=Paper=
{{Paper
|id=Vol-3741/paper70
|storemode=property
|title=Food Certification through Collaborative Sensory Analysis Methods and Tools
|pdfUrl=https://ceur-ws.org/Vol-3741/paper70.pdf
|volume=Vol-3741
|authors=Ada Bagozi,Devis Bianchini
|dblpUrl=https://dblp.org/rec/conf/sebd/BagoziB24
}}
==Food Certification through Collaborative Sensory Analysis Methods and Tools==
https://ceur-ws.org/Vol-3741/paper70.pdf
Food Certification through Collaborative Sensory
Analysis Methods and Tools
(Discussion Paper)
Ada Bagozi1,* , Devis Bianchini1
1
University of Brescia, Dept. of Information Engineering
Via Branze 38, 25123 - Brescia (Italy)
Abstract
In the current global food market, there exists a vital need for data-driven tools that ensure the highest
quality of food. Guaranteeing food quality demands meticulous control across the entire production
chain, while adhering to best practices and legal regulations. However, beyond objective metrics for
evaluating food quality, subjective elements derived from sensory analysis hold paramount importance.
Sensory analysis involves assessing food through the five senses: taste, sight, touch, smell, and hearing.
It significantly influences food choices and dietary preferences. The process of preparing a sensory
analysis panel is complex and includes panel leaders, tasters and sensory analysis experts, who are
in charge of analysing the panel results. Therefore, the process can greatly benefit from the use of
specialised tools. These tools must facilitate all phases of the panel, from selecting tasters and food
samples to analysing and visualising results, and must be properly integrated to maximise the outcome
of the sensory analysis. They must also help in appropriately weighing tasters’ input based on their
experience and on-the-fly comparison against other tasters during the panel, ultimately culminating in
the issuance of a food certification. To this aim, in this discussion paper we discuss a comprehensive
suite of tools developed to manage sensory analysis panels. These tools are grounded on a shared
conceptual data model and are specifically designed to evaluate food quality and generate a food certifi-
cate, ensuring that the highest standards are met throughout the food production and assessment process.
Keywords
Agri-food 4.0, Sensory Analysis, Smart Food, Food Certification
1. Introduction
In recent years, consumers are demanding for high-quality food, that calls for meticulous controls
across the entire production chain and compliance with best practices and regulations. In this
context, sensory analysis plays a pivotal role in evaluating food quality, as it provides invaluable
insights into the sensory attributes that define consumers’ preferences. Collected sensory data
encompasses a wide range of attributes, including taste, aroma, texture, appearance, and sound
perception. Nonetheless, the variability of human sensory perception adds complexity to data
collection, emphasising the need for standardised evaluation methods to ensure consistency
and reliability in data interpretation.
SEBD 2024: 32nd Symposium on Advanced Database Systems, June 23-26, 2024, Villasimius, Sardinia, Italy
*
Corresponding author.
$ ada.bagozi@unibs.it (A. Bagozi); devis.bianchini@unibs.it (D. Bianchini)
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
Going beyond objective measurements on food quality and compliance, such as laboratory
tests, sensory analysis has a significant role in food choice and dietary patterns [1]. Data
collected for sensory analysis presents specific features [2] and requires an interdisciplinary
approach to address subjectivity and variability in perception, ensure data quality through proper
data analysis and interpretation, allow data security and privacy, implement longitudinal data
management for continuous monitoring, and handling the volume and complexity of large
datasets. Different types of users, such as sensory scientists, statisticians, data scientists, and
IT professionals, could benefit from the adoption of proper tools that might support all phases
of the sensory analysis process, ranging from the selection of tasters and food samples, to the
analysis and visualisation of results, to outcome a food certification.
In this discussion paper, we introduce a comprehensive suite of tools developed to support a
tasters association in the preparation, execution and analysis of the sensory analysis process.
