LRCM ON SATURDAY, DECEMBER 14TH, 2013

Depending on the data that is stored, it is common to impose certain constraints, or semantic rules, that the database instances must obey (e.g. two different students cannot have the same id number). In various settings, we may have to handle databases that are inconsistent, i.e. databases violating the constraints.

About fifteen years ago, the principled approach of consistent query answering has been introduced in order to answer queries over such databases. Intuitively, a piece of data belongs to the consistent answer of a query if no matter how we minimally restore the consistency of the database, the data is selected by the query.

In the last few years, the complexity of querying inconsistent databases in that framework has been thoroughly investigated. An important problem is to identify classes of constraints and queries for which the problem can be solved efficiently. Another related problem is to solve the dichotomy problem for consistent query answering w.r.t. a class C of constraints and queries. Such a result is usually hard to obtain and would provide an automatic classification of the complexity of all the consistent query answering problems over the class C.

The goal of the talk is to introduce the notion of consistent query answering and to present the related  complexity issues and main results. We also give a brief overview of the dichotomy problem for consistent query answering.

As part of the requirements phase of building a software system, a conceptual model could be created which describes the information and processes of the business. This conceptual model is often expressed as a UML class diagram. In an attempt to validate the UML class diagram, stakeholders test various usage scenarios against it. The availablity of formal reasoning procedures will greatly ease finding associated inconsistencies. To date, substantial research has been done on transforming UML class diagrams into different Description Logics. In this paper we explore how Description Logics can be used to fascilitate scenario testing as a means to validate the UML class diagrams. Since identity constraints on UML class diagrams have not been mapped to a Description Logic as yet, we evaluate the suitability of using either DLRifd or SROIQ(D) for this purpose.

Traditional conceptual data models, such as EER, ORM, and UML Class dia- grams, are time-independent, meaning that they do not represent time changing data. Many applications across subject domains, such as in medicine, engineer- ing, manufacturing, administration, however, need to store time-dependent data. For example in manufacturing plants, the successful completion of a product triggers a new process of shipping, and employees evolve to play dierent roles in the organisation throughout their employment. As time flows, data evolves, which induces the need for temporal conceptual data models as a means to repre- sent information in an implementation-independent way. Research on temporal data models spans about 15 years, which focuses typically on extending EER diagrams [2-7]. However, there is no accepted standard temporal ER that has met all requirements for a temporal model [3, 8], in particular when also a solid, unambiguous, model theoretic semantics is required, except for ERVT [6] and MADS [9]. ERVT is a temporal extension of EER that uses the very expressive Description Logic DLRUS [10] to formalize the model's entity types, relation- ships, and constraints. Concerning the temporal aspects, object migration for ERV T has been formalised [6] and relationship migration was added afterward [11-13]. However, there is no formalization of a full temporalization of attributes. In conceptual data models, however, attributes play an important role, and so do temporal attributes. For instance, a regular Employee who is promoted to Man- ager may not have her Union membership set to inactive and her Salary must have a value 1million Rand, unlike regular employees, and a HIV positive outpatient will migrate to AIDS patient when her CD4 count drops below 180. The aim of our research is to give a model theoretic semantics for attributes, and temporal attributes in particular, extending ERVT and DLRUS to ensure the complete formalization of a temporal conceptual data modelling language. That is, we focus on both the logic reconstruction of atemporal attributes and all temporal aspects, i.e., the scope is comprehensiveness for modelling temporal information. While we can avail of the notions of status classes, status relations, and the evolution constraints introduced in [6, 11, 12] to have "status attributes" in a similar way, there are some variations in the details and, notably, in the logical implications, which we will present. Since DLRUS is undecidable, it does not look good for automated reasoning, but it can inform other works in extending computationally well-behaved languages with attributes [14] and it will provide insight in what is still feasible for temporal attributes in TDL-lite for ontology- based data access [15] that, so far, does not consider temporal attributes.

The paper aims at discussing the possibility to exploit the introduction of a defeasible semantics for a number of ORM2 constraints, in order to support the schema design and modelling activity in this language. The paper relies on previous theoretical results by the same authors concerning the definition of defeasible ORM2 constraints, and their interplay with their classical version. The nonmonotonic semantics introduced by the authors allows for the desing and implementation of several interesting reasoning tasks that can be offered to a potential ORM2 user, such as automatic entity types and predicates consistency checking, constraint entailment, or constraint refinement.