https://orcid.org/0000-0002-9853-7185
Figures
Abstract
Front-of-package (FOP) is one of the most direct communication channels connecting manufacturers and consumers, as it displays crucial information such as certification, nutrition, and health. Traditional methods for obtaining information from FOPs often involved manual collection and analysis. To overcome these labor-intensive characteristics, new methods using two artificial intelligence (AI) approaches were applied for information monitoring of FOPs. In order to provide practical implementations, a case study was conducted on infant food products. First, FOP images were collected from Amazon.com. Then, from the FOP images, 1) the certification usage status of the infant food group was obtained by recognizing the certification marks using object detection. Moreover, 2) the nutrition and health-related texts written on the images were automatically extracted based on optical character recognition (OCR), and the associations between health-related texts were identified by network analysis. The model attained a 94.9% accuracy in identifying certification marks, unveiling prevalent certifications like Kosher. Frequency and network analysis revealed common nutrients and health associations, providing valuable insights into consumer perception. These methods enable fast and efficient monitoring capabilities, which can significantly benefit various food industries. Moreover, the AI-based approaches used in the study are believed to offer insights for related industries regarding the swift transformations in product information status.
Citation: Kim D, Kim S-Y, Yoo R, Choo J, Yang H (2024) Innovative AI methods for monitoring front-of-package information: A case study on infant foods. PLoS ONE 19(5): e0303083. https://doi.org/10.1371/journal.pone.0303083
Editor: Muhammad Khalid Bashir, University of Agriculture Faisalabad, PAKISTAN
Received: July 11, 2023; Accepted: April 19, 2024; Published: May 16, 2024
Copyright: © 2024 Kim et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: In compliance with the limitation that the data we used is available for research purposes only, it cannot be directly shared as third-party data. However, we have sufficiently described the data source and the procedures for data collection and processing in the Materials and Methods section. Furthermore, to enhance reader convenience, we have provided additional detailed URLs and access procedures in a 'Supporting Information'. Any additional inquiries about the dataset for research purposes can be directed to the corresponding author Professor Hee Yang at Kookmin University, Republic of Korea (yhee6106@kookmin.ac.kr).
Funding: This research was supported by a grant (21162MFDS076) from the Ministry of Food and Drug Safety in 2023 and the Korea Institute of Planning and Evaluation for Technology in Food, Agriculture and Forestry(IPET) through the Export Promotion Technology Development Program, funded by the Ministry of Agriculture, Food and Rural Affairs(MAFRA)(617078-06). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [Funder 1] Full name of the funder 1: Ministry of Food and Drug Safety Grant number: 21162MFDS076 Recipient initial: JG Choo Funder website: www.mfds.go.kr/eng/index.do [Funder 2] Full name of the funder 2: Korea Institute of Planning and Evaluation for Technology in Food, Agriculture and Forestry (IPET), Ministry of Agriculture, Food and Rural Affairs(MAFRA) Grant number: 617078-06 Recipient initial: SY Kim Funder website: https://www.mafra.go.kr/english/index.do.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Front-of-Package (FOP) information, which refers to the content prominently displayed on the front side of a product’s package [1], holds paramount importance by enabling consumers to easily understand products, whether in online listings or on physical store shelves [2, 3]. Based on standardized regulations, especially in the food market, FOP information becomes more standardized and refined to prevent consumers from being confused or misleading [4]. For example, the U.S. government’s Nutrition Labeling and Education Act of 1990 mandated food labeling rules, granting the Food and Drug Administration (FDA) authority to require nutrition labeling and nutrient content claims [5]. In 2013, the FDA introduced "gluten-free" certification standards [6, 7], and voluntary standards, such as organic guidelines and the Non-GMO Project Standard, also emerged.
Since product information, including FOP information, serves as a significant influencing factor in purchase decisions [8–10] and is also a crucial tool for companies to communicate their product’s value to consumers [11–13], several earlier studies sought to investigate the written information on FOPs for monitoring markets by analyzing the statement across different food categories or the alignment of statements with legal requirements, as well as scientific evidence [14, 15]. Recently, with the increasing emphasis on product quality and functional benefits, certification, nutrition, and health-related information on FOPs has also been highlighted [16, 17]. However, the manual analysis of collected data required a considerable investment of time and effort from the researchers [18], which imposed challenges in expanding the size of the data pool.
