Semantic Query Expansion (tool SQEx)

Searching for concepts within the ontology using heuristic inferences and logical deductions.

Institution: Slovak University of Technology
Technologies used: Java, Protege OWL API, Jena2 OWL Library
Inputs: Ontologies, Query terms in natural language, Parameters for execution control
Outputs: Concepts from the ontology, expanded query terms

Addressed Problems

Reasoning and semantic ontology searching is one of the main benefits of the Semantic Web. Ontologies allow for intelligent searching within concepts. But, users who want to search the ontologies may not be aware of neither concepts in the ontology nor relationships among them.  Explicit conceptualization of a domain which contains all the relationships between concepts brings the possibility of expanding the user’s query along these relations. However, ontologies could be inconsistent or could not contain every logical relationship. Various techniques to overcome this problem are known but there is not any algorithm which combines them together. Therefore we proposed a method for semantic ontology searching which uses reasoning techniques to evaluate user’s query. The method does not depend on the ontology and overcomes various types of possible inconsistencies in the ontology.

Description

The method combines the approaches of logical reasoning on OWL-DL, heuristic inferences and lexical analysis of the users query to find related concepts in the ontology and evaluate their relatedness.

Logical reasoning is formal semantic reasoning based on the explicitly defined relations between concepts in the ontology. The main logic reasoning expansions are expansion to equivalent concepts, expansion to broader or narrower concepts and expansion the concepts with common superclass.

Heuristic inferences are used to find related concepts which relatedness is not explicitly modeled e.g. due to inconsistent ontology.

Lexical query analysis is about finding linguistic relations between concepts. This includes the word normalization (removing morphological endings), stop-words removal and finding lexically related words as synonyms. The latter step is performed by cooperating with external tool named Morphonary.

The method for expansion of the user’s query defines number of consequent steps and orders described expansion techniques in one sequence:

  1. Initialization phase(ontology selection - the user can specify ontology URL, path of the filesystem or select one of the cached ontologies,  configuration of the parameters to control the expansion execution, separation of the query to individual terms,  application of the heuristic “Acronym expansion”,  lexical expansion of the terms (normalization, synonyms, etc.),  finding the terms in the concept names and datatype properties using fulltext search engine,  creation of the initial set of concepts.
  2. Logical expansion phase ( expansion to equivalent concepts,  lexical expansion of new concepts (synonyms, etc.),  finding the new terms in the concept names and datatype properties using fulltext search engine, expansion of classes with their instances and subclasses,  expansion of the instances with their classes, expansion the concepts with common superclass.
  3. Heