Dooley et al., 2018)	FoodOn contains vocabulary for naming food materials and their anatomical and taxonomic origins, from raw harvested food to processed food products, for humans and domesticated animals.	foodon:FoodProduct is a subclass of nl:Product
GoodRelations (gr) (Hepp, 2008)	GoodRelations is a lightweight ontology for exchanging e-commerce information, namely data about products, offers, points of sale, prices, terms and conditions, on the Web.	$∙$ nl:Product is a subclass of gr:ProductOrServiceModel $∙$ nl:gtin is an equivalent property gr:hasGTIN-14
AGROVOC (Caracciolo et al., 2012)	AGROVOC is a relevant Linked Open Data set about agriculture available for public use and facilitates access and visibility of data across domains and languages.	agrovoc:products is an instance of nl:Product
schema.org	schema.org is a collaborative, community activity with a mission to create, maintain, and promote schemas for structured data on the Internet, on web pages, in email messages, and beyond.	The integration of NutriLink with schema.org concepts is done through its integration with GoodRelations and FoodOn, which already integrate schema.org concepts.

$^{a}$ foodon: <http://purl.obolibrary.org/foodon/>

nl: <http://purl.org/nutrilink#>

gr: <http://purl.org/goodrelations/v1#>

agrovoc: <http://aims.fao.org/aos/agrovoc/>

Integrating NutriLink with the FoodOn ontology provides a larger amount of food-related definitions, broadening the scope of NutriLink to non-consumer-oriented domains such as food production. Reciprocally, NutriLink’s deeper coverage of consumer-oriented information, even beyond our specific CQs, supplements FoodOn’s existing data. Concretely, the FoodProduct class in FoodOn is equivalent to NutriLink’s Product class. Regarding NutriLink’s integration with GoodRelations, nl:Product is a subclass of gr:ProductOrServiceModel. In addition, the data property nl:gtin is equivalent to gr:hasGTIN-14. Linking with GoodRelations enhances NutriLink’s data interoperability, facilitating compatibility with the existing Web standards. On the other hand, NutriLink complements GoodRelations by offering more advanced features. For instance, the qudt:Quantity used in NutriLink has a richer and more detailed representation of quantities, including dimensions and types, compared to the gr:QuantitativeValue¹⁶ in GoodRelations. Connecting NutriLink with AGROVOC extends its coverage beyond food to agriculture-related domains, providing multilingual support. The agrovoc:products is an instance of nl:Product. The integration of NutriLink with GoodRelations and FoodOn automatically links the relevant schema.org concepts these ontologies employ (see Section 2.3).

The W3C Web Ontology Language (OWL) formalization of our model and all relevant SPARQL queries are openly available in our GitHub repository.¹⁷ Below, we present parts of NutriLink by introducing how it implements the CQs (see Section 3.1) in the context of a fully automated dietary counseling system based on digital receipts, FoodCoach (Wu et al., 2025).

3.2.1. Overview of the NutriLink Ontology

To provide an overview of NutriLink and its semantic scope, we present its key concepts (namespace nl omitted) with brief comments when the names are not self-explanatory: Abstention (of a user from a product ingredient), Allergen, Basket, BasketAggInfo (aggregated basket information such as total price), DietCoachMajorCategory (major food category such as processed food), DietCoachMinorCategory (minor food category such as salty snacks), Dietitian, Item (items on receipts, identified by item names), MarketRegion (the geographical availability of a product), NutriScoreCategory, NutritionInformation with the three subclasses Nutrient, NutriScore, and OfComDetail (i.e., the detailed FSA scores of a product); furthermore Office (exact supermarket store), Product (products on FCD, identified by Global Trade Item Names (GTINs)), Retailer, and User.

Basket is the fundamental class for describing digital receipts. It represents the collection of Item instances from a specific basket or a single-shopping session, uniquely identified by a hash of the checkout timestamp, user identifier, and retailer outlet identifier (i.e., Office). A successful match of an Item instance to a Product instance links receipts to nutrition information. A Product instance is typically linked to multiple instances of the NutritionInformation class, which contains three subclasses: Nutrient, NutriScore, and OfComDetail. The Nutrient class describes the nutrient content per 100 g of the product for a specific nutrient. Based on this information and following the FSA-NPS DI definition, the detailed product FSA scores are described in the class OfComDetail; the NutriScore class represents the Nutri-Score values of products and baskets. For a basket overview, NutriLink introduces the BasketAggInfo class. It provides aggregated information for all Item instances in a Basket instance, including total expense, discounts, and nutritional data (e.g., nutrient content, average FSA score, and Nutri-Score) when applicable.

Figure 1 shows how the classes Basket, Item, Product, Nutrient, qudt:Quantity, and BasketAggInfo are connected, using an example Item instance named Broccoli that is linked to GTIN 2124321000000. For more details about the ontology, refer to our GitHub repository.17

Figure 1.

Semantic description of an Item instance Broccoli using NutriLink. The Item instance matches a Product instance with gtin 2124321000000. The Product instance is connected to multiple Nutrient instances, for example, dietaryFiber. These linked nutritional information supports calculating BasketAggInfo instances for corresponding Basket instances. qudt:Quantity is used to represent quantities and quantity units.

3.2.2. NutriLink and its CQs

In the following, we discuss how the NutriLink ontology can be used to answer our CQs.

Accessing Quantities of Products in a Basket (CQ1)

NutriLink uses qudt:Quantity to represent quantities related to classes Product, Nutrient, and BasketAggInfo, and to permit consistent representation and handling of quantity values across the ontology. We described the quantities of Item instances using amount, because receipts do not provide quantity units. Answering CQ1, Listing 1 shows a SPARQL query retrieving the quantity of an item named “Broccoli” and its product quantity.

Accessing Aggregated Basket Information (CQ2)

To address CQ2, the classes Basket, Item, Product, Nutrient, and OfComDetail are needed to describe specific information from individual baskets. The required information includes item names, item quantities, item prices, matched product quantity values and units, matched product nutrient quantities and units per 100 g product, and matched product FSA score details. The basket Nutri-Score is derived from the basket FSA-NPS DI, that is, the energy-weighted average of the FSA scores of all products in the basket, as introduced in Section 2.1. We provide a SPARQL query retrieving these information for recent baskets on our GitHub repository.¹⁸

Accessing Nutri-Score Values from Shopping History (CQ3)

To address CQ3, the classes Basket, Item, Product, BasketAggInfo, and NutriScore are needed to describe basket timestamps, basket Nutri-Score values, item names, item quantity, and product Nutri-Score values. A SPARQL query to retrieve this information from the knowledge base can be found in our GitHub repository.¹⁹

Monthly Nutrition and Expense Analysis (CQ4)

To answer CQ4, energy content and prices of items that were purchased in a given month need to be retrieved for nutrition and expense analysis. The detailed SPARQL query can be found in our GitHub repository.²⁰

Access Dietary Recommendations (CQ5)

In the fully automated dietary counseling system FoodCoach (Wu et al., 2025), we generate structured dietary recommendations based on the last 10 baskets. We retrieve various information from the knowledge base, including item amount, product quantity values and units, corresponding food categories, product nutrient/energy content, and product FSA scores. The SPARQL queries that retrieve this information can be found in our repository.²¹ $^{,}$ ²² In the future, we plan to collaborate with our hospital partners to enable registered dietitians to give structured dietary recommendations based on loyalty card data of patients (or study participants). We will provide dietitians with nutrition information of baskets similar to what we retrieve for automated recommendations according to NutriLink, however in greater detail with respect to the provenance of macronutrients. To access structured dietary recommendations (e.g., Reduce sodium from cheese products) through NutriLink, Recommendation instances are linked to food categories (i.e., DietCoachMinorCategory instances, such as cheese products) and to nutrients (i.e., Nutrient instances, such as sodium). A recommendation is furthermore associated with a Boolean flag through the property increaseContent that signifies whether the user should increase (or decrease) the consumption of the target nutrient in products from the target category. Recommendations carry, as further properties, a recommendationPriority (integer), a human-readable explanation (string), and are associated with a User instance and a timestamp; furthermore, to permit tracking the origin of a recommendation, they are linked to a Dietitian instance.

4. Case Study: Fully Automated Dietary Counseling through NutriLink

The basis of the NutriLink ontology is FoodCoach (Wu et al., 2025), a fully automated dietary counseling system based on digital receipts from Migros and/or Coop loyalty cards in Switzerland. This system has been developed with our hospital partners and pilot-tested with 76 users, including 15 test users.

The digital receipt integration was facilitated by a third-party service that manages user consent and digital receipt retrieval from the retailers.²³ The transferred receipts contain information including article names, quantities, prices, discounts (if applicable), receipt timestamps, store names, and store location (latitude and longitude). Furthermore, we receive metadata including a user identifier and consent record. We removed the exact store names and locations, and reduced timestamp precision from milliseconds to days, further reducing potential user reidentification risks. The sanitized receipt data is then inserted—according to the NutriLink ontology—into a GraphDB²⁴ instance running on premises in our institute.

The study team has been maintaining an FCD compliant with the requirements outlined in Section 2.2. This database, which includes over 50,000 products from 126 categories (Fuchs et al., 2020), is based on Trustbox Switzerland (GS1 Switzerland) and has been supplemented with manually added products to expanded coverage. The FCD used in this study employed the pre-2023 algorithm for Nutri-Score calculation. Data quality is improved via automated validation (e.g., the sum of all nutrients should not exceed 100 g per 100 g product) and manual verification.

Receipt items are matched to FCD products using automated methods (e.g., regular expressions) and manual verification. Matching considers similarities between receipt item names and product names from retailer websites, supplemented by price differences. The resulting matching table contains item names and prices from receipts, GTINs from the products table, and manually assigned weight factors and confidence levels (high, medium, or low). Across previous datasets, approximately 60%–70% of receipt items are successfully matched. Low-confidence matches are cross-reviewed by at least two team members for a more accurate conclusion. The three confidence levels allow database users to balance matching coverage and data quality. Nonfood items (e.g., shopping bags) are flagged and excluded from exact matching, as they are irrelevant for nutritional analyses.

Item weights are estimated as follows: for weighed products (whose quantities can be noninteger e.g., name: Tomatoes, amount: 0.526), kilograms are used as the unit; for products sold by pieces, we first extract the quantity from receipts or product websites if available (e.g., Bread 500 g on receipts, or Chocolate cookies, 225 g on the retailer’s website). If the quantity is not specified (e.g., name: Pomelo, amount: 1), average weights are estimated based on retailer-specific data (e.g., 1,000 g for a Pomelo). Items with the same name but different prices (e.g., duo- or four-packs of yogurt) are accounted for using weight factors. Data quality is continuously improved, and basket-level aggregations are periodically updated.²⁵

If an item is successfully matched, the FCD returns the following product attributes: GTIN; product name (in several languages); product size and unit; energy content; nutrient content (total carbohydrates; sugars; total fats; saturated fats; proteins; fiber; salt; sodium); product FVLNO share; product image(s); product ingredients (string); allergens (NutriLink supports 14 different allergens); minor food category; major food category; Nutri-Score product category; Nutri-Score; and FSA score details. This information, which is then available via our GraphDB and structured according to the NutriLink ontology, forms the basis of FoodCoach.

Our system can be used by users whose data has successfully been integrated through the mobile-first Web application FoodCoach.²⁶ Following the Backend for Frontend pattern and similar to the method proposed in Mayer et al. (2017), this application—whose features correspond to the introduced CQs—communicates with the GraphDB indirectly: The application triggers the SPARQL queries that correspond to its currently relevant CQs through a backend and renders the results in a format that makes them easily accessible to users. The resulting views are shown in Figure 2:

Corresponding to CQ2, the view Nutri-Score of Baskets presents the Nutri-Score values of the seven most recent baskets of a user. The view additionally compares the energy-weighted Nutri-Score of these baskets of the user to the values achieved by all FoodCoach users over the last 4 weeks.

Corresponding to CQ3, the view Shopping History displays the user’s purchase history, product Nutri-Score values, and aggregated basket Nutri-Score values.