The process firstly involves panel leaders, who are in charge of preparing the product profile
(i.e., a list of descriptors used to characterise a food product using five senses), select the tasters
and food samples and supervise the execution of the tasting panel. Tasters are asked to assign a
vote in a discrete scale to each descriptor (e.g., on a scale from 0 to 9, how much astringency is
perceived or how much attractive the product appears) according to their sensory experience
(stimulation caused by aromas and flavours of the product). Based on the answers collected
through multiple panels, a food product certificate is prepared, to establish the quality of
the product for future comparisons and quality assessment. These steps are supported by an
integrated set of software tools that guide the involved actors [3]. To the best of our knowledge,
this is the first attempt to provide such a completely integrated suite of tools in this domain,
bridging together data management about the portfolio of tasters, statistical analysis to judge
tasters according to their experience and data modelling to pursue food quality and certification.
This is the first step in a project where food certificate will be correlated with Big Data collected
on field during food production (implementing the so-called Agri-Food 4.0) and the feedback of
the final consumers.
The paper is organised as follows: Section 2 provides a Sensory Analysis overview; Section 3
introduces the Sensory Data Model; Section 4 delves into Sensory Data Analysis; Section 5
outlines the software architecture of the tool suite; Section 6 reviews Related Work. Finally,
Section 7 offers concluding remarks and future directions.
2. Sensory Analysis Overview
Figure 1 depicts the sensory analysis process, that begins with a tasters’ association receiving
a commission for food evaluation (for instance, from a chocolate manufacturer interested in
certifying the quality of products) and culminates in the issuance of a food certificate based on
the analysis results. The Panel Leader, an expert with significant experience in sensory analysis,
embarks on the critical task of preparing the sensory panel. This phase is marked by a series of
deliberate and well-considered actions aimed at ensuring the panel’s effectiveness and reliability
and improving the panel compliance to standardised procedures. The panel leader’s role is
pivotal, involving the meticulous selection of tasters who have received extensive training
to participate in food evaluation panels. The panel leader is in charge of defining the set of
Figure 1: Sensory Analysis overview
sensory attributes (descriptors) that are relevant for product evaluation. For instance, in the
case of chocolate, examples of descriptors could be creamy texture and floral aroma. The panel
leader also selects the most suitable methodology for the session, guaranteeing an efficient and
effective assessment of the product’s sensory quality.
With the sensory panel in place, the evaluation of samples can begin. This phase is charac-
terised by a structured and controlled approach, essential for maintaining the integrity of the
sensory data collected. The tasters, now thoroughly trained, systematically evaluate the food
product using their calibrated senses to determine its quality. To avoid any bias, samples are
presented in a neutral and standardised manner. The tasting sessions are held in environments
specially designed to minimise external influences and distractions. Factors such as lighting,
temperature, and even the room’s color are meticulously controlled, complemented by the
integration of a user-friendly graphical user interface (GUI) for data collection. Employing blind
testing methods, where the identity of the samples remains hidden, ensures tasters base their
evaluations solely on the product’s sensory attributes, thus relieving tasters from the influence
of environmental factors, for example allowing the chocolate manufacturer to objectively as-
sess the appeal and flavour profile of the product. This harmonised setting, together with the
advanced collection tool, streamlines the evaluation process, allowing tasters to efficiently and
accurately record their observations. Tasters assess the product using the previously defined set
of descriptors, rating each attribute on scales measuring intensity, preference, or quality. The
tasters’ responses are compiled, resulting in a comprehensive dataset that mirrors the collective
sensory experience of the product.
Advanced statistical methods are then applied to analyse the collected data. Through sta-
tistical analysis, subtle nuances and significant trends are uncovered, providing a profound
understanding of the product’s sensory attributes. This methodical approach enables the quan-
tification of sensory perceptions, converting subjective experiences into objective data. A crucial
aspect of the analysis involves evaluating the tasters’ performance to ensure consistency and
alignment with the calibration phase standards.
