The expressiveness of a conceptual model depends on the set of language symbols used for representation. UML is one of the most commonly used languages for representing conceptual models. However, issues remain regarding expressiveness that the language OntoUML proposes to resolve. Therefore, we performed an experiment involving eight professionals and eighty students to evaluate the expressiveness of both languages. The overall analysis showed that OntoUML was selected by the participants the most expressive language in 42% of the situations, while in 39% it was selected as having the same level of expressiveness as UML. After further analyses, we identified situations in which OntoUML was the most expressive.
ontologically well-founded primitives that help represent the reality of a problem’s domain as precisely as possible.
Considering these issues,
In his thesis,
The remainder of this paper is organized as follows. In Section 2 we present some basic concepts related to OntoUML. In Section 3, we present our research method. Section 4 discusses the results of our experiment. Section 5 includes final considerations and indication for future studies.
OntoUML was proposed by
In this study, only the main constructs that comprise the object type category are presented (Guizzardi et al., 2011). In this category, constructs are more closely related to the static conceptual modeling of a domain. The hierarchical structure of these models is presented in Fig. 1. The object type constructs may be sortal and non-sortal. Sortals provide identity and individuation principles to their instances, whereas non-sortals do not supply any clear identification principles. Sortal constructs are classified as rigid and anti-rigid sortals. A sortal is said to be rigid if it is necessarily applied to all its instances in all possible worlds. A sortal is said to be anti-rigid if it is not necessarily applied to all its instances. Rigid sortals include kind and subkind categories. A kind is a rigid sortal and thus has intrinsic material properties that provide clear identity and individuation principles. It determines existentially independent classes of things or beings and are said to be functional complexes. A subkind is also a rigid type that provides an identity principle and has some restrictions established and related to the kind construct. Every object in a conceptual model must be an instance of only one kind.
Two sub-categories of anti-rigid sortals exist: phases and roles. In both cases, instances may change their types without affecting their identities. During the phase construct, changes may occur as a result of changes to intrinsic properties. By contrast, in the role construct, changes occur because of relational properties.
Type
ObjectType Sortal Type
Non-Sortal Type Rigid Sortal Type
Anti-Rigid Sortal Type Kind subKind
Phase
Role
This section describes the phases of our experiment conducted to evaluate the expressiveness of the OntoUML and UML languages in a specific context.
The first step in our experiment consisted of defining a context for the construction of the conceptual model. The objective was to select an uncommon domain, that is one not commonly known (e.g., a library, a university.) with a smaller scope to lend feasibility to the experiment. We believe that an uncommon domain brings more discussion to find their concepts and relationships.
In accordance with these criteria, the software requirement specifications for an electronic proxy software program were obtained from specialists in the domain. Based on these specifications, a description of the main software features was written. This description is presented in Table 1.
Only the organization’s representative can grant an electronic proxy.
An organization may have one or more representatives.
To allow the grantor to indicate an active user in the Receita-PR database to grant the condition of the grantee.
Only one grantee per proxy.
Only one proxy per grantor and the same grantee.
The granting of a proxy is restricted to organizations with a record in the ICMS database.
To display the services to be granted.
To display a list of organizations (in which the grantor is the organization’s representative) to be granted.
To select the organization allowed to perform all services.
To allow the grantor to revoke a proxy. 3.2 Construction of a Conceptual Model in OntoUML Based on the scope defined in Section 3.1, specialists were selected to construct a conceptual model in OntoUML. As OntoUML is still not a widely used language on the market, few specialists in this language exist. One of the groups trained for this task is the Ontology and Conceptual Modeling Research Group (NEMO). This group works on research related to ontologies as well as OntoUML, and is led by Professor Giancarlo Guizzardi, the creator of OntoUML. Considering their competence in this activity, an email was sent to the NEMO group with a description of the domain (Table 1), and the construction of the respective conceptual model was requested. The constructed model was a collaboration of the three members of the group. Some e-mails were exchanged between the researchers and the specialists until a consensus was reached on the representation of the model.
For constructing the conceptual model in UML, three specialists in the language were selected, all of whom held advanced degrees in the field of software engineering and had professional industry and academic experience. The description of the domain (Table 1) was sent through e-mail to each specialist with a request to construct a conceptual model based on the description. E-mails were exchanged between the researchers and the specialists until a consensus was reached on the representation of the model.
The objective of this phase was to evaluate the expressiveness of the two conceptual models constructed by the specialists (OntoUML and UML). Twelve statements were derived from these models. Using these statements, an instrument was prepared to evaluate if the statements were more clearly represented in the conceptual model in OntoUML or UML, or whether both languages exhibited the same level of clarity. The instrument created for the evaluation is wholly included in Appendix A.
