Modeling, Discovering, and Exploiting Complex Semantic Relationships

1. Modeling, Discovering, and Exploiting Complex Semantic Relationships Amit Sheth, I.Budak Arpinar and Vipul Kashyap

2. Identification, discovery, validation and utilization of relationships- critical on the Semantic Web • Types of Semantic relationships 1. Using predefined multi-ontology relationship 2. Relevancy ranked indirect relationships 3. User-defined relationships • Challenges in finding semantic relationships 1. Each document might describe many entities 2. Number of relationships in the KB is very large.

3. Taxonomy of Relationships based on Information Content • Content Independent Relationships • Content Dependent Relationships 1. Direct Content Dependent Relationships 2. Content Descriptive Relationships - Direct Semantic Relationships - Complex Transitive Relationships - Inter-domain Multi-ontology Relationships - Semantic Proximity Relationships

4. Representation of Relationships A fundamental representation between two concepts is a mathematical structure denoting it as a mapping between the instances belonging to the two concepts. These mappings can be characterized along following dimensions. • Arity • Cardinality • Direct vs Transitive • Crisp vs Fuzzy • Properties vs Relations • Structural Composition

5. Computation and Exploitation of Relationships Four main computations • Identify • Discover • Validate • Evaluate

6. SCORESemantic Content Organization and Retrieval Engine • Ontology with Definitional Component and Assertional Component. • Using relevant ontology, domain specific metadata can be extracted from a document, thus enhancing its meaning.

7. Semantic Document Enhancement in SCORE system

8. An Example Ontology and Knowledge Base

9. Ways to improve Efficiency of the Semantic Association Discovery • Assigning more weights to certain entities • Specification of Relevant Context • Ranking relations

10. Knowledge Base (KB) • Contains “Entities”(name and a classication type) and “Relations”(name and a vector of classification types) • Entity Classification Hierarchy – similarities among the entity classification • Relationship Hierarchy – similarities among existing relationships

11. Types of Semantic Queries • Keyword Queries • Entity Queries • Relationships Queries • Path Queries • Path Discovery Queries

12. Semantic Index (SI) • Constitutes a foundation for the design of a suitable semantic query engine.

13. Rho (ρ) operator • It is an approach for computing complex semantic relations. • Intended to facilitate complex path navigation of metadata as well as schema/taxonomies in KBs. • Specifically it provides the mechanism for reasoning about semantic associations that exist in KBs.

14. Binary form of the operator is ρT(a,b)[C,K] where C= context given by user K = constraints that includes user associations to a specific region in the KB. There are 4 types of the ρ operator: PATH, INTERSECT, CONNECT and ISO • ρPATH(a,b): Given the entities a and b, looks for directed paths from a to b and returns a subset of possible paths.

15. Human Assisted Knowledge Discovery (HAND) • Users are able to pose questions that involve exploring complex hypothetical relationships amongst concepts within and across domains, in order to gain a better understanding of their domains of study, and the interactions between them. • Could include complex information requests involving user defined functions and fuzzy or approximate match of objects thus requiring richer environment of expressiveness and computation.

16. Eg: Does Nuclear Testing cause Earthquakes? • Correlation of data from different domains like Natural Disasters, Nuclear Testing • Meaning of “cause” should be clearly understood. • Refining relations and posing other questions based on the results presented may lead to better understanding of the nature of interactions between two events.

17. Information Scapes (Iscapes) • “A computing paradigm that allows users to query and analyze the data available from diverse autonomous sources, gain better understanding of the domains and their interactions as well as discover and study relationships.” • An Iscape is defined in terms of relevant ontologies, inter-ontological relationships and operations. • InfoQuilt uses Iscapes. Supports user defined operations. • Eg: Find all earthquakes with epicenter in a 5000 miles radius area of the locations at latitude 60.7 North and longitude 97.5 East.

18. Evaluations Involving Semantic Relationships • A user query formulated using terms in a domain ontology is translated by using terms of other domain ontologies. • Substitution of terms by traversing inter-ontological relationships like synonyms, hypernyms or hyponyms. • When a query is posed : 1. The user browses the available ontologies and chooses a user ontology that includes terms needed to express the semantics of the query. 2. If the user is not satisfied with the answer, the system retrieves more data from other ontologies to enrich the answer. 3. In doing so a ‘target ontology’ is created. 4. Full/partial translation.

19. Drawbacks • Semantics of the query may change. • “loss of information” • Can estimate the “loss of information” and set a threshold.

20. Conclusion • Ontologies provide the semantic underpinning, while relationships are the backbone for semantics in the Semantic Web. • Attention needs to shift from searching relevant documents to an approach of exploiting data with knowledge.