Ultimately, the Panel Leader utilises the analysis results to produce a food certificate. This
document consolidates the sensory evaluation findings, offering a formal and authoritative
SampleCategory
* 1
…
1 Sample Equipment
* 1 … * *
* Method
ConformityProfile …
PanelConfiguration * Sample
…
Template: Boolean 1 *
* …
1 *
1 * 1 * 1
SamplePosition 1
* SensoryPanel …
*
… 1 *
DescrioptorInConformity *
Profile * * * *
Panel Leader 1
Range
PanelSession 1 * EvaluationSheet
…
… Value
*
InputOptions * 1
…
*
1 *
InputType 1
… * Taster ValueAssign
1 Value
…
* *
*
1 Descriptor 1
…
Figure 2: Sensory data conceptual model on which the suite of tools relies.
assessment of the product sensory quality, which can greatly enhance the manufacturer’s
market positioning.
3. Sensory Data Model
The unified view over the data exploited by different components of the tool suite is one of
the strengths of the approach, with respect to the existing systems for sensory analysis. The
collection of sensory data is initiated through the Input Sensory Software (ISS), designed to
organise panel sessions as standard procedures. Afterwards, data is exported to be analysed
using the Big Sensory Software (BSS), that implements several statistical metrics. Finally, sensory
data collected over time is stored to enable longitudinal data management through the Data
Sensory Software (DSS).
Figure 2 illustrates a portion of the conceptual data model which the tools suite is based on.
The Panel Leader is in charge of creating and overseeing a Sensory Panel. For this purpose,
the Panel Leader selects the Samples for evaluation, associated to a Sample Category (e.g. red
wine or white chocolate). The evaluation Method, encapsulating a set of rules and guidelines, is
selected to make the evaluation process compliant to best practices, such as the proper sequence
in which the samples are presented to the tasters. Moreover, the Panel Leader defines the
Descriptors set (e.g., astringency or flower aroma). A descriptor is tailored to a specific Sample
Category — for instance, creamy texture and shiny appearance can be descriptors used for
chocolate, whereas tannin taste or earthy aroma better suit wine. Descriptors are characterised
by an Input Type (e.g., number, text, range, select), to enable data collection. Some Input
Types may feature a predefined set of Input Options. Dropdowns, sliders, and other kinds of
predefined input sets are tailored as usual to reduce ambiguity and enhance the precision of
data entry, directly addressing the complexity of sensory data translation. Sample Category,
Figure 3: Sensory analysis result for tasters (tasters) evaluation.
Method and Descriptors composes the Panel Configurations. These configurations are crucial
for loading the appropriate descriptors for the selected sample category.
After establishing the panel, the Tasters start the Panel Session. During the panel session,
tasters evaluate one or more food samples. Each sample is associated with an Evaluation Sheet,
that contains the Assigned Values of descriptors for the assessment. Using the ISS, tasters
assign ratings to each descriptor of samples, allowing the Panel Leader to monitor the panel
progress. The Panel Leader also plays an important role in sustaining the tasters’ engagement
throughout the session.
Once the panel has been completed, the Panel Leader exports the collected evaluations
for further analysis. The analysis is aimed at preparing a Conformity Profile for a Sample
Category. Conformity Profile ensures that the product meets specific criteria established by
regulatory bodies, industry standards, or internal benchmarks. Such criteria are expressed as
expected ranges of values for each descriptor. For example, in order to assure quality of the
tasted chocolate, the flower aroma descriptor range should be from 6 to 8.
4. Sensory Data Analysis
Sensory analysis experts leverage the Big Sensory Software (BSS), a sophisticated web-based
tool, for statistical analysis of product evaluations collected by the Panel Leader. The involved
statistical techniques are tailored to handle the inherent variability and subjectivity of sensory
perception. For example, the analysis of variance (ANOVA) is used for identifying significant
differences between sample groups. The use of statistical techniques serves a specific purpose in
analysing different aspects of sensory data, ranging from detecting differences in mean ratings
to assessing the reliability and effectiveness of sensory descriptors. The BSS is designed with the
overarching aim of enhancing the quality certification of each product through comprehensive
sensory analysis, encompassing data collected throughout the entire product lifecycle—from
production to distribution. This data, typically gathered via sensors, offers a basis for corre-
lating product quality levels identified through sensory analysis, thereby informing potential
improvements within the food production chain. As depicted in Figure 3, the BSS features an
intuitive Graphical User Interface (GUI), empowered with a statistical engine, completed with
an extensive array of indices for assessing the significance of panel results. This interface allows
experts to meticulously examine the quality of the product and validate the accuracy of tasters’
evaluations using various analytical techniques. For instance, majority-based techniques can
identify and possibly exclude evaluations from less skilled tasters, ensuring the reliability of the
analysis. Such a measure is crucial when tasters may not meet the requisite quality standards for
reliability due to factors like distraction or fatigue, thereby mitigating the risk of compromising
the analysis with unreliable data.
