Developing the Ontology

There are no hard and fast rules to the development of an ontology according to Noy and McGuinness (2000) but consideration must be made to the purposes of the ontology in the first place. In this instance we are primarily concerned with the structure of a given organisation so the main class will be something along the lines of employee. But to get to this conclusion I needed to approach someone who has a wider view of the organisation.

Collecting the data

I scheduled an appointment with a highly placed manager who I had discussed matters of an IT nature with before and asked if he could explain the structure of the organisation to me. This meeting took over an hour as I came to realise how complex the management structure of the trust is and continued to ask questions and seek clarification. This perhaps over-zealous inquisition helped considerably in the formulation of the following ontology and was chosen because, as Hughes (1999) points out, "In it the analyst is personally involved, and consequently gains information at first hand via direct contact with the users and clients, who also get feedback and information" (p17). That is to say that the manager I interviewed had a greater view over the structure than I did in my relatively low position within the organisation.

It would seem that my enquiry also came about while the trust itself was engaged in a period of self-examination because when I sought further data by approaching a clerical worker close to the apex of the trust I was emailed a number power-point slides in response. These slides had an unconfirmed status and so were deemed inappropriate to form the basis of the present ontology; this illustrates the point, again by Noy and McGuinness (2000), that an ontology can only express a snapshot image of a field of discourse rather than ever claim to be a definitive description which will stand the test of time. The slides must, I felt, be ratified by all parties before inclusion in an updated ontology.

From these initial discussions I produced a rough draft of how I saw the overall structure of the trust (Figure 6), I decided to produce this visual representation because as is pointed out by AbsInt (2003) "visualization provides for much better and faster understanding".

This illustration was originally produced in CorelDraw due to its ability to exporting vector illustrations to the SVG (Scalable Vector Graphics) format. SVG allows for scaling and panning by the viewer when displayed in a browser. The full listing for Figure 6 is available in the appendix.

The diagram was unsatisfactory not least because it didn't illustrate the tiers of management associated with the organisation and so was reformulated as in figure 7. This process involved looking at each level of the organisation and its relationship with the levels above and below (Please refer to the appendix for a copy of the table which formed the basis for the illustration.), while still not ideal this formulation is certainly more acceptable than the rather chaotic structure offered by figure 6. Should the tiers be analysed: the first and second tier each has one member a piece; the third has twelve; the forth has forthy-three; the fifth has seventy-three; the sixth has sixty-six and the seventh again has twelve. This figure was somewhat skewed as at times it seemed as though the disparate nature of the differing constituent organisations meant that some branches of the trust had completely different management structures, with another level of management being evident in one particular instance. I say "constituent organisations" as the trust is a relatively new entity being made up of the mental health focused segments of other NHS trusts; this means that the original "parents" of each branch had different ways of working and organising themselves.

Figure 7 is far more elongated in order to accommodate the members in each tier but it does so in a more logical and aesthetically pleasing way than figure 6. This process of modelling is problematic because the analyst has to be selective in approach while consideration must be made for possible applications of the final utility. Consciousness of at least a certain amount of cultural imperialism is imperative when modelling data associated with any organisation and this subject of bias is particularly relevant when discussing ontologies as debate rages about what is or isn't important or worthy of inclusion in a model. Indeed the relationships between entities is subject to bias, this particular situation has been discussed at length on the Protégé newsgroup with emphasis being placed on the question of the development of multi-lingual ontologies (please see the tread started by Vatant, 2003).

Defining the Classes

Once the data was collected and collated in a presentable format analysis was made of the possible classes which could apply to the data as a whole. This process is somewhat similar to that which is gone through while analysing relational data in preparation for the implementation of a relational database, not least when looking at normalising the data. There are few entities shared between the 4 branches though each branch has the same node as its head; the Area Director. Each branch also has a similar collection of Professional Leads and a similar selection of Service Managers. Between the four branches there are Adult Mental Health Services, Child and Adolescent Mental Health Services, Older People's Mental Health Services, Learning Disability Services, Substance Misuse Services, Prison In-reach Services and Ward and Day Services. These services and Service Managers are however spread throughout the Trust and there are times when the titles given to the heads of the various Services are different - again this is probably due to their initial differences and inherited from their original owners' management structures. This created substantial problems when considering avenues for normalisation and the initial Logical Data Structure Diagram (LDS) created to these ends (Figure 8) was unsatisfactory for all purposes except for, as noted above, developing a relational database. Thus the precepts of normalisation were not followed, rather an approach similar to the one used by Apitz (2002) was used when she developed the Space Shuttle Crew ontology. That is to say that each and every entity as illustrated in figure 7 would be defined as a class apart but as a subclass of the employee class.

This approach seems the most sensible and negates any confusion about the identity of the separate entities which might have occurred when looking at the Professional Leads - for instance which professional and in which area.

Thus the basic class would be that of employee, it might be argued that according to the structure the Chairman would have no employers but this isn't the case in that in the wider context of the society which the Trust plays a part all tax-payers are the employers of the Chairman, even if we should narrow the focus of the context the Chairman is directly responsible to any number of separate government bodies. The Trust under examination is merely the primary care aspect of an apparatus which has as its apex the Secretary of State for Health who oversees the work at the Department of Health who in turn appoints other bodies to oversee particular aspects of mental health care provision. Even the relationships between the seven tiers illustrated in figure 7 become almost simplistic after further examination of the wider context of health provision within the NHS with its myriad, and growing, collection of concerned parties.

Thankfully, in terms of logical progression, there were no clear instances of any entity having more than one direct manager though the relationships between the various members of the medical profession and the wider trust were open to debate with the original diagrams showing a link between Service Managers and Medics but with these relationships being annotated with "not direct reporting line". This situation is by no means new and the relationship between Doctors and the NHS has, since its inception, been loose in that Medics are at liberty to contract to undertake private consultancy should that meet certain recently introduced criteria. After consultation with a doctor it became apparent that medical staff belong to a whole different management structure parallel to the one being examined at this time.

The final leaves of figure 7 represent, to use a hospital-centric perspective, are the heads of each ward, as such they do not represent the totality of the management structure in that a ward is managed by a Ward Manager who in turn generally has two deputies and they in turn assist in the management of the other qualified members of staff; further down there is a collection of unqualified staff, who are managed by the qualified staff. Further, the ward has catering and cleaning staff and the hospital itself requires a team of people to provide care for the basic infrastructure in terms of building maintenance, etc. These diverse entities are not illustrated in the Trusts documentation and mapping their position in the larger scheme of things would be prohibitive in terms of time but they warrant mention as this situation illustrates the adage that the more one examines a thing the more one is able to describe about that thing. That isn't to say that ones description might ever hope to capture everything but merely that the description might grow exponentially in its complexity without furthering knowledge, sometimes a rock is just a rock.

Thus the employee is the initial class with all other entities inheriting its properties. These properties, in the first instance will be the post-holders name, contact number, who they are directly managed by and who they manage directly. One is conscious that this is simplistic in that it will not allow for historical perspectives in any OWL based database produced but that functionality might be incorporated in further refinements of the deliverable. That is to say that any query will show the properties of the present post-holder not previous holders of the same posts, nor will it allow for a search of previous posts held by an individual. These are all worthwhile services and worthy of inclusion in the future, if only for intellectual purposes