Corresponding to CQ4, the views Nutrition Analysis and Spending Analysis visualize calorie and expense contributions across seven food categories (protein foods; vegetables; fruits; processed food; grains, potatoes and legumes; beverages; and oils, fats, sauces, nuts and seeds).

Corresponding to CQ5, the view Automatic Recommendations provides dietary recommendations along with explanations. It also shows healthier alternative products in the same food category, based on the user’s last 10 baskets containing identified food products.

All shown views depend on answering CQ1 to retrieve information about absolute or relative product quantities.

Figure 2.

Visualization of data that conforms to the NutriLink ontology in the FoodCoach application. Each view corresponds to one competency question introduced in Section 3.1.

5. Limitations and Future Work

The NutriLink ontology has several limitations. Most importantly, the ontology was created mainly based on (food and nonfood) products on digital receipts from two major loyalty card systems in Switzerland. The applicability of our ontology has hence not been verified for products from paper receipts, farmers’ markets, restaurant meals, or home-made dishes. We argue that NutriLink should still be applicable to these products, since our CQs and queries depend on standard retail data (amount, price, weight, etc.) and nutrition information (energy in kJ, sugar in g, etc.). The main challenge in extending NutriLink to additional products is retrieving this information in digital formats. Potential solutions include a combination of digital tools and manual data correction. To digitalize paper receipts, optical character recognition combined with transformer-based models for layout understanding provides potential solutions (Abdallah et al., 2024; Ren et al., 2021). However, this manual scanning approach is burdensome and error-prone, due to potential duplicate or missing scans. For restaurant meals, obtaining accurate nutrition information is challenging despite EU Regulation No. 1169/2011 requiring its provision. One solution is to scrape nutritional data from restaurant websites where available, or estimate average values for local meals. Similarly, home-made meals could be approximated using local averages. For farmers’ market products, national FCDs, for example, the Swiss Food Composition Database,²⁷ could provide nutritional data, but product sizes and prices may need to be manually entered.

Second, the NutriLink ontology only models the nutrition information of raw products, neglecting practical factors that could heavily affect nutritional intake, such as food preparation and food waste. This limitation could be mitigated by integrating people’s eating/cooking behavior captured via questionnaires, at the cost of requiring more effort by the users.

Third, receipts themselves can occasionally be unreliable, for example, because of product returns. Although food product returns rates are in general low (e.g., even for online purchases, only 1%–3.9% of food products were returned in Switzerland in 2022; Zumstein et al., 2022), we can extend NutriLink to cover this aspect in the future.

Fourth, as NutriLink has been geared toward dietary analysis and the provisioning of dietary recommendations, it might overlook other relevant aspects of food products. For instance, NutriLink lacks sustainability characteristics of a food product. Environmental aspects like greenhouse gas emissions could be integrated into nl:Product using databases such as the World Food LCA database²⁸ (cf. Pilz et al., 2022), or retailer websites, when available. Similarly, social aspects like Fair Trade certification²⁹ could be incorporated when suitable standards are identified. Lastly, the adoption of NutriLink faces social and organizational barriers, including varying digital literacy among researchers and practitioners and complex institutional decision-making processes.

As next steps, we plan to expand NutriLink to include more food products, especially from beyond supermarkets, and to incorporate sustainability aspects as mentioned above. More research in the user experience and user interface aspects should also be conducted for broader adoption and applicability of NutriLink. Additionally, we will explore potential collaborations with international partners to assess NutriLink’s applicability in other regions and support related research, such as the work of the International Network for Food and Obesity/Non-communicable Diseases Research, Monitoring, and Action Support (INFORMAS; Swinburn et al., 2013).

6. Conclusions

We presented the NutriLink ontology that links digital receipts to food nutrition information and dietary recommendations. Compared to the existing food ontologies, NutriLink additionally covers fine-grained details of food products for Nutri-Score computation. The integration of NutriLink with FoodOn, GoodRelations, AGROVOC, and schema.org concepts significantly enhance its coverage, data interoperability, and multilingual support globally. We verified the utility of the ontology by implementing it in a fully automated dietary counseling system that has been used by 76 users. We propose that NutriLink can allow researchers to better integrate and analyze food-related data, advancing digital monitoring and intervention in nutrition and beyond. Additionally, it can contribute to the development of evidence-based policies and interventions to promote healthy food purchasing, hence helping combat NCDs.