The BSS GUI offers diverse methods for visualising the results of sensory analysis, ranging
from tables to charts, thus providing a comprehensive view of the data at hand. Moreover,
it enables experts to compile and export detailed reports, encapsulating all findings from the
sensory analysis. This functionality not only facilitates a deeper understanding of the product
sensory attributes, but also supports informed decision-making regarding quality assurance
and product development.
5. Tools Architecture and Implementation
Figure 4 presents the architecture of the Sensory Analysis Tools. This project has been designed
as a microservices architecture to ensure modularity and scalability. The system uses two
databases: a MySQL database for storing relational data, including Sensory Panel and configu-
ration data, and a MongoDB database for holding all the collected through panel sessions.
On top of these databases, a suite of micro-services operates, each one belonging to different
components of the tool suite. Among the ISS Services, the Panel Service is invoked to manage
sensory panels. This service closely interacts with the Taster Service and the Sample Service,
which are responsible for managing tasters and samples, respectively. Additionally, the Collec-
tion Data Service generates evaluation forms and compiles evaluation data to be saved on the
MongoDB database.
Conversely, the BSS Services include the Sample Evaluation Service, which is in charge of
starting the sensory analysis starting from data collected through ISS services. The Sensory
Figure 4: Sensory Analysis Tools architecture
Analysis Service is invoked to produce analysis results, and uses the Conformity Profile Service,
which establishes product profiles for evaluating results and generating certifications through
the Food Certification Service.
Finally, the Data Sensory Software (DSS) is in charge of managing and navigating the historical
data, thought the Historical Data Service.
All the tools 1 , namely ISS, BSS and DSS, have been developed in a modular way. The GUI is
crafted with ReactJS, while the server-side logic employs NodeJS with TypeScript. The Sensory
Analysis Service, derived from a legacy and standalone implementation of the statistical methods,
was available in Perl and integrated in the overall tool suite. To booster the efficiency, reliability,
and scalability of the server-side applications, the NestJs framework has been integrated. A
unified API Gateway sits atop the micro-services, although the architecture permits the creation
of multiple gateways for different deployment strategies.
For web application integration and management within a single development environment,
the NX build system has been used. Considering the variety of technologies involved and to
1
Tools Demo: https://youtu.be/03Mr3uXA5kg
streamline the release process of the integrated web applications, Docker containers have been
adopted. The current setup comprises three Docker containers: one for each database and
another deploying the APIs.
6. Related Work
Transitioning from paper to digital sensory analysis tools has been a pivotal advancement,
addressing inefficiencies and enhancing data management. Platforms like Compusense, RedJade,
Fizz Software, SIMS 2000, and EyeQuestion have significantly contributed to this field by offering
advanced functionalities for sensory evaluation and are widely used in the field [4, 5, 6, 7, 8, 9, 10].
Moreover, the introduction of the BioSensory app, which records participants’ biometric re-
sponses, represents an innovation in incorporating biometrics into sensory analysis [11]. Despite
such progresses, the high costs and limited remote analysis capabilities of these tools present
barriers, indicating a need for more accessible and functional solutions. Traditional tools, while
capable in their own rights, often fall short in facilitating effective remote sessions. In contrast,
the proposed tools suite is specifically designed to support remote analysis, incorporating
features that ensure participant engagement and the integrity of data collected in such settings.
This adaptability positions the suite as particularly relevant in the current global context.
The adaptation of digital tools like Zoom and Google Forms for remote sensory sessions
during the pandemic demonstrates the field resilience and flexibility [12]. However, these
general-purpose tools lack the specialised features necessary for detailed sensory analysis.
