Validators are used to ensure that internet based services present information in an adequate way to the end users. However, today's validators are not transparent making it di cult for the users to understand their feedback. To enhance the transparency we propose a MDE based approach to program validation, where domain requirements are presented in a metamodel. The validation is based on a bottom up approach where programs are parsed and represented as models before they are checked against the metamodel.
As the internet becomes an increasingly more important tool for communication,
it is necessary to ensure that not only is the quality of internet based digital
services adequate, but that they are also equally accessible to all end users.
Furthermore, as web based technologies continue to develop rapidly, the need arises for
automated tools that can ensure that web pages and digital services comply with
the latest recommended standards. These standards are designed for the bene t
of the end user, ensuring that the provided digital service functions adequately
and is accessible to all end users, especially those with disabilities. Moreover, by
following the appropriate guidelines for the digital service, the developer can be
certain that the digital service they have designed, is of a decent level of quality.
In other words, these standards are bene cial to both those who develop the
digital services, and those who use them. One of the better known accessibility
standards is the The Web Content Accessibility Guidelines 2.0 (WCAG 2.0) [
It is common practice to determine the quality of a web page with the aid
of an automated web evaluation tool. A web evaluation tool can be de ned
as an application that inspects if a web page satis es a set of requirements.
Any elements contained within the web page that violates one or more of the
requirements are agged as invalid, and included in a report that is presented
at the end of the evaluation process. Some of the more advanced evaluation
tools, such as Wave [
Furthermore, the lack of transparency in the evaluation process makes it
di cult for the user to evaluate the e ectiveness and legitimacy of the web
evaluation tool. In [
Finally, the lack of consensus in the number of errors reported by existing web
evaluation tools indicate a need for clearer and more precise formulation of the
requirements in the WCAG 2.0 guidelines [
To solve these problems we introduce a validation tool based on the
bottom up approach to program validation. The approach utilizes methods and
techniques from Model Driven Engineering [
The remainder of this paper is organised as follows: In section 2 we give a high level description of the The Bottom-Up Model Validation approach before we
Technological Space Conceptual Space Web page .html .css .js extract extract extract
HTML Model Discoverer
CSS Model Discoverer
JS Model Discoverer generate generate generate Phase One: Model Discovery
The WCAG 2.0 model
Conformance
Check Derived Model model-to-model transformation Initial Model
Phase Three: Model Validation Phase Two: Model Understaning give a more detailed description of each of the phases: in section 2.1 we describe the metamodel used in the bottom up validation approach, in section 2.2 we describe the model discovery phase, in section 2.3 the model understanding phase is explained and in section 2.4 is it explained how the model is validated against the WCAG 2.0 standard. In section 3 we discuss how the bottom up approach is related to other contributions before we conclude the paper and envision some possible further work in section 4. 2
The Bottom-up Model Validation approach, hereby referred to as the Bottom-up
approach, is inspired by the interactive Bottom-Up Meta-Modelling approach
described in [
Our solution consists of two parts; the rst part focuses on designing a DSML
capable of providing a clear and precise description of the mandatory
requirements. The second part combines the DSML with Model Driven Reverse
Engineering (MDRE) techniques, to determine if the digital service in question
satis es all the necessary requirements [
The basic idea behind the Bottom-up approach is to extract an abstract model representation of a web page using Reverse Engineering (RE) combined with MDE. The extracted model can be manipulated, transformed and validated using techniques from MDE. By con rming that the model is valid, our evaluation process can state that the corresponding artefact is also valid. The actual evaluation process itself is reduced to a conformance check between the extracted model representation of the artefact, and the meta-model that models the necessary requirements. Figure 1 presents an overview of the process, and as we can see in the gure, the bottom-up approach consists of three phases: { Phase One: An initial model is extracted from the artefact using platform speci c model discoverers. The main goal of this phase is to transfer the problem from the technological space to the conceptual space as early as possible in the validation process, details will be given in section 2.2. { Phase Two: The initial model is manipulated using techniques from MDE.
The main goal of this phase is to transform the initial model into a model more suitable for our purposes. The transformed model is referred to as the derived model. More details about this will be presented in section 2.3. { Phase Three: In the third phase, a conformance check between the derived model and the metamodel is executed. The metamodel contains the requirements the artefact must meet. The goal of this phase is to determine if the abstract model representation of the artefact is a valid instance of the metamodel. This will be discussed further in section 2.4. 2.1
As previously mentioned, the main purpose of the DSML is to provide a clear
and precise de nition of the requirements our artefact must satisfy. One of the
advantages with describing the requirements using a DSML is that the domain
expert get the opportunity to con rm that the requirements have been
correctly interpreted by the evaluation process, thus providing the means to reduce
the number of false negatives and false positives in the results. The
BottomUp approach uses a multi level metamodelling hierarchy [
Value AttributeValue
AttributeName
Event Attribute
Reference [irr]
Element [irr]
Containment types belonging to the domain, and how they are related. Figure 2 shows an example of the top level metamodel describing an element in a web page. The gure shows that each element may have an arbitrary number of attributes and events, and that each element may relate to other elements by either containing them, or by referring to them. The irre exive constraint, represented using the [irr] -notation, prevents any element from containing or referring to itself.