Footnotes

Acknowledgments

We would like to thank Simeon Pilz and our partners at the University Hospital of Bern—Melanie Stoll and Lia Bally—for their continued collaboration and support. GPT-5.1 and Claude Sonnet 4.5 were used for text refinement.

ORCID iDs

Jing Wu

Kimberly Garcia

Simon Mayer

Jan Albert

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by the Swiss National Science Foundation under Project 188402.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

1.

—Article 8.6 mandates that basic personal information as well as personal health, physiological, educational, and occupational information need to be provided to third parties upon request of the data subject.

2.

3.

; accessed on February 27, 2025.

4.

The time span covered by these records depends on both the shopper (duration of loyalty program membership) and the retailer (retention period of shopping data).

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

This service has since been replaced by an internal solution, see .

24.

25.

We have since updated the FCD and integrated the most recent Nutri-Score algorithm. Data in this system is manually validated by a dietitian; it is available at

26.

27.

28.

29.

References

Abdallah

Eberharter

Pfister

Jatowt

(2024). Transformers and language models in form understanding: A comprehensive review of scanned document analysis. arXiv preprint arXiv:2403.04080.

Appelhans

B. M.

French

S. A.

Tangney

C. C.

Powell

L. M.

Wang

(2017). To what extent do food purchases reflect shoppers’ diet quality and nutrient intake? International Journal of Behavioral Nutrition and Physical Activity, 14(1), 1–10. https://doi.org/10.1186/s12966-017-0502-2

Bakland

Rosenvinge

J. H.

Wynn

Sørlie

Sundgot-Borgen

Fostervold Mathisen

Hanssen

T. A.

Jensen

Innjord

Pettersen

(2020). A new treatment for eating disorders combining physical exercise and dietary therapy (the PED-t): Experiences from patients who dropped out. International Journal of Qualitative Studies on Health and Well-being, 15(1), 1731994. https://doi.org/10.1080/17482631.2020.1731994

Blick (2021). So viel wissen Migros und Coop über uns. Retrieved on November 19, 2025, from https://www.blick.ch/wirtschaft/blick-macht-den-grossen-datentest-so-viel-wissen-migros-und-coop-ueber-uns-id8439476.html

Bommer

Sagalova

Heesemann

Manne-Goehler

Atun

Bärnighausen

Davies

Vollmer

(2018). Global economic burden of diabetes in adults: Projections from 2015 to 2030. Diabetes Care, 41(5), 963–970. https://doi.org/10.2337/dc17-1962

Boulos

M. N. K.

Yassine

Shirmohammadi

Namahoot

C. S.

Brückner

(2015). Towards an “internet of food”: Food ontologies for the internet of things. Future Internet, 7(4), 372–392. https://doi.org/10.3390/fi7040372

Caracciolo

Stellato

Rajbahndari

Morshed

Johannsen

Jaques

Keizer

(2012). Thesaurus maintenance, alignment and publication as linked data: The AGROVOC use case. International Journal of Metadata, Semantics and Ontologies, 7(1), 65–75. https://doi.org/10.1504/IJMSO.2012.048511

Çelik

(2015). FoodWiki: Ontology-driven mobile safe food consumption system. The Scientific World Journal, 2015, 475410. https://doi.org/10.1155/2015/475410

Chen

Gemming

Hanning

Allman-Farinelli

(2018). Smartphone apps and the nutrition care process: Current perspectives and future considerations. Patient Education and Counseling, 101(4), 750–757. https://doi.org/10.1016/j.pec.2017.11.011

10.

Coop (2025). Annual report 2024 coop group. Retrieved on November 19, 2025, from https://report.coop.ch/pdfs/Coop_GB_24_en_a463e1279e.pdf

11.

Dooley

D. M.

Griffiths

E. J.

Gosal

G. S.

Buttigieg

P. L.

Hoehndorf

Lange

M. C.

Schriml

L. M.

Brinkman

F. S.

Hsiao

W. W.

(2018). FoodOn: A harmonized food ontology to increase global food traceability, quality control and data integration. npj Science of Food, 2(1), 23. https://doi.org/10.1038/s41538-018-0032-6

12.