The proposed tools suite fills this critical gap, offering both the flexibility required for remote
analysis and the analytical depth needed for comprehensive sensory evaluations.
In conclusion, while the foundational work of existing sensory analysis software has signifi-
cantly contributed to the field, our tools aim to build upon this foundation by addressing the
critical needs of cost, data integrity, and remote analysis capability.
7. Concluding remarks
In this paper, we discussed an integrated tools suite designed to implement food quality as-
sessment through the lens of sensory analysis. Our approach provides a novel contribution to
the emerging field of Agri-Food 4.0, setting the stage for a future, where food certification is
strictly linked with comprehensive big data collected throughout the food production cycle. As
the tools suite is operative in test mode at the moment, current work is focused on conducting
thorough evaluations of users’ experience, including among users experts of sensory analysis,
panel leaders and professional tasters. Moreover, we recognise the potential of blockchain
technology in fostering transparency, immutability, and trust among all participants in the
food quality assurance ecosystem. Therefore, our future work will explore the integration of
blockchain technology to safeguard the integrity of data exchanged across the food supply
chain. In this scenario, the proposed tool suite can interact directly with the blockchain. By
doing so, we aim at establishing a new benchmark for data quality and security in food quality
assessment, paving the way for more informed decision-making and consumers’ confidence.
References
[1] A. Drewnowski, Taste preferences and food intake, Annual review of nutrition 17 (1997)
237.
[2] G. Rocchetti, G. P. Braceschi, L. Odello, T. Bertuzzi, M. Trevisan, L. Lucini, Identification
of markers of sensory quality in ground coffee: An untargeted metabolomics approach,
Metabolomics 16 (2020) 1–12.
[3] S. Fuentes, E. Tongson, C. Gonzalez Viejo, Novel digital technologies implemented in
sensory science and consumer perception, Current Opinion in Food Science 41 (2021)
99–106.
[4] B. Kemp, S. Trussler, J. Willwerth, D. Inglis, Applying temporal check-all-that-apply (tcata)
to mouthfeel and texture properties of red wines, Journal of Sensory Studies 34 (2019)
e12503.
[5] C. Sharma, S. S. Jayanty, E. Chambers, M. Talavera, Segmentation of potato consumers
based on sensory and attitudinal aspects, Foods 9 (2020). doi:10.3390/foods9020161.
[6] W. Yin, M. Washington, X. Ma, X. Yang, A. Lu, R. Shi, R. Zhao, X. Wang, Consumer
acceptability and sensory profiling of sesame oils obtained from different processes, Grain
Oil Science and Technology 3 (2020) 39–48.
[7] Y. Kong, C. Sharma, M. Kanala, M. Thakur, L. Li, D. Xu, R. Harrison, D. D. Torrico, Virtual
reality and immersive environments on sensory perception of chocolate products: A
preliminary study, Foods 9 (2020).
[8] B. M. Adhikari, T. Truong, S. Prakash, N. Bansal, B. Bhandari, Impact of incorporation of
co2 on the melting, texture and sensory attributes of soft-serve ice cream, International
Dairy Journal 109 (2020) 104789.
[9] S. Fuentes, Y. Y. Wong, C. Gonzalez Viejo, Non-invasive biometrics and machine learning
modeling to obtain sensory and emotional responses from panelists during entomophagy,
Foods 9 (2020).
[10] S. Biju, S. Fuentes, C. Gonzalez Viejo, D. D. Torrico, S. Inayat, D. Gupta, Silicon supple-
mentation improves the nutritional and sensory characteristics of lentil seeds obtained
from drought-stressed plants, Journal of the Science of Food and Agriculture 101 (2021)
1454–1466.
[11] S. Fuentes, C. Gonzalez Viejo, D. D. Torrico, F. R. Dunshea, Development of a biosen-
sory computer application to assess physiological and emotional responses from sensory
panelists, Sensors 18 (2018).
[12] C. Gonzalez Viejo, H. Zhang, A. Khamly, Y. Xing, S. Fuentes, Coffee label assessment
using sensory and biometric analysis of self-isolating panelists through videoconference,
Beverages 7 (2021).