Referring Expression based Identification in Information Systems

点击翻译按钮获取中文摘要

Management of identity is an essential component of database and information systems: when entities, and relationships between entities, are stored in an information system, a clear way of uniquely identifying such entities must be found. Current solutions typically rely on introducing universal identifiers ("surrogates") in object-based approaches or require external value-based keys, tuples of attributes that uniquely identify objects in relational systems. Both these approaches suffer from numerous shortcomings. The former essentially requires all known individual objects to have unique distinguishing identifiers. These identifiers are often insufficient to allow human users to figure out what real-world object they refer to, since they are semantically opaque. A specific example are identifiers that individual authors or the system must invent in community-developed ontologies such as Freebase (where the identifier of the "Synchronicity" album by the Police is "/guid/9202a8c04000641f8000000002f9e349"). On the other hand, using external keys leads to problems when data from heterogeneous information systems must be integrated.******Additional problems for finding identifying attributes arise in conceptual modelling. For example, in the case where Extended Entity-Relationship modelling creates a new heterogeneous entity set by generalization, a new auxiliary and an artificial identifier is commonly required to apply uniformly to all instances, despite the fact that all the entities are already properly identified.******The goal of the proposed research is to develop a comprehensive theory for identifying entities in information systems based on the notion of a referring expression. In linguistics, a referring expression is any noun phrase identifying an object in a way that will be useful to interlocutors. In information systems, these interlocutors range from end-users issuing queries, to other computer systems communicating with the system at hand. In both cases, a common way of identifying objects is essential for such communication to make sense.******The proposal is structured around the notion of referring expressions expressed as unary formulas in first-order logic that, in the appropriate context, are true of a single individual in the universe of discourse and in this way indirectly identify such an individual. We call such formulas singular referring expressions.******The research will address many issues relating to the use of the singular referring expressions in information systems ranging from theoretical issues related to general expressive power, computational properties, and on the limits of logic-based referring expressions, to designing fragments of first-order logic with more amenable computational properties, e.g., in the area of query answering, to design methodologies for developing information systems with focus on identity management.