With advancements in artificial intelligence (AI)-based image data processing [18], some studies with food products have applied AI techniques for various purposes. Object detection, a part of computer vision (CV) [19], detects objects within an image and assigns labels to them, indicating what they are [20]. This technique predicts consumers’ food choices from vending machines [19] or calculates calories from food images [20]. Optical character recognition (OCR) is another type of CV technology that locates and recognizes characters from an image and converts them into machine-encoded texts [21]. This technique extracts detailed information about products, such as expiration dates [22, 23] and a list of halal ingredients for verifying accuracy [24]. However, for the purpose of monitoring the status of food products and detecting market changes, there are still few studies incorporating AI, such as object detection and OCR, to efficiently extract and analyze product information from FOP images, which may overcome the aforementioned limitations of manual methods.
Infant food products, which are one of the food categories with heightened consumer involvement, were chosen for the case study. FOPs of infant foods convey crucial nutrition and health-related information, significantly influencing purchasing decisions [25–27]. Furthermore, especially among parents who actively engage in social activities, the demand for more comprehensive and detailed information is increasing [17]. Previous studies manually analyzed infant food information, encountering limitations that arose due to the time and effort required for manual analysis [28, 29].
Here, we propose a new AI-based monitoring method for FOP information on food items in this study. This study aims to efficiently analyze the image and text information using AI techniques to comprehend the information provided on FOPs of infant food products. Applied with AI techniques of object detection and OCR, the new monitoring method has the advantage of expanding the analyzing data size and efficiently extracting different types of certification marks and various statements, including nutrition and health-related texts, from FOP images. Subsequently, we applied frequency and network analysis, as it can provide notable interpretations by quantifying and visualizing the relations between texts [30, 31]. The output from our new method is verified by analyzing the results of infant foods obtained by manual analysis in other previous studies.
Materials and methods
FOP data collection
FOP images of infant food products were collected from Amazon’s website (www.amazon.com). Products within the five categories of infant foods, including meals, formula, snack foods, beverages, and cereal & porridge, were collected in February 2021 using the Python web scraping library ‘Selenium’. For each product, the image, name, category, and Amazon standard identification number were crawled. During the data collection process, advertising data unrelated to infant foods were removed, as were duplicate data and product data with image quality below 50KB. The number of product image data after pre-processing was 1,176.
Development of a certification mark detection model
An object detection model was initially developed to recognize several certification marks on FOP images using Google’s AutoML Vision to determine the certification status of infant food products. This model was selected for its ability to quickly achieve optimal performance, even with limited training data [32].
Certification marks indicate symbols highlighting a product’s characteristics that a manufacturer has verified compliance with specific standards or regulations, and the model was trained to identify four types of certification marks on FOP images: USDA Organic, Non-GMO, Kosher, and Gluten-Free. The specific criteria for each certification are presented in Table 1. USDA Organic is an organic certification mark supervised by the U.S. Department of Agriculture (USDA) and awarded to products with an organic content of over 95% [33]. Non-GMO Project Verified is a certification mark widely used in the United States for food certified by the non-profit organization, the Non-GMO Project. The certification is given to products that contain less than 0.9% genetically modified organisms (GMOs), adhering to the strictest EU standards [34]. The U-shaped Kosher certification mark, maintained by the Orthodox Union, is the most common Kosher certification mark in the United States [35]. Gluten-Free, maintained by the Gluten-Free Certification Organization (GFCO), is also widely used worldwide. GFCO only grants Gluten-Free certification if the gluten content of a food is less than 10 ppm [36].
Fig 1 illustrates the process of training and inferencing the object detection model. The training dataset consisted of images gathered from various categories on Amazon, excluding the infant foods category. In this study, four classes were trained: USDA Organic, Non-GMO, Kosher, and Gluten-Free. Since each certification mark has a unique shape and color, as shown in Fig 1, a relatively small number of training examples were required for each class. 129 images were trained for USDA Organic, 128 for Non-GMO, 154 for Kosher, and 100 for Gluten-Free. Some of the trained images contained multiple certification marks and were thus used to train more than one class, resulting in a total of 328 FOP images used as the training dataset for the object detection model.
The overall process can be divided into two distinct stages: model training and inference. During the training stage, the training dataset is collected, pre-processed, labeled with the certification marks, and used to train the model. During the inference stage, pre-processed infant food products’ images are used as the test dataset to grasp the certification status.