After the instrument was prepared, a profile for the participants in the evaluation was defined. Two distinct groups were selected, the first composed of eight professionals educated in the field of computing with experience in UML modeling, and the second group consisted of eighty students from undergraduate courses in the field of computing. The experiment was performed only with classes that had already completed the course on UML. Neither group (i.e., neither professionals nor students) had prior knowledge of OntoUML. The experiment was first performed with the group of professionals, a smaller and more experienced group that could validate the instrument. Suggested improvements and corrections could thus be collected for later use with the group of students. One of the improvements applied to the students was the creation of two models of the instrument. In the first model (Model 1), UML appeared as the first option in the list and this UML model appeared as Attachment 1. In the second model (Model 2) (see Appendix A), OntoUML appeared as the first option in the list and the OntoUML model appeared as Attachment 1. These were necessary to eliminate any bias related to the order in which options in the list and models were presented. Thus, Models 1 and 2 were distributed in alternation to the participants.
In this section, results are presented and discussed. First, the results concerning the construction of the conceptual models by specialists are presented; afterwards results on the evaluation of the expressiveness of the models by professionals and students.
The specialists in OntoUML revealed that the language’s richer nature generated several questions concerning the domain, even after the scope was sent. The specialists also noted that much of the information that is implied in a model must become explicit when OntoUML is used. In all rounds, specialists revealed information that should be included in the description of the scope so that creating a final model would be possible. Much of the information was implied.
In this experiment, we observed that a high degree of formality and consistency in OntoUML generated a variety of questions that perhaps would not occur with other languages. This feedback reinforces the belief that conceptual models in OntoUML may lend positive support to eliciting software requirements.
The three versions delivered differed considerably. Possible reasons for this include the lower degree of formality of the language allows for distinct representations of the same context, and the lack of semantic restrictions does not encourage questioning during construction. In the delivered versions, a representation focused on data persistence in a software program instead of on the concepts of a domain. This bias may be indicative of the lack of use of conceptual models in UML for understanding a domain. These observations should be studied in greater depth in future studies.
The version delivered by Specialist 3 was the closest to the representation of the scope. An in-person meeting was held among specialists to complete the final version presented in Fig. 3. With the two conceptual models (UML and OntoUML) constructed by the specialists, the next phase of the experiment was to evaluate the expressiveness of the models. The results are presented as follows.
First, results are presented for the pilot experiment performed with the professionals. Table 2 presents an overview of the results. Considering that eight professionals evaluated twelve statements, ninety-six choices were derived. Among these choices, twelve (13%) indicated UML the most expressive, forty indicated OntoUML (42%) the most expressive, and forty-four (46%) indicated the languages exhibited the same level of clarity.
UML OntoUML Both Total
Based on these initial results, we observed situations in which the languages exhibit the same level of clarity, and others in which OntoUML exhibits a greater level of clarity than UML. Only some situations occurred in which UML was more expressive. Considering this first result, each statement was analyzed to identify the situations in which the languages stood out. Fig. 4 presents the results for each statement. Fig. 4 shows that for Statements 1, 2, 7, 8, and 12, both languages exhibited the same level of clarity. For Statements 3, 4, 5, 6, and 11, OntoUML exhibited a greater level of clarity. Statements 9 and 10 revealed that OntoUML and both languages exhibited equal clarity. There has not been any statement in which UML had been the preferred one.
Since the professionals had more practical experience with modeling, they may have had a different viewpoint than students from the field of computing, who have not yet had considerable practical experience. Thus, the same experiment was performed with students from different computing majors to identify their perceptions relative to the expressiveness of the models.
Table 3 presents an overview of the results from the students. Eighty students evaluated twelve statements, thus totaling nine hundred and sixty choices. Among these choices, one hundred and eighty-one (19%) indicated UML the most expressive language, four hundred and six (42%) indicated OntoUML the most expressive, and three hundred and seventy-three (39%) indicated that the languages exhibited the same level of clarity. The perception of the students, despite having less knowledge about modeling, was very similar to those of the professionals. The students also identified situations in which the two languages exhibited the same level of clarity, as well as situations in which OntoUML exhibited a greater level of clarity than did UML. Only some situations occurred in which UML was indicated the most expressive.
Table 3 presents the overall results for the eighty students. However, because these are distinct groups (i.e., with different majors and class schedules) an analysis of each class was also performed. Table 4 presents these individualized results. In addition, Table 4 lists the major, the current semester of each student, and the number of participating students. All classes agreed that OntoUML was more expressive for the majority of statements. The exception was Class 3 in which OntoUML and Both got the same percentage (45%). No classes considered UML to be the most expressive overall. However, the perception of Class 1 and 5, showed a considerable difference relative to UML: 6% and 29%, respectively. In other words, Class 5 considered UML more expressive than OntoUML in at least 29% of the situations analyzed, whereas Class 1 considered UML more expressive in only 6% of the situations. This difference may be related to the extent of student knowledge of UML. However, the reasons behind their decisions cannot be determined only based on the results of this experiment.
