4. PROCESOS DE ATRIBUCIÓN Y TEORÍAS DE LA
4.1. Procesos de atribución
4.1.4. La respuesta emocional y actitudinal a las noticias sobre terrorismo
Whilst it is not possible to quantifiably measure the quality of data captured by the system, one of the important theoretical benefits of the system is producing a set of captured data, that is consistent with a defined domain ontology, thus in principle supporting a high quality of said data. To do this the specialised domain model produced by the specialisation process must remain consistent with the domain ontology. The system has been successfully used to produce a number of specialised domain models that are consistent with the underlying ontology.
A number of constraints on user behaviour are used to ensure this consistency.
6.5.1.1 Use of ontology relationships and terms
The descriptive elements that users must use for the specialised domain model are based on defined descriptive terms from the ontology. No new terms can be added to the specialised domain model by users and existing terms cannot have their underlying ontology-based term definitions altered. Primarily users select from elements based on terms and explicit relationships from the ontology. However users are able to add some new relationships to cover such areas as universal description objects which the ontology permits but does has not have an explicit relationship for.
Users cannot create new relationships between terms that are not allowed for in the underlying ontology. This constrains users from moving description objects to different places on the description object hierarchy (other than sibling order changes) as this would involve creating new description object to description object primary organising relationships. These relationships cannot be changed as they may be fundamental to domain understanding of the description object. For example a ‘branch’ as part of a ‘stem’ is different to ‘branch’ as part of an ‘inflorescence’, both of
these are supported by the ontology and could be in the specialised domain model but ‘branch’ as part of a ‘leaf’ is not.
Users are equally constrained from adding attributes for description objects that are not supported in the ontology. Adding such a new a description object – attribute relationship could lead to applying such an attribute in a context that the ontology designer never intended which could lead to misuse.
Value domain relationships (attribute – value object relationships) are used to constrain users from using value objects for purposes and in contexts for which they are not intended.
The constraints to stop using the various descriptive elements in contexts that the ontology builder did not intend is important, as the definitions attached to the terms can then be specific to their possible usage context. This allows them to be more accurate, without making the definitions themselves directly reliant on other definitions. It was important that the definitions should be as self-contained as possible, avoiding a highly elaborate definition space as that would more difficult to source than an ontology that did not require such a rich information structure.
6.5.1.2 Ontology-based logical consequences
Logical consequences of user actions upon the domain model data structures are enforced by the system to maintain consistency with the ontology. This constraint ensures that when an attribute is included in the specialised domain model that the parent description object is also added. Equally when a description object is included, that a path up to the root of the description object exists within the specialised domain model, including other description objects as necessary to ensure this. Including a value object of an attribute provokes similar behaviour to ensure the attribute is included in the specialised domain model.
Logical consequences also flow down the hierarchical data structures, ensuring that if a description object is removed from the specialised domain model that no attributes of it nor descendent description objects remain in the specialised domain model. Similar behaviour is enforced for removing attributes. This behaviour must be
enforced as the description objects, attributes and value objects do not contain sufficient data in themselves to give their contextual relationships to other objects in the ontology.
Whilst users would be unlikely to wish to perform operations that would result in nonsense data, these constraints remain necessary. The system aims for domain users to edit the specialised domain model. Such users however may lack the expertise or knowledge to understand the details of the underlying data structures or the effects of their editing decisions upon the structures.
6.5.1.3 Users and consistency
During the user tests, users did exhibit behaviour that attempted to perform operations that would break consistency.
Users exhibited a wish to add new description elements based on new terms to the specialised domain model. Some users indicated such a desire to use their own favoured terms in early interviews, but the majority disagreed or accepted the value of only using those from an accepted common ontology (especially if it was their view over what terms to use, that was the accepted view). During later wide tests, a minority of users (15% during the final wide test) made some comment during speak-aloud observation, indicating a wish to add terms when they found they could not immediately find terms they wanted. This was sometimes due to failures of omission in the ontology (especially in earlier tests with an early version of the angiosperm ontology) and sometimes due to failure to find the desired term. In the latter case, being unable to add a new term, usually led the user to find existing equivalent terms that existed in the ontology. Enforcing this constraint does however, make the effectiveness of the system more reliant on the completeness of the underlying ontology being used. Enforcing it also ensures that spurious non-comparable terms are not used where equivalent terms are already available.
Users (7% in final wide test) also attempted to use the facility for changing the order of sibling description object nodes to actually move description objects to different places on the description object hierarchy. In doing so, they were attempting to place the description objects in a context, which better matched their own view of the data as
opposed to the ontology builders. Again this was most evident in early tests, with an early version of the ontology, where weaknesses of reliance on the ontology were being exposed.