After uploading the entire training dataset to Google Cloud, each training image data was labeled with a bounding box indicating the location of a certification mark. The labeled data were then randomly split into three sets: 80% for training, 10% for validation, and 10% for testing.
Data extraction and analysis
The trained object detection model was applied to the pre-processed infant food products’ FOP images. The confidence threshold—i.e., the reference probability required for the model to return a predicted value—was set at 0.5. By applying this threshold to the model’s predictions, the certification usage status of each product was determined.
To determine the frequency of nutrition and health-related texts -the list of contained macronutrients, micronutrients, and their expected effects in terms of physical and mental health- as well as to find the relations within health-related texts of infant foods, all texts were initially extracted from the FOP image dataset pre-processed in the aforementioned FOP image data collection and pre-processing stage using Naver’s OCR application programming interface (API). This API was selected for its superior accuracy level in extracting texts [37]. Fig 2 shows the process of extracting and analyzing nutrition and health-related texts.
From pre-processed infant food product images, the texts are extracted using OCR, then processed and filtered using nutrition and health-related databases. Following this, the nutrition and health-related texts are used for frequency analysis, and the association of co-occurring health-related texts is further identified by network analysis.
To confirm the accuracy of the extracted texts, the Python text pre-processing library ‘Natural Language Toolkit’ (NLTK) was used to ascertain whether the extracted words were present in the standard dictionary. Irrelevant stopwords, such as syntactic words (e.g., is, the), as well as misextracted words, were removed. Furthermore, lemmatization was performed to transform inflected words to their base forms [38]. The USDA National Nutrient Database for Standard Reference [39] and Dietary Reference Intake (DRI) [40] were used to recognize nutrition and health-related texts. Moreover, Medical Subject Headings (MeSH) terms were utilized to identify health-related texts. The extracted texts were cross-referenced with texts in the database and terminology list, and only the matching texts were included in the analysis.
Nutrition and health-related texts were classified into separate categories. Nutrition-related texts were grouped as “Macronutrients” (carbohydrate, protein, fat), “Vitamins” (e.g., vitamin C, vitamin D, biotin), “Minerals” (e.g., iron, zinc, calcium), “Lipids” (e.g., DHA, EPA, cholesterol, choline), “Disaccharides & Polysaccharides” (e.g., sugar, fiber) and other nutrients (e.g., water, lutein, caffeine). Although Lipids, Disaccharides & Polysaccharides are generally recognized as fats and carbohydrates, the two categories were not identified as part of the “Macronutrients” category in this work. This classification was based on the specific portrayal of these nutrients on product labels, as texts including DHA and sugar on the FOPs often indicated alternative interpretations. For the health-related texts, the categories were developed based on the International Classification of Functioning, Disability, and Health (ICF) [41]. The categorization was organized as follows: the categories encompassed “Brain Development” (e.g., learn, brain), “General Growth” (e.g., grow), “Immunity” (e.g., immune, allergy), “Digestion” (e.g., digest, fussiness, diarrhea), “Eye Development” (e.g., eye, vision), “Bone & Tooth Development” (e.g., bone, tooth), “Malnutrition” (e.g., malnutrition), and “Calming” (e.g., sleep, anger, stress).
In the first step of the analysis process, the frequency of texts belonging to each nutrition or health category was examined. Then, network analysis, which identifies meaningful patterns within complex data by analyzing the properties and structures of the network, which is a graph consisting of nodes and edges representing co-occurrences between nodes, was performed between the eight health-related categories using the ‘networkx’ library based on the co-occurrence of health-related texts.
Results
Certification mark usage status of infant food products
After training the certification mark detection model, an overall performance of 94.9% was obtained. Performance refers to a score that indicates the confidence level of the model for each class provided by Google. Table 2 exhibits the performances of USDA Organic, Non-GMO, Kosher, and Gluten-Free; 97%, 96.8%, 75.9%, and 97.3%, respectively (Table 2).
Table 3 presents the certification usage status of infant food products across five categories: meals, formula, snack foods, beverages, and cereal & porridge. The proportions of organic certifications in the meals category were higher than in other categories, with USDA Organic and Non-GMO accounting for 72.4% and 61.3% of products, respectively. A significant proportion of products had both USDA Organic and Non-GMO certifications. The snack foods category also showed a high proportion of USDA Organic certification, accounting for 52.3% of products. By contrast, within the beverages, cereals & porridge, and formula categories, the proportion of USDA Organic certification was around 25%. The ratio of Kosher certification varied across all categories, ranging from 49.3% in the meals category to 92.8% in the formula category. Interestingly, Non-GMO certification was not used at all in the formula category, while Gluten-Free certification was rare across all the infant product categories.