As it happened with the professionals, the experiment with the students yielded statements in which OntoUML was the most expressive and other statements in which Both (OntoUML and UML) had same level of clarity. Thus, the results per statement were evaluated. The overall results are presented in Fig. 5, which shows that for Statements 1, 2, 7, 9, and 12, the two languages exhibited the same level of clarity. For Statements 3, 4, 5, 6, 8, 10, and 11, OntoUML exhibited a greater level of clarity. No statements were identified in which UML was most frequently selected.
These results indicate situations in which OntoUML is more expressive than UML and situations in which the two exhibit the same level of clarity. To better understand these situations, we examined statements indicating consensus that OntoUML was more expressive. The results were grouped by professionals, students (all eighty), and class. This grouping is presented in Table 5. Table 5 reveals consensus for Statements 3, 4, 5, 6, 8, 10, and 11 (in gray) in which the representation in OntoUML was the most expressive.
In OntoUML, the Role construct was used to represent the concept in Statements 4, 5, 6, 10, and 11. Specifically, OntoUML used the role construct to establish that it is relationally dependent on a universal concept, which carries the principle of identity and individuation. Representing a relationship of specialization is then required. In addition, in UML, because of the lack of semantic restrictions, the concepts for these same statements were represented by means of associative relationships, in which the origin of the concept is unclear. This finding became clearer when we evaluated Statement 9. In this statement, the perceptions of the participants were identical. Although in OntoUML, the role construct was also used, in UML a specialization was employed to represent the concept, which, based on the perceptions of the participants, resulted in the same level of expressiveness.
In Statements 3 and 8, the Relator construct was used. This construct allows the
multiplicities of a specific relationship to be expressed. In UML, associative
relationships were used. These do not allow for the expression of multiplicity in a
certain relationship. For example, in UML, representing that a relationship between the
same grantor and proxy occurs only once is not possible, even though a grantor may be
associated with various proxies and a proxy may be associated with various grantors.
We believe that if the participants knew OntoUML and the meaning of its constructs, the results would have been even more positive. One example is the representation of Statement 9. OntoUML can represent the fact that it is not sufficient for the grantor to be the representative of an organization, but that the grantor must be the representative from the same organization referenced in the proxy. In UML, only the grantor as the representative of an organization is represented, and this may not necessarily be the organization referenced in the proxy.
Conceptual models are considered crucial instruments to achieve consensus and
understanding of a domain. Thus, conceptual models are allies that support requirements
elicitation in unknown domains. However, the use of a certain language to represent the
model may undermine its expressiveness. UML is one of the most commercially popular
languages. However, according to
The objective of this study was to collect these perceptions and identify situations in which OntoUML is more expressive in an information systems context. Although our participants lacked knowledge of the constructs of OntoUML, overall it was considered more expressive than UML. In addition, various situations occurred in which consensus was reached between the participating groups that OntoUML better represents certain concepts. In addition, when conceptual models were constructed by specialists during our experiment, OntoUML was determined to have a high degree of formality and consistency. Our study showed that OntoUML causes modelers to question the situations of a domain that are not explicit. Thus, models that are more consistent and faithful to reality were built.
These results reinforce the need for a conceptual model represented in OntoUML to support software requirements elicitation. This research can evolve many different directions. One is developing a computational environment to support constructing a conceptual model in OntoUML. This conceptual model can then support the derivation of functional software requirements. Some results were present in Valaski et al. (2014)
Name: _____________________________________ Read the statements given below that was extracted from the domain electronic proxy. Analyze the corresponding representation in the conceptual model (Attachment 1 OntoUML and Attachment 2 - UML) and enter X for the model that best represents (represents most clearly) what it is being affirmed. 1. An Organization may have one or more representatives. 2. A Representative is a person that represents one or more organizations.
Option 1 – It is clearer in the OntoUML model.
Option 2 – It is clearer in the UML model.
Option 3 – Both exhibit the same level of clarity.
Option 1 – It is clearer in the OntoUML model.
Option 2 – It is clearer in the UML model.
Option 3 – Both present the same level of clarity.
Option 1 – It is clearer in the OntoUML model.
Option 2 – It is clearer in the UML model.
Option 3 – Both present the same level of clarity. 3. A Representation is a relationship established between an organization and one or more representatives. 4. Organization ICMS Database is an organization having a record in the ICMS database.
Option 1 – It is clearer in the OntoUML model.
Option 2 – It is clearer in the UML model.
Option 3 – Both present the same level of clarity. 5.Receita User is a person having a record in the Receita database.