Erkkola

Kinnunen

S. M.

Vepsäläinen

H. R.

Meinilä

J. M.

Uusitalo

Konttinen

Saarijärvi

Fogelholm

Nevalainen

(2022). A slow road from meat dominance to more sustainable diets: An analysis of purchase preferences among Finnish loyalty-card holders. PLoS Sustainability and Transformation, 1(6), e0000015. https://doi.org/10.1371/journal.pstr.0000015

13.

European Union. Regulation (EU) No. 1169/2011 of the European Parliament and of the Council. Retrieved on December 2, 2025, from https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:02011R1169-20250401

14.

European Union (2016). Regulation (EU) 2016/679 of the European Parliament and of the Council. Official Journal of the European Union. Retrieved on December 2, 2025, from https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32016R0679

15.

Fuchs

K. L.

Haldimann

Zeltner

Ilic

(2020) Icon-based digital food allergen labels for complementation of text-based declaration. https://aisel.aisnet.org/ecis2020_rp/111

16.

Gemming

Mhurchu

C. N.

(2016). Dietary under-reporting: What foods and which meals are typically under-reported? European Journal of Clinical Nutrition, 70(5), 640–641. https://doi.org/10.1038/ejcn.2015.204

17.

Green

M. A.

Watson

A. W.

Brunstrom

J. M.

Corfe

B. M.

Johnstone

A. M.

Williams

E. A.

Stevenson

(2020). Comparing supermarket loyalty card data with traditional diet survey data for understanding how protein is purchased and consumed in older adults for the UK, 2014–16. Nutrition Journal, 19, 1–10. https://doi.org/10.1186/s12937-020-00602-3

18.

GS1 Switzerland. Trustbox: Product master data directly from the source. https://trustbox.gs1.ch/en

19.

Hepp

(2008). Goodrelations: An ontology for describing products and services offers on the web. In Knowledge engineering: Practice and patterns: 16th international conference, EKAW 2008, Acitrezza, Italy, September 29–October 2, 2008. Proceedings 16 (pp. 329–346). Springer. https://doi.org/10.1007/978-3-540-87696-0_29

20.

Hercberg

Touvier

Salas-Salvado

(2021). The Nutri-Score nutrition label. Hogrefe AG. https://doi.org/10.1024/0300-9831/a000722

21.

Jenneson

V. L.

Pontin

Greenwood

D. C.

Clarke

G. P.

Morris

M. A.

(2022). A systematic review of supermarket automated electronic sales data for population dietary surveillance. Nutrition Reviews, 80(6), 1711–1722. https://doi.org/10.1093/nutrit/nuab089

22.

Julia

Hercberg

(2017). Nutri-score: Evidence of the effectiveness of the French front-of-pack nutrition label. Ernahrungs Umschau, 64(12), 181–187. https://doi.org/10.4455/eu.2017.048

23.

Justice

B. E.

Nielsen

M. G.

Tyson

(2021). CPRA countdown: It’s time to brush up on California’s latest data privacy law. National Law Review, XI(352). https://natlawreview.com/article/cpra-countdown-it-s-time-to-brush-california-s-latest-data-privacy-law

24.

Kolchin

Zamula

(2013). Food product ontology: Initial implementation of a vocabulary for describing food products. In Proceeding of the 14th conference of open innovations association FRUCT, Helsinki, Finland (pp. 11–15). https://web.archive.org/web/20180412015813id_/https://www.fruct.org/publications/abstract14/files/Kol_21.pdf

25.

Lee

C. L.

Aveyard

P. N.

Jebb

S. A.

Piernas

(2021). Using supermarket loyalty card data to provide personalised advice to help reduce saturated fat intake among patients with hypercholesterolemia: A qualitative study of participants’ experiences. Nutrients, 13(4), 1146. https://doi.org/10.3390/nu13041146

26.

Mayer

Fuchs

Brügger

D. J.

Lian

Ciortea

(2021). Improving customer decisions in web-based e-commerce through guerrilla modding. Nature Machine Intelligence, 3, 1008–1010. https://doi.org/10.1038/s42256-021-00424-w

27.

Mayer

Hodges

Kritzler

Michahelles

(2017). An open semantic framework for the industrial internet of things. IEEE Intelligent Systems, 32(1), 96–101. https://doi.org/10.1109/MIS.2017.9

28.

Merz

Temme

Alexiou

Beulens

J. W. J.

Buyken

A. E.

Bohn

Ducrot

Falquet

M. N.

Solano

M. G.

Haidar

Infanger

Kühnelt

Rodríguez-Artalejo

Sarda

Steenbergen

Vandevijvere

Julia

(2024). Nutri-Score 2023 update. Nature Food, 5(2), 1–9. https://doi.org/10.1038/s43016-024-00920-3

29.

Migros (2025). How many people have signed up for cumulus? Retrieved November 19, 2025, from https://support.migros.ch/hc/en-us/articles/23004428522258-How-many-people-have-signed-up-for-Cumulus

30.

Mönninghoff

Fuchs

Albert

Mayer

(2022). The effect of a future-self avatar mobile health intervention (FutureMe) on physical activity and food purchases: Randomized controlled trial. Journal of Medical Internet Research, 24(7), e32487. https://doi.org/10.2196/32487

31.

Naska

Lagiou

(2017). Dietary assessment methods in epidemiological research: Current state of the art and future prospects [version 1; peer review: 3 approved]. F1000Research, 6(F1000 Faculty Rev), 926. https://doi.org/10.12688/f1000research.10703.1

32.

Odunitan-Wayas

F. A.

Okop

K. J.

Dover

R. V.

Alaba

O. A.

Micklesfield

L. K.

Puoane

Levitt

N. S.

Battersby

Meltzer

S. T.

Lambert

E. V.

(2021). Food purchasing behaviour of shoppers from different south African socio-economic communities: Results from grocery receipts, intercept surveys and in-supermarkets audits. Public Health Nutrition, 24(4), 665–676. https://doi.org/10.1017/S1368980020001275

33.

Open Food Facts. Open food facts ontology. https://wiki.openfoodfacts.org/Structured_Data#Food_Ontology

34.

Parker

H. W.

De Araujo

Thorndike

A. N.

Vadiveloo

M. K.

(2021). The utility of household grocery purchase quality index scores as an individual diet quality metric. British Journal of Nutrition, 126(6), 933–941. https://doi.org/10.1017/S0007114520004833

35.

Peroni

(2016). SAMOD: an agile methodology for the development of ontologies. In Proceedings of the 13th OWL: Experiences and directions workshop and 5th OWL reasoner evaluation workshop (OWLED-ORE 2016) (pp. 1–14). https://doi.org/10.6084/m9.figshare.3189769

36.

Pilz

Merian

Mayer

Fuchs

(2022). BetterPlanet: Sustainability feedback from digital receipts. In Advances in mobile computing and multimedia intelligence: 20th international conference, MoMM 2022, Virtual Event, November 28–30, 2022, Proceedings (pp. 102–107). Springer. https://doi.org/10.1007/978-3-031-20436-4_10

37.

QUDT.org. Quantities, units, dimensions and types (QUDT). Retrieved on December 2, 2025, from https://doi.org/10.25504/FAIRsharing.d3pqw7

38.

Ren

Zhou

Chen

Shao

Zhang

(2021). A transformer-based decoupled attention network for text recognition in shopping receipt images. In Neural computing for advanced applications: Second international conference, NCAA 2021, Guangzhou, China, August 27–30, 2021, Proceedings 2 (pp. 563–577). Springer. https://doi.org/10.1007/978-981-16-5188-5_40

39.

Sainz-De-Abajo

García-Alonso

J. M.

Berrocal-Olmeda

J. J.

Laso-Mangas

De La Torre-Díez

(2020). FoodScan: Food monitoring app by scanning the groceries receipts. IEEE Access, 8, 227915–227924. https://doi.org/10.1109/ACCESS.2020.3046031

40.

