Enterprises arise from creative human endeavours, articulated through business concepts encoded in enterprise information systems through a modular Enterprise Information Model (EIM). The EIM thus brings the productivity of computers to bear. Essentially, the EIM represents conceptual structures, which align the computer's structured way of working with the human's conceptual way of thinking. Using an industrial-strength SAP exemplar known as `Global Bike Inc.', and expressing its EIM's meta-objects as meta-object ! relation ! metaobject, Conceptual Graphs (CGs) simpli ed the EIM's modules, which consist of four business layers and two information systems layers. The logical simpli cation of these modules is extended into four levels of detail that culminate in performance indicators being assigned to each of the six layers. From the CGs, Formal Concept Analysis (FCA)'s CGtoFCA algorithm was used to generate the meta-objectarelation ! meta-object binaries that identi ed the pathways layer-wise and level-wise between the meta-objects. Through the interoperability of CGs and FCA, gaps in the conceptual structure of the EIM as highlighted by its performance indicator or measure, implying that the layer is not as modular as intended.
Enterprise information systems are nowadays mainstream to business
activity [
SAP's systems tend to be purchased and used by larger scale enterprises,
with consequently large-scale data sets. Consequently, learning SAP for
newcomers is a major task without having a reference implementation of this scale
to understand its capabilities and potential [
An opportunity therefore arises to explore the e ectiveness of this
contemporary approach using Conceptual Structures (CS), given its stated purpose of
harmonising the creativity of humans with the productivity of computers [
5 The SAP UA GBI environment consists of a set containing three parts: 1) SAP solutions (landscape), 2) GBI model company, and 3) curriculum material (including slides, exercises, case studies) enterprise, level 2 a department or organisational unit (e.g. sales) within the enterprise, level 3 the actual workplace (e.g location) in the organisational unit, and level 4 is the performance indicator and other measures used to evaluate the alignment and e ectiveness of its above 3 layers. Level 1 is thus the most conceptual and level 4 the most structural. As concepts and structures are thereby aligned as meta-objects in the matrix given by this table, it o ers an initial conceptual structure of the EIM. 3
From this starting point, Conceptual Graphs (CGs) enabled us to focus on the
meta-objects and their relations, through CGs' simple concept ! relation !
concept nature [
The Value CG depicts the each meta-object name (i.e. Vision, Mission,
Strategy, Goal) as a CG type label. To instantiate it an a particular meta-object, a
unique identi er appears in the referent eld. For example, v1V denotes that
a meta-object that is Vision (v), Level 1 (1), and V (Value layer). Likewise,
g3V for example describes Goal, Level 3, Value and so on. The [Enterprise:
@enterprise] concept follows an alternative pattern where (@enterprise) is
a CGs measure referent. The pointer to Enterprise follows that of the previous
work [
The CG and the FCL for the next layer i.e. Competency is not shown due to space restrictions. Fig. 2 shows the following Service layer CG, which nonetheless highlights similar ndings to that of Competency. It exhibits the same [Type-Label: Referent] and (relation) pattern as illustrated by Fig. 1 and its explanation above. So, for example bs1S in [Business Service: bs1S] describes bs for Business Service, 1 is Level 1, and S is Service. In this layer there is an (occurrence copy) relation too. Essentially, this relation describes two meta-objects that are synonymous, except they appear in di erent layers. For example, [Business Service: bs1S] ! (occurrence copy) ! [Business Process: bp1P]. They are therefore not co-referent, and might be described as `pseudo-synonym' meta-objects. The FCL generated by CGtoFCA for the Service layer is shown by gure 3.
Fig. 3 is the result of the CGtoFCA algorithm transforming the meta-object!
relation ! meta-object triples in the CG of Fig. 2 to meta-objectarelation!
meta-object binaries7. Like in the previous work, it thus identi ed the pathways
layer-wise and level-wise between the concepts (meta-objects) and (semantic)
relations [
An inspection of Fig. 2 shows that there are the three (occurrence copy) relations that do not point directly or indirectly through the other concepts to [Enterprise: @enterprise] in that gure. As they are not co-referent they cannot simply be removed. To do so would break the link with the associated meta-objects in the Process layer. Nor should vertical relations be (arbitrarily) added; such relations would anyway be outside of the semantic scope of this 6 We could have stated [Enterprise: GBI] but given our earlier remark about the generality of the EIM we chose @enterprise. 7 The CGtoFCA algorithm is implemented in software (http://cgfca.sourceforge.net/) and uses the CGIF le format output from CharGer (http://charger.sourceforge.net/), in which the CGs were drawn. Concept Explorer (http://conexp.sourceforge.net/) was used to generate the FCL. layer, which is Service not Process. Rather, these (occurrence copy) destination meta-objects in Process represent a messaging structure between these layers. As it stands, the layer is not a self-contained semantic unit. Although not shown the same happens for the Competency layer meta-objects. In passing it is worth remarking in ARIS that these occurrences are the same object, so in that sense the problem does goes away. But they are `pseudo-synonyms', evidenced by them not being co-referent semantically. Put simply, Process needs Service to be a semantic unit, and vice versa. The same issue applies to Competency and its same dependencies with Service.
The CGs and FCL for the remaining layers (i.e. Process, Application, Data)
are not shown once more due to brevity of space. Application and Data are
however illustrated elsewhere [
GBI, ARIS and SAP Solution Manager provide a comprehensive and detailed EIM (Enterprise Information Model), in the pursuit of conceptual structures. CGs (Conceptual Graphs) and FCA (Formal Concept Analysis) highlighted the gaps in the existing EIM. These discoveries were achieved by a focus on the layers (from Value to Data) and their levels, ranging from 1 to 4 denoting their level of detail going from the most conceptual to the most structural. Through the interoperability of CGs and FCA, gaps in the conceptual structure of the EIM were highlighted in certain layers (i.e. Competency, Service, Process and Application) and its performance indicator or other measure (i.e. level 4), implying that these layers (unlike Value and Data) may not be as modular as intended. Enterprise systems could thereby bene t from layers made up of semantic units with better-de ned interfaces (borrowing ideas from object-orientation) rather than as present. Taking this approach, enterprise systems can increase their effectiveness in contributing to the enterprise's human-driven purpose.