Data visualizations and dashboards are powerful means to convey information to large audiences. However, the design and understanding of these tools are not straightforward because several factors are involved. It is essential to rely on theoretical frameworks to design and implement data visualizations for these reasons. In this context, we propose a meta-model to identify and arrange the main characteristics and elements of data visualizations and dashboards. The proposed meta-model provides a powerful artifact to generate information visualizations and dashboards automatically, but also a learning resource to understand how data visualizations elements interact and influence each other. However, it is necessary to validate this artifact to ensure its quality and usefulness. In this paper, we present a work-inprogress or a quality assessment and content validation of the me-ta-model to seek weaknesses and tackle them in subsequent iterations.
Visual explanations are everywhere: they convey complex information, raise attention over target
topics, improve the understandability of particular domains, etc. [
These tools are also powerful because they let users visually perceive information to generate knowledge. However, the complexity involved in this domain can hamper the understanding of the displayed data.
Data visualizations are composed of different visual elements, including shapes, visual encodings (i.e., visual characteristics like color, size, position, etc.), and visual aids (i.e., legends, axes, etc.), but also abstract concepts like scales, data domain, data operations, among others.
If these concepts are not fully understood, they can lead to improper designs (on the developer’s side) and to wrong conclusions (on the audience’s side).
It is essential to consider all the concepts and factors involved in the data visualization domain to provide adequately designed and reliable visualization methods.
These factors can be considered through domain experts (i.e., information visualization experts) who also know the visualization’s data do-main and can provide a well-designed product through its expertise.
However, it isn’t easy to have this expertise or domain knowledge levels for every practitioner who uses information visualizations to convey information. For all these reasons, it is crucial to deeply understand all the elements that compose data visualizations and how they relate to and in-fluence each other.
The abstraction of these elements and their relationships can provide a framework to improve the knowledge about the most primitive aspects of data visualizations, no matter if they are tangible (understood as the elements that are directly displayed through data visualizations, such as shapes, encodings, layouts, etc.) or conceptual (data domains, data transformations, user characteristics, etc.).
To achieve this, paradigms like the model-driven architecture [
One of the main benefits of relying on a meta-model is that it can be used as a conceptual map and
educational tool to guide the design of data visualizations, but also as an artifact to automatically
generate analytical dashboards in different domains (health [
In this paper, we provide a proposal to validate the content of the dashboard. By validating the content of the meta-model, it is possible to identify potential limitations and drawbacks of the dashboard domain representation and address them before using this artifact to instantiate real-world dashboards.
The rest of this paper is organized as follows. Section 2 describes previous works related to metamodel validations. Section 3 describes the methods employed to carry out the meta-model, the automatic generation of dashboards, and the experts’ validation questionnaire. Finally, section 4 discusses the results, and section 5 offers the conclusions derived from this work.
A meta-model is considered of quality if it contains the required elements to instantiate a model that
adequately represents the elements in the domain and is technically built using solid principles. The
first concern is related to meta-model validation, which is necessary to ensure that we are building the
right meta-model [
There are few quality models or frameworks to measure the quality of a meta-model [
On the other hand, Marín, et al. [
Regarding validation techniques, there is no specific literature focused on validation methods for
meta-models. According to the systematic literature review of meta-models for software quality and its
evaluation [
Although meta-models are prone to extensions and modifications, the main entities and relationships must be identified to ensure the proper development of dashboards, in this case.
The dashboard meta-model has been developed using a domain engineering [17, 18] and an example-driven approach [19], in which different dashboards and data visualizations were analyzed to extract main and standard features among their elements. This methodology allows the identification of abstract qualities that can be arranged into a meta-model to obtain fully functional dashboard instantiations.
We developed the first version of an automatic dashboard generator based on the meta-model to test if it supports the generation of real-world dashboard and visualization examples. The code generator inputs a set of parameters that account for the elements and attributes of the meta-model, and the result is the source code of a dashboard according to the provided configuration.
The approach taken to automatically generate the source code is based on the software product line (SPL) paradigm [20, 21] and we developed different HTML and JavaScript code templates [22] to materialize the variability points of the product line [23].
During the development and testing of the dashboard generator, the meta-model was subject to modifications because this process raised new dimensions and relationships that need to be accounted for during the design and development of these tools. Figure 1 presents the current version of the dashboard meta-model.
We created an online questionnaire in which six different sections of the meta-model (dashboard layout, user characteristics, goals and tasks, user and dashboard relationships, data visualization primitives, and data domain and operations), in addition to the whole meta-model, were scored in terms of the mentioned dimensions using a 1-4 scale, where one implies that the section does not meet the criterion, and four that it highly meets the criterion. Figure 2 shows the rubric used to score the different dimensions based on previous works on content validation by experts [25].
We also added a brief explanation of the meta-model section and a “Yes/No” question to test if the representation meets the intended goal of the section. Finally, we included an open text field to collect any qualitative feedback or justification that experts might have. The questionnaire was configured in a customized installation of LimeSurvey and sent by e-mail to different domain experts from both the software engineering and data visualization fields.
This paper presents our work-in-progress for validating a meta-model framed in the data visualization and dashboard domain. The meta-model is set to be used as a foundational learning resource for users to understand the primitive elements of data visualizations, which is, in turn, important to reach proper insights.
The meta-model has been subject to continuous modifications following the domain engineering approach, including not only tangible and visual elements but also abstract concepts related to the user and data domain. The resulting outcome is a complex meta-model (Figure 1) that tries to enclose every primitive element involved in the data visualization domain.
Due to the meta-model’s inherent complexity and our intention to use it as a learning resource, we proposed a content validation of this artifact before integrating it into real-world processes.
The validation questionnaire has been designed to measure the coherence, relevance, and clarity of the dashboard meta-model. It was necessary to divide the meta-model into different sections to ease the analysis. This questionnaire has been sent to experts in the data visualization and software engineering domain with the goal of obtaining different perspectives about the content of the meta-model.
The measured dimensions will provide crucial information to improve the meta-model, which will be translated into a better learning resource that provides the most relevant elements in data visualizations in an understandable, clear, and coherent manner.
Future research will involve the analysis of the questionnaire responses as well as the improvements related to the potential issues that might arise from the expert validation.
Once the validation has been carried out, we plan to use the meta-model as a learning resource by implementing an educational tool focused on improving the understandability of the elements involved in data visualizations and information dashboards.
This research work has been supported by the Spanish Ministry of Education and Vocational Training under an FPU fellowship (FPU17/03276). This research was partially funded by the Spanish Government Ministry of Economy and Competitiveness through the DEFINES project grant number (TIN2016-80172-R) and the Ministry of Science and Innovation through the AVisSA project grant number (PID2020-118345RB-I00).
[15] S. Wagner et al., "Operationalised product quality models and assessment: The Quamoco approach," Information and Software Technology, vol. 62, pp. 101-123, 2015/06/01/ 2015, doi: https://doi.org/10.1016/j.infsof.2015.02.009. [16] P. Mohagheghi and V. Dehlen, "A metamodel for specifying quality models in model-driven engineering," in Proceedings of the Nordic Workshop on Model Driven Engineering, 2008, pp. 5165. [17] G. F. Arango, "Domain engineering for software reuse," 1988. [18] K. C. Kang, S. G. Cohen, J. A. Hess, W. E. Novak, and A. S. Peterson, "Feature-oriented domain analysis (FODA) feasibility study," Carnegie-Mellon University, Software Engineering Institute, Pittsburgh, PA, USA, 1990. [19] J. J. López-Fernández, J. S. Cuadrado, E. Guerra, and J. De Lara, "Example-driven meta-model development," Software & Systems Modeling, vol. 14, no. 4, pp. 1323-1347, 2015. [20] P. Clements and L. Northrop, Software product lines. Addison-Wesley, 2002. [21] K. Pohl, G. Böckle, and F. J. Van der Linden, Software Product Line Engineering: Foundations,
Principles and Techniques. New York, NY, USA: Springer-Verlag New York, Inc., 2005. [22] A. Vázquez-Ingelmo, F. J. García-Peñalvo, and R. Therón, "Addressing Fine-Grained Variability in User-Centered Software Product Lines: A Case Study on Dashboards," in World Conference on Information Systems and Technologies, 2019: Springer, pp. 855-864. [23] A. Vázquez-Ingelmo, F. J. García-Peñalvo, and R. Therón, "Taking advantage of the software product line paradigm to generate customized user interfaces for decision-making processes: a case study on university employability," PeerJ Computer Science, vol. 5, p. e203, 2019/07/01 2019, doi: 10.7717/peerj-cs.203. [24] R. Skjong and B. H. Wentworth, "Expert judgment and risk perception," in Proceedings of the Eleventh (2001) International Offshore and Polar Engineering Conference (Stavanger, Norway, June 17-22, 2001): International Society of Offshore and Polar Engineers, 2001. [25] J. Escobar-Pérez and A. Martínez, "Validez de contenido y juicio de expertos: Una aproximación a su utilización," Avances en Medición, vol. 6, pp. 27-36, 01/01 2008.