Schönenberger

K. A.

Cossu

Prendin

Cappon

Fuchs

K. L.

Mayer

Herzig

Facchinetti

Bally

(2022). Digital solutions to diagnose and manage postbariatric hypoglycemia. Frontiers in Nutrition, 9, 855223. https://doi.org/10.3389/fnut.2022.855223

41.

Snae

Bruckner

(2008). Foods: A food-oriented ontology-driven system. In 2008 2nd IEEE international conference on digital ecosystems and technologies (pp. 168–176). IEEE. https://doi.org/10.1109/DEST.2008.4635195

42.

Swan

W. I.

Vivanti

Hakel-Smith

N. A.

Hotson

Orrevall

Trostler

Howarter

K. B.

Papoutsakis

(2017). Nutrition care process and model update: Toward realizing people-centered care and outcomes management. Journal of the Academy of Nutrition and Dietetics Home, 117(12), 2003–2014. https://doi.org/10.1016/j.jand.2017.07.015

43.

Swinburn

Sacks

Vandevijvere

Kumanyika

Lobstein

Neal

Barquera

Friel

Hawkes

Kelly

L’Abbé

Lee

Macmullan

Mohan

Monteiro

Rayner

Sanders

Snowdon

Walker

(2013). INFORMAS (International Network for Food and Obesity/non-communicable diseases Research, Monitoring and Action Support): Overview and key principles. Obesity Reviews, 14, 1–12. https://doi.org/10.1111/obr.12087

44.

Swiss Confederation (2022). Bundesrat veröffentlicht Bericht über die Wirksamkeit des Nutri-Score. Retrieved on March 5, 2026, from https://www.news.admin.ch/de/nsb?id=92042.

45.

swissinfo.ch (2025). Coop overtakes Migros in battle of Swiss food retailers. Retrieved on November 19, 2025, from https://www.swissinfo.ch/eng/agribusiness/coop-overtakes-migros-in-core-business/90267923.

46.

World Health Organization (2024a). Diabetes. Retrieved on March 5, 2026, from https://www.who.int/news-room/fact-sheets/detail/diabetes

47.

World Health Organization (2024b). Noncommunicable diseases. https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases

48.

Fuchs

Lian

Haldimann

M. L.

Schneider

Mayer

Byun

Gassmann

Brombach

Fleisch

(2022). Estimating dietary intake from grocery shopping data—A comparative validation of relevant indicators in Switzerland. Nutrients, 14(1), 159. https://doi.org/10.3390/nu14010159

49.

Mayer

Pilz

Antille

Y. S.

Albert

J. L.

Stoll

Garcia

Fuchs

Bally

Eichelberger

Schneider

Tiefenbeck

Merian

Orban

(2025). FoodCoach: Fully automated diet counseling. IEEE Journal of Biomedical and Health Informatics, 29(7), 5257–5269. https://doi.org/10.1109/JBHI.2025.3540899

50.

Zumstein

Oswald

Brauer

(2022). Online retailer survey 2022. Institute of Marketing Management: Winterthur, Switzerland. Retrieved on December 2, 2025, from https://handelsverband.swiss/wp-content/uploads/2022/10/221013_Zumstein-et-al_OnlineRetailerSurvey_2022.pdf

NutriLink: An Ontology for Linking Digital Receipts to Food Nutrition Information and Dietary Recommendations

Abstract

Keywords

1. Introduction

2. Related Work

2.1. Nutri-Score and the British Food Standards Agency Nutrient Profiling System Dietary Index

2.2. Provenance of Food Purchase and Food Composition Data

2.3. Food and Grocery Shopping Ontologies

3. The NutriLink Ontology

3.1. Motivating Scenario and Competency Questions

3.2. Semantic Model

Accessing Quantities of Products in a Basket (CQ1)

Accessing Aggregated Basket Information (CQ2)

Accessing Nutri-Score Values from Shopping History (CQ3)

Monthly Nutrition and Expense Analysis (CQ4)

Access Dietary Recommendations (CQ5)

4. Case Study: Fully Automated Dietary Counseling through NutriLink

6. Conclusions

Footnotes

Acknowledgments

ORCID iDs

Funding

Declaration of Conflicting Interests

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

References