Nutrition and health-related information of the infant food products
The findings of frequency analysis conducted on nutrition-related texts (Table 4) presented “Minerals” (33.3%) as the largest category, with iron being the most prevalent text (294), followed by mineral (92), and zinc (72). Regarding “Macronutrients” (18.1%), the analysis revealed that protein was the most frequently observed macronutrient (238), followed by fat (44) and carbohydrate (8). Among “Vitamins” (14.6%), while the term vitamin was highly frequent (215), specific types of vitamins, such as vitamin C, D, and E, were scarce. Within the “Lipids” (13.3%) category, the frequency was high in the order of DHA (173), choline (32), EPA (5), and cholesterol (3). The “Disaccharide & Polysaccharide” (12.1%) category consisted of two nutrients: fiber (103) and sugar (91). Lastly, the category of “Other Nutrients” (8.7%) was identified with lutein (44) and caffeine (3).
In terms of health-related texts (Table 5), the analysis indicated that the texts related to “Brain Development” were most frequent with a total of 328 (30.9%), followed by the texts associated with “General Growth” (323, 30.4%), “Immunity” (157, 14.8%), and “Digestion” (146, 13.7%). The texts related to “Eye Development” (73, 6.9%) occurred with relatively lower frequency, and those related to “Bone & tooth development” (23, 2.2%), “Malnutrition” (10, 0.9%), and “Calming” (3, 0.3%) were also observed.
Fig 3 describes the relationship between health-related texts on infant food products identified by their co-occurrence frequencies. Strong correlations were found in the occurrences of specific health-related texts such as “Immunity”, “General Growth”, and “Brain Development”, as indicated by the thickness of the lines in the network. Also, the occurrence of “Eye Development” and “Digestion” correlated with those of other health-related texts was observed, including “Brain Development” and “General Growth”. However, the “Calming” category consisted of the most uncommon texts and had no connections with other categories.
Discussion
This study introduces innovative AI techniques for efficient monitoring of FOP information, overcoming limitations associated with traditional manual collection and analysis methods. By conducting a case study on infant food products, the research utilizes object detection to identify certification usage. Sequentially, optical character recognition (OCR) is applied to extract nutrition and health-related texts, while frequency or network analysis provides the frequencies or associations among this information. These approaches for identifying product information statuses are expected to offer related industries with insights into the rapidly transforming market.
The utilization of two methods—one for analyzing certification marks and another for nutrition or health-related texts—applied to assess infant food products on the American Amazon platform revealed a trend consistent with previously studied outcomes through conventional research methods. This alignment underscores the validation of both models’ efficacy and sufficient accuracy in this analysis. As presented in Table 3, the object detection model provided a comprehensive certification usage status, demonstrated with a case study of infant food products. The study found that at least half of the analyzed infant foods were certified as organic, either through USDA Organic or Non-GMO, which was consistent with the results of Cheikhrouhou et al.’s study [42]. Notably, it was also confirmed that a substantial number of products in the meals category had both USDA Organic and Non-GMO certifications, supporting the findings of Castellari et al. [34]. Also, we confirmed the efficacy of OCR-based text mining in the analysis of the information on FOPs. By conducting a frequency analysis of nutrition and health-related texts (Tables 4 and 5), the most commonly occurring phrases were identified. The findings indicated that the texts iron (in “Minerals”), protein (in “Macronutrients”), vitamin (in “Vitamins”), and DHA (in “Lipids”) were frequently mentioned in the FOPs of infant and toddler food products, which aligned with previous studies emphasizing the importance of these nutrients for early childhood development [43–45]. Especially, the observation of iron and protein as the most frequently occurring nutrients in the FOPs of infant foods (Table 4) supported the benefits of these nutrients for the development of infants and toddlers [46–48], as well as suggested supplementation of iron and protein for breast milk [49]. The study also exhibited consistent results with previous work by Berry & Gribble [50], which analyzed health information from online advertisements of infant formula, and the main claims included supporting brain and eye development, bone growth, and immune system, as well as aiding digestion.
Intriguingly, besides the statement based on legal or scientific evidence, the FOP information also partially reflects general consumers’ perceptions. The presence of the Kosher certification mark was found to be widespread in all categories of infant food products. This finding supports previous research conducted by Della Corte et al. [51], emphasizing North America’s substantial contribution to the global Kosher product market. Cohen et al. [52] and Lytton [53] suggested that Kosher certification serves as an indicator of product safety and quality to consumers—thereby impacting their buying choices—regardless of their religious convictions. The data analysis conducted in this study further strengthens the notion that Kosher certification operates as a widely recognized promotional indicator for infant food products compared to other certifications. Moreover, the co-occurrence network analysis of each category type in Fig 3 revealed significant associations among the health-related texts. Notably, “Brain Development” frequently appeared alongside “Immune” or “General Growth” categories. This association could be linked to parental perceptions associating these claims with healthier infant foods [54, 55]. Addressing parental concerns, FOPs often contain information catering to these expectations. Belamarich et al. [56] noted a prevalent advertisement trend regarding digestion-related claims in US infant formula products, consistent with this study’s findings. Likewise, our method, the visualization of the co-occurrence network between text information of FOPs, could offer intuitive comprehension, as a starting point for new research endeavors. This underscores the need for further investigation to understand manufacturers’ practices on FOPs better, particularly considering consumers’ perceptions in the development of infant foods.
Our proposed approach reduces time and manual efforts over traditional monitoring methods across the entire process, from data collection and pre-processing to analysis. Previously, it took several weeks to months to collect FOPs by purchasing or photographing a substantial number of food products, ranging from over 300 [57] to as many as 3,000 [58], in offline stores. On the other hand, we crawled 1,176 infant products on Amazon within a few hours. Also, in contrast to previous studies, which restricted the scope of collected products due to manpower limitations, this study thoroughly examined all products within the infant product category. In the pre-processing phase after data collection, previous studies manually filtered out duplicate products by capturing images of shopping carts, providing a quick overview of the collected products. In contrast, we automatically removed them using the products’ unique Amazon standard identification numbers. Additionally, instead of manual data entry for information on FOPs, the implementation of OCR enhanced efficiency, reducing human typing errors. It allowed for rapid processing of vast data volumes, optimizing resources for comprehensive market understanding. Furthermore, the certification status of infant foods was analyzed easily using an object detection model. The development of this model, based on Google’s Auto ML, required a small amount of training dataset, consisting of 328 FOP images. Although it involved labeling a total of 511 instances of four certification marks, taking several hours initially, once the model was developed, it became available for continuous use, even when dealing with large volumes of data. Therefore, it enabled quick model training and application, requiring minimal manual effort.
However, there remains careful consideration for data collection when we utilize such an AI-based monitoring method as our model because data availability may limit the method’s scope; focusing on high-market-share brands could affect the result of the analysis. Nevertheless, the utilization of AI-based monitoring methods will become more feasible in the future. The evolving consumer perception of food quality is expanding to encompass functional, social, cultural, and ethical considerations, resulting in diverse certification marks [34]. Additionally, FOPs increasingly incorporate various information such as additives and allergens, even introducing the concept of ‘advice’ through color in Nutrient Facts, which are previously listed on the back of the package as a text, demonstrating a broader range of details. This may help food companies monitor what information competitors convey through FOPs especially when exporting food. Additionally, although governmental efforts standardized mandatory and voluntary labeling, the online shopping era introduced varied information beyond FOPs, revealing issues like insufficiently presenting essential data or exaggerating health claims restricted on FOPs [28, 59]. Regulatory authorities can leverage these methods to acquire extensive fundamental data on a large scale, serving as a crucial foundation for the development of well-informed policies and regulations. Consequently, it is suggested that our AI-based monitoring methods developed in this study can be a basis model to facilitate further research in these directions.
Conclusions
This research introduced innovative AI-based methods for product investigations, employing object detection and OCR techniques, and exemplified through a case study on infant food products. The object detection model achieved an accuracy performance rate of 94.9% in identifying certification marks, revealing prevalent certifications like Kosher across various infant food categories. Furthermore, the frequency and network analysis of nutrition and health-related texts shed light on common nutrients such as iron and protein, and health associations among brain development, immune, and general growth found on FOP labels in infant food products. These findings highlighted the benefits of employing these AI techniques to alleviate the labor-intensive challenges in prior studies. Additionally, given their capacity for comprehensive market analysis, these techniques suggest promise for further research, encompassing the monitoring of various types of information across wide range of product categories.
Supporting information
S1 File. Detailed listings of data sources and specific AI platforms with corresponding URLs for our method.
https://doi.org/10.1371/journal.pone.0303083.s001
(DOCX)
References
- 1. Oswald C, Adhikari K, Mohan A. Effect of front-of-package labels on consumer product evaluation and preferences. Curr Res food Sci. 2022;5: 131–140. pmid:35036933
- 2. Hersey JC, Wohlgenant KC, Arsenault JE, Kosa KM, Muth MK. Effects of front-of-package and shelf nutrition labeling systems on consumers. Nutr Rev. 2013;71: 1–14. pmid:23282247
- 3. Temple NJ. Front-of-package food labels: A narrative review. Appetite. 2020;144: 104485. pmid:31605724
- 4. Kuchler F, Greene C, Bowman M, Marshall KK, Bovay J, Lynch L. Beyond Nutrition and Organic Labels-30 Years of Experience With Intervening in Food Labels. 2017 [cited 10 Dec 2023]. Available: www.ers.usda.gov
- 5. Greenberg EF. The Changing Food Label: The Nutrition Labeling and Education Act of 1990. Loyola Consum Law Report. 1990;3. Available: https://heinonline.org/HOL/Page?handle=hein.journals/lyclr3&id=10&div=6&collection=journals
- 6.
Gluten-Free Labeling of Foods | FDA. [cited 10 Dec 2023]. Available: https://www.fda.gov/food/food-labeling-nutrition/gluten-free-labeling-foods
- 7.
Federal Register:: Food Labeling; Gluten-Free Labeling of Foods. [cited 10 Dec 2023]. Available: https://www.federalregister.gov/documents/2013/08/05/2013-18813/food-labeling-gluten-free-labeling-of-foods
- 8. Magnier L, Schoormans J. Consumer reactions to sustainable packaging: The interplay of visual appearance, verbal claim and environmental coEncern. J Environ Psychol. 2015;44: 53–62.
- 9.
Silayoi P. The importance of packaging attributes: a conjoint analysis approach Mark Speece. [cited 18 Dec 2023]. https://doi.org/10.1108/03090560710821279
- 10. Wansink B, Sonka ST, Hasler CM. Front-label health claims: when less is more. Food Policy. 2004;29: 659–667.
- 11. Hawkes C. Government and voluntary policies on nutrition labelling: a global overview. Innov Food Label. 2010; 37–58.
- 12. Küster I, Vila N, Sarabia F. Food packaging cues as vehicles of healthy information: Visions of millennials (early adults and adolescents). Food Res Int. 2019;119: 170–176. pmid:30884646
- 13. Vilnai-Yavetz I, Koren R. Cutting through the clutter: purchase intentions as a function of packaging instrumentality, aesthetics, and symbolism. Int Rev Retail Distrib Consum Res. 2013;23: 394–417.
- 14. Miklavec K, Pravst I, Raats MM, Pohar J. Front of package symbols as a tool to promote healthier food choices in Slovenia: Accompanying explanatory claim can considerably influence the consumer’s preferences. Food Res Int. 2016;90: 235–243. pmid:29195876
- 15. Pinto VRA, Freitas TB de O, Dantas MI de S, Della Lucia SM, Melo LF, Minim VPR, et al. Influence of package and health-related claims on perception and sensory acceptability of snack bars. Food Res Int. 2017;101: 103–113. pmid:28941673
- 16. Sagaceta-Mejía J, Tolentino-Mayo L, Cruz-Casarrubias C, Nieto C, Barquera S. Understanding of front of package nutrition labels: Guideline daily amount and warning labels in Mexicans with non-communicable diseases. PLoS One. 2022;17: e0269892. pmid:35749373
- 17. Asioli D, Aschemann-Witzel J, Caputo V, Vecchio R, Annunziata A, Næs T, et al. Making sense of the “clean label” trends: A review of consumer food choice behavior and discussion of industry implications. Food Res Int. 2017;99: 58–71. pmid:28784520
- 18. Zhu L, Spachos P, Pensini E, Plataniotis KN. Deep learning and machine vision for food processing: A survey. Curr Res Food Sci. 2021;4: 233–249. pmid:33937871
- 19. Kakani V, Nguyen VH, Kumar BP, Kim H, Pasupuleti VR. A critical review on computer vision and artificial intelligence in food industry. J Agric Food Res. 2020;2: 100033.
- 20. Wang D, Wang JG, Xu K. Deep Learning for Object Detection, Classification and Tracking in Industry Applications. Sensors 2021, Vol 21, Page 7349. 2021;21: 7349. pmid:34770656
- 21. Mittal R, Garg A. Text extraction using OCR: A Systematic Review. Proc 2nd Int Conf Inven Res Comput Appl ICIRCA 2020. 2020; 357–362.
- 22. Gundimeda V, Murali RS, Joseph R, Naresh Babu NT. An automated computer vision system for extraction of retail food product metadata. Adv Intell Syst Comput. 2019;815: 199–216.
- 23. Hosozawa K, Wijaya RH, Linh TD, Seya H, Arai M, Maekawa T, et al. Recognition of Expiration Dates Written on Food Packages with Open Source OCR. Int J Comput Theory Eng. 2018;10: 170–174.
- 24. Kartiwi M, Gunawan TS, Anwar A, Siti , Fathurohmah S. Mobile Application for Halal Food Ingredients Identification using Optical Character Recognition. Meas Appl. 2018.
- 25. Malek L, Fowler H, Duffy G, Katzer L. Informed choice or guessing game? Understanding caregivers’ perceptions and use of infant formula labelling. 2018 [cited 18 Dec 2023]. pmid:30477595
- 26. Ferrante MJ, Moding KJ, Bellows LL, Bakke AJ, Hayes JE, Johnson SL. Examining Front-of-Package Product Names and Ingredient Lists of Infant and Toddler Food Containing Vegetables. J Nutr Educ Behav. 2021;53: 96–102. pmid:33573771
- 27. Munsell CR, Harris JL, Sarda V, Schwartz MB. Parents’ beliefs about the healthfulness of sugary drink options: opportunities to address misperceptions. Public Health Nutr. 2016;19: 46–54. pmid:25757372
- 28. Koo Y-C, Chang J-S, Chen YC. Food claims and nutrition facts of commercial infant foods. PLoS One. 2018;13: e0191982. pmid:29489848
- 29. Pomeranz JL, Romo Palafox MJ, Harris JL. Toddler drinks, formulas, and milks: Labeling practices and policy implications. Prev Med (Baltim). 2018;109: 11–16. pmid:29339115
- 30. Makinei L V., Hazarika MK. Flavour network-based analysis of food pairing: Application to the recipes of the sub-cuisines from Northeast India. Curr Res food Sci. 2022;5: 1038–1046. pmid:35789802
- 31. Matsuoka H, Uchiyama Y, Woraitthinan K, Kohsaka R. Does novel food differ in cultural contexts? A comparative analysis of Japanese and Singaporean cultural acceptance through text analysis of mass media. Curr Res Food Sci. 2023;6: 100436. pmid:37377498
- 32. Zhang J, Zheng Y, Qi D, Li R, Yi X, Li T. Predicting citywide crowd flows using deep spatio-temporal residual networks. Artif Intell. 2018;259: 147–166.
- 33. Guilabert M, Wood JA. USDA Certification of Food as Organic: An Investigation of Consumer Beliefs about the Health Benefits of Organic Food. J Food Prod Mark. 2012;18: 353–368.
- 34. Castellari E, Soregaroli C, Venus TJ, Wesseler J. Food processor and retailer non-GMO standards in the US and EU and the driving role of regulations. Food Policy. 2018;78: 26–37.
- 35. Havinga T. REGULATING HALAL AND KOSHER FOODS: DIFFERENT ARRANGEMENTS BETWEEN STATE, INDUSTRY AND RELIGIOUS ACTORS. www.erasmuslawreview.nl Erasmus Law Rev. 2010;3.
- 36. Thompson T, Simpson S. A comparison of gluten levels in labeled gluten-free and certified gluten-free foods sold in the United States. Eur J Clin Nutr. 2015;69: 143–146. pmid:25271013
- 37. Eurviriyanukul K, Phiewluang K, Yawichai S, Chaichana S. Evaluation of Recognition of Water-meter Digits with Application Programs, APIs, and Machine Learning Algorithms. 2020 8th Int Electr Eng Congr iEECON 2020. 2020.
- 38.
<Anjali M, Jivani G. A Comparative Study of Stemming Algorithms. [cited 18 Dec 2023]. Available: www.ijcta.com
- 39.
USDA Department of Agriculture. USDA national nutrient database for standard reference. Available: https://data.nal.usda.gov/dataset/usda-national-nutrient-database-standard-reference-legacy-release
- 40. Medicine I of. Dietary Reference Intakes: The Essential Guide to Nutrient Requirements. 2006 [cited 18 Dec 2023].
- 41.
International Classification of Functioning, Disability and Health (ICF). [cited 18 Dec 2023]. Available: https://www.who.int/classifications/international-classification-of-functioning-disability-and-health
- 42.
Cheikhrouhou S, Bélisle D, Legendre S. Tell me more! An analysis of certifications and claims display practices on organic food front-of-packages. https://doi.org/10.1362/204440820X16003501529202
- 43. Innis SM. Human milk: maternal dietary lipids and infant development. Proc Nutr Soc. 2007;66: 397–404. pmid:17637092
- 44. Lönnerdal B. Infant formula and infant nutrition: bioactive proteins of human milk and implications for composition of infant formulas. Am J Clin Nutr. 2014;99. pmid:24452231
- 45.
New Zealand Ministry of Health. Healthy Eating Guidelines for New Zealand Babies and Toddlers (0–2 years old). [cited 18 Dec 2023]. Available: www.health.govt.nz
- 46. de Almeida CC, Baião DDS, Leandro KC, Paschoalin VMF, Costa MP da, Conte-Junior CA. Protein quality in infant formulas marketed in brazil: Assessments on biodigestibility, essential amino acid content and proteins of biological importance. Nutrients. 2021;13: 3933.
- 47. Nutrition Bhargava A., health, and economic development: Some policy priorities. Food Nutr Bull. 2001;22: 173–177.
- 48. Krebs NF. Dietary zinc and iron sources, physical growth and cognitive development of breastfed infants. J Nutr. 2000;130. pmid:10721906
- 49. Dewey KG. Nutrition, Growth, and Complementary Feeding of The Brestfed Infant. Pediatr Clin North Am. 2001;48: 87–104. pmid:11236735
- 50. Berry NJ, Gribble KD. Health and nutrition content claims on websites advertising infant formula available in Australia: A content analysis. Matern Child Nutr. 2017;13. pmid:27739216
- 51.
Corte V Della, Gaudio G Del, Sepe F. Ethical food and the kosher certification: a literature review. [cited 18 Dec 2023]. https://doi.org/10.1108/BFJ-09-2017-0538
- 52. Cohen E, Schwartz Z, Antonovski R, Saguy S. Consumer perceptions of kosher products. J Foodserv Bus Res. 2002;5: 69–88.
- 53. Lytton T. Kosher Certification as a Model of Private Regulation: Third-Party Certification Has Benefits Over Both Government Regulation and Unregulated Markets. Regulation. 2013;36. Available: https://readingroom.law.gsu.edu/faculty_pub/1986
- 54. Richter APC, Duffy EW, Smith Taillie L, Harris JL, Pomeranz JL, Hall MG. The impact of toddler milk claims on beliefs and misperceptions: A randomized experiment with parents of young children. J Acad Nutr Diet. 2022;122: 533. pmid:34391941
- 55. Duffy EW, Taillie LS, Paula A, Richter C, Higgins ICA, Harris JL, et al. Parental Perceptions and Exposure to Advertising of Toddler Milk: A Pilot Study with Latino Parents. 2021 [cited 18 Dec 2023]. pmid:33435227
- 56.
Belamarich PF, Bochner RE, Racine AD. A Critical Review of the Marketing Claims of Infant Formula Products in the United States. http://dx.doi.org/101177/0009922815589913. 2015;55: 437–442. https://doi.org/10.1177/0009922815589913
- 57. Lwin MO, Vijaykumar S, Chao J. “Natural” and “Fresh”: An Analysis of Food Label Claims in Internationally Packaged Foods in Singapore. J Food Prod Mark. 2015;21: 588–607.
- 58. Kanter R, Reyes M, Corvalań C. Photographic Methods for Measuring Packaged Food and Beverage Products in Supermarkets. Curr Dev Nutr. 2017;1. pmid:29955678
- 59. Maganja D, Davies T, Sanavio L, Louie JCY, Huffman MD, Trieu K, et al. Current food labelling practices in online supermarkets in Australia. Int J Behav Nutr Phys Act. 2023;20: 1–11. pmid:37749593