CAPÍTULO II: EL PROCESO DE DESARROLLO E INTEGRACIÓN FRONTERIZA
2. Zona fronteriza peruano-chilena
2.3 Distrito de Tacna y Comuna de Arica
Authors in [Noy & Musen, 2002] developed Promptdiff to compare different versions of an ontology. It detects changes in two versions of an ontology and presents the differences. At the end of the evolution process, ontology editors use promptdiff to review changes and approve or reject those changes. Currently, promptdiff does not support OWL2 on- tologies. However, there are different successors of promptdiff [Tudorache et al., 2008] [Redmond et al., 2008] that use the heuristics used in promptdiff. [Redmond et al., 2008] suggest a system that manages changes using version control systems. The authors propose a system which addresses the existing problems of ontology version control systems. This includes addressing problems in concurrent editing, complete change tracking, scalability, and performance. They focus on add, delete and rename operations and perform analysis using diffs between two ontology versions.
The authors [Redmond & Noy, 2011] present a pluggable difference engine which aligns ontology entities before conducting comparison. The difference engine uses an alignment phase and explanation phase. The explanation phase organizes the output of the alignment phase and presents the difference in a human understandable and organized way. The differ- ence engine highlights additions, removals and renaming of entities. This approach requires two versions to compare changes. It does not consider the change operations that are the sources of the change.
The authors [Ruiz et al., 2009] propose content CVS (Concurrent Versioning System) for building and editing ontologies collaboratively. They use a CVS paradigm used in
software engineering to build ontologies and manage changes. In content CVS the most recent version of the ontology is kept in a shared repository in a server and each developer keeps a local copy. Whenever the developer makes a change to the local copy, he/she has to submit the latest local version to the server. The system compares the request with the most recent version. The developer can access the repository using export, check-out, update and commit operations. If the local version of the ontology is not changed, it means there is no meaningful change committed. Otherwise, if the local version is up-to-date and not in conflict with the recent version, the local version will replace the recent version.
This approach uses change detection, conflict detection and conflict resolution. It uses structural conflict resolution and a combination of structural and semantic conflict resolu- tion. The authors implement deductive difference which computes the logical consequence of the new version with the previous version to identify semantic differences. It uses rea- soners to conduct deductive reasoning and semantic conflict refers to the conflict due to inferred axioms. Once the difference is calculated, if there are conflicting axioms or unin- tended entailments, the users are presented zero or more options to choose. The content CVS allows the user to choose the most suitable minimal plan to avoid the conflicts. If there is no plan, the conflict resolution process ends and the ontology rolls back to the old version.
A closely related work on concurrent development and editing of ontologies is given in [Ruiz et al., 2011]. This work extends content CVS to incorporate several developers to make changes concurrently. This work focuses on conflict detection among change requests from different developers, and resolving the conflicts by employing structural and semantic differences. Semantic conflicts are addressed using logical reasoners.
The authors [Hartung et al., 2012] propose a tool that allows determining semantic changes between two versions of an ontology. A web-based tool, CODEX (COmplex Ontology Diff EXplorer), is proposed. The tool contains a repository for calculating diffs at the backend. The backend computes diffs and presents the changes using statistical measures. This in- cludes: number of changes, diff sizes and growth rates of the changes. It allows exploration of elements that have been influenced by the changes. It further includes change impact
analysis to find out the elements that are affected.
CODEX can provide information similar to our change impact analysis tool. However, it follows a similar approach used in diff and in content CVS. Even if we do not follow the ubiquitous diff approach, our change impact analysis tool provides rich analysis and additional semantic impacts other than the semantic changes presented in CODEX. Our approach not only focuses on terminologies, but also analyses impacts on instances and annotation triples. Change impact analysis deals with impacts of changes on annotated documents. It presents impacts of changes on ontology entities and information sources that consume the ontologies.
The authors in [Konev et al., 2012] propose a new version of CEX versioning tool which extends the original CEX [Konev et al., 2008] to incorporate three distinct logical differ- ences. These are: concept inclusion, answers to instance query and answers to conjunctive query. CEX is applicable for acyclicEL terminologies and the proposed version extends it to ELH+ which admits role inclusion, range and domain restrictions. This enables users to perform concept diff, instance diff and query dif. This work is close to our work by considering semantic changes on instances (ABox Statements).
In the above approaches, changes are made concurrently and two or more versions of ontologies are compared structurally and/or semantically. Semantic difference focuses on the logical difference of axioms based on inputs from a reasoner. Our approach is different in the following ways. We view impacts from change operations perspectives. First, we focus on the impacts of the change operations that are requested by the user and generated by the system. Second, our notion of structural and semantic impacts is broader than the structural and semantic changes discussed in the above papers. We further incorporate the implication and interpretation of the changes. Third, a minimal plan, in content CVS, refers to a plan that avoids inconsistencies and errors caused by arbitrary entailments with minimum removal of additions or deletions.
Our approach provides detailed information about the requested and the derived change operations, the impacts of the change operations, the affected entities due to a given change operation and the severity of the impact on the entities in the system. Our concern is not only
finding the additions and deletions, but also how the entities are impacted, which change operations impact them, how two or more changes impact an entity and the severity of the impacts.
In comparison to the CVS approach, a CVS presents “what” has changed, but the details about how the changes affect the dependent entities, why a given entity is affected and the severity of the effect is missing. Evolution of ontologies using such analysis as an input for selecting an optimal strategy and evolving ontologies is the major concern of this research. At this stage, this research does not analyse impacts of changes on inferred axioms.