• Information Integration

This is one of the possible Use Cases.

1. Abstract

Information integration uses cases focus on the merging of multiple sources of information to present a unified view to the user. Integrating information often involves merging information with different semantic sources (databases versus readings from physical devices versus accessing content in the form of web pages, knowledge bases), different syntactic forms (RDF, OWL, HTML, raw data streams) and across different ownership domains.

2. Status

This is an abstraction of a number of use cases that fall into the general category of information integration.

Edited by FrankMcCabe.

3. Links to Related Use Cases

• Frame-based representation, Inheritance of defaults, Reification

• Information Integration with Rules and Taxonomies

• Internet search: combining query language, rule languages and scoped negation

• Managing incomplete information

• Ontology Mapping with OWL and Rules

• Organizing a Vacation with Friends

• Rule-Based Combined Access to XML and RDF Data

• Rule-Based Intelligent Guiding

• Rule-Based Reactive Organizer

• Scoped negation, Encapsulation

• Situation Assessment and Adaptation

• SW rules for Health Care and Life Sciences

• Enterprise Information Integration

4. Relationship to OWL/RDF Compatibility

OWL and RDF compatibility are important in that information integration is fundamentally about integrating multiple sources of information, and some of those sources will be in OWL and RDF.

5. Examples of Rule Platforms Supporting this Use Case

All rule platforms?

6. Benefits of Interchange

The primary expected benefits of RIF for information integration are:

• Support the integration of heterogeneous sources of information be providing a common platform of information representation.
• Support the inter-working of information where the semantic basis of the information is different
• Leverage existing sets of information and knowledge in new contexts

7. Requirements on the RIF

• Scoping of negation and other inference rules

  • Use cases Scoped negation, Encapsulation, Internet search: combining query language, rule languages and scoped negation and Managing incomplete information highlight some of the issues involved in the semantic bases of different source of information, particularly differences in RDF/OWL and traditional logics. This strongly suggests support for multiple kinds of inference and for scoping of global inferencing rules (such as negation).

• Multiple theories and logics

  • Since information is often combined across ownership domains, it is important to be able to support rules that access multiple sources. It appears to be unlikely that all these sources of information will ever have the same logical basis; therefore, support for multiple kinds of logics appears to be important. One approach to this is suggested in the Frame-based representation, Inheritance of defaults, Reification use case.

• Compatibility with legacy information There is an increasingly large body of information that is available in 'legacy' formats -- such as XML. A RIF should be able to permit rules access such legacy information in as smooth a way as possible.

8. Breakdown

8.1. Actors and their Goals

• The primary parties are the users of information and the multiple providers of that information

• Provider offers taxonomy information represented in OWL.

• Seller offers real-world data in the form of RDF, including product information and pricing.

• Employer has eCommerce-style business logic represented as Prolog rules.

• Regulatory body has legal definitions and regulations represented in various forms.

• User uses an application that requires information provided by all of above. Also has a repository of personal information in RDF.

8.2. Main Sequence

1. User wishes to purchase goods that are potentially subject to regulatory control (medicine, weapons, sensitive information)

2. User contacts Seller with an intention of purchasing some good or service

3. Seller confirms category of good by consulting taxonomy provided by Provider

4. User consults with Employer with regard to purchasing policy for this category of goods.

5. Seller requests information from Regulatory body about possible additional requirements. These are forwarded to User.

6. User responds based on internal repository.

7. Seller confirms contract with User and, if necessary, with other stakeholders of the transaction -- such as in the case of restricted drugs.

9. Narratives

At each step in the main sequence there is the potential for requiring knowledge in one format to be integrated with knowledge in other formats. Since each party is inherently autonomous, it is not feasible to require all parties to adopt a common language or even a common semantic basis for its information.

For example, in the situation where the Seller consults with the Regulatory body about a possible application of a legal constraint. This will typically take the form of a query. The language of the query will be fundamentally determined by the Regulatory body; but the Seller must integrate information provided by the User and also by the Provider, neither of which it can control.

Of particular concern are the semantics of certain kinds of inference (such as negation) and the partitioning of the knowledge used. Proper scoping is critical to effective integration.

Of particular note here are the use cases Scoped negation, Encapsulation, Frame-based representation, Inheritance of defaults, Reification, and the work by Gabbay and others on combining multiple forms of logic in a single system.

10. Commentary

Information integration is one of the oldest uses of knowledge bases systems. It is no surprise that there are many use cases for a RIF that exemplify this. Some particular aspects that become important with a RIF are the ability to handle information whose semantic basis is itself heterogeneous (for example, an LP-based system has information that needs to be combined with an OWL-based system and a 'raw' XML-based system) and also an ability to be able to freely combine information across ownership domains (i.e., combining information belonging to more than one party).

This has implications for the kinds of inference needed and for the kinds of scoping (such as in negation) needed. This is over and above normal engineering requirements arising from combining information in a variety of syntactic forms.

11. Editorial Notes

The use case Automatically generated rules does not appear to fit with this theme.

The use case Enterprise Information Integration has not yet been considered in this abstraction, but it is certainly relevant to this category of use cases.