The metamodel at the next level in the metamodelling hierarchy describes the
actual requirements we wish to impose upon the artefact. At this level we include
in addition to the typing of the artefacts and the relationships between them,
modelling constraints that ensure that the corresponding part of the artefact is
valid. Figure 3 display a model of an actual requirement from WCAG 2.0. The
1.1.1 Non-text Content requirement [
The bottom metalevel of the hierarchy is where the derived model is placed after the model understanding phase, this will be elaborated in section 2.4. 2.2
The Model Discovery phase is the rst part of the MDRE process, a raw model
is extracted from the artefact in question, see gure 4. These raw models are
referred to as initial models and are extracted as early as possible during the
MDRE process in order to avoid losing any essential information [
We now continue our example involving the 1.1.1 Non-text Content
requirement. At this stage in the bottom-up approach we do not know what parts of
the HTML-document are of importance when determining if all non-text
content are adequately described. Therefore the initial model is extracted using a
HTML speci c model discoverer, which transforms the HTML document model
to a generic XML based model that conforms the HTML metamodel. This model
is of a low level of abstraction ensuring that as little information as possible is
lost during the transition from the technological space to the conceptual space.
The main purpose of this generated initial model is to serve as a starting point
for the model understanding phase [
During the Model Understanding phase, chains of model manipulation
techniques are used to query and transform the initial model in order to obtain a
model of the artefact at a higher level of abstraction, see gure 4. The result of
this phase is the Derived Model [
Model Understanding Phase HTML-metamodel
Metamodel A
Metamodel driven Legacy Artefact
Extract
Platform specific Model Discoverer
Populate
Initial Model Representation of
Conforms to dMreivteanmodel
Conforms to Transformation
Derived Model A Fig. 4. The Model Discovery Phase and the Model Understanding Phase Meta-model Derived Model Webpage
img1 : HtmlImageElement
img2 :
HtmlImageElement <img src="car.png" /> <img src="car.png" alt="An image of a nice red car" />
alt :AttributeType altValue
text :AttributeValue attr alt : alt value An image of a nice red car :
text HTML element is excluded from the process. This phase transforms the HTML element to a generic a set of nodes and edges representing the HTML element and its attributes with a graph-based visual syntax. Irrelevant information such as user events attached to the element are not included in the transformation, ensuring that the new representation of the HTML image element only contains the information that is necessary in order to con rm that it satis es the 1.1.1 Non-text Content requirement. 2.4
The nal step in the Bottom-Up approach consists of checking if the derived model successfully conforms to the metamodel, if any part of the derived model is incorrectly typed, or doesn't satisfy the corresponding predicate, it is agged as invalid. This encompasses the actual evaluation process that determines if the artefact contains any components that fail to meet the necessary requirements. The feedback presented to the user is purely graphical, consisting of a graph-based model representation of the HTML-document and a graph-based representation of the metamodel containing the constraints the HTML-document must satisfy. Any parts of the model representation that violates the metamodel, is highlighted red, warning the user of the violation. Concluding the example from the two previous steps, any HTML image element that is missing the alt attribute will not conform to the metamodel reporting that the multiplicity predicate has not satis ed. Figure 5 illustrates the model validation phase using two image elements. The image element img1 does not contain an alternative text, therefore it does not conform to the meta-model and is agged as invalid. The image element img2 however, does contain an alternative text and therefore conforms to the meta-model. 3
We now compare the Bottom-up approach to related approaches and existing
web evaluation tools. The model driven reverse engineering framework MoDisco
introduced in [
Software Modernization is a typical application of MDRE according to [
The Bottom-up approach suggests an alternative use of the derived model.
Where [
AChecker presented in [
Transparency Problem Completeness Problem Correctness Problem Speci city Problem Ambiguity Problem
No No No Yes Yes No No Yes Yes No No Yes Partly Partly Yes Yes
N/A N/A N/A Yes
The Bottom-Up approach and the validation tool indicate that it may serve as
a viable automated web evaluation tool that addresses several critical concerns.
Based on what is presented in [
We now envision some possible extensions of the bottom up approach.
Model Veri cation: Model conformance is just the tip of the iceberg when
discussing model validation and veri cation, another factor that should be
examined is the models consistency. This is addressed in the satis ability problem
which investigates if a meta-model contains any contradicting constraints, if so
it will be impossible for a model to successfully conform to the meta-model. This
problem could be applied to the metamodel describing the requirements, where
it could be determined if the artefact can actually satisfy all of the requirements
included in the metamodel, or if changes to the requirements con ict with other
existing ones. In the future we will use similar techniques as presented in [
Model Adequacy: The Bottom-Up approach does not evaluate the actual
artefact, but rather an abstract model representation of it. This means that in
order for the evaluation process to present accurate results, the derived model
needs to be adequate. An adequate derived model must contain all the necessary
information that is relevant to the evaluation process. In order to ensure that
the model is adequate, some sort of automated mechanism should be included
in solution. The problem with model adequacy was rst addressed in [
Completing the Metamodel: The metamodel described in this paper only
covered 1 of the 35 mandatory requirements described in the WCAG 2.0
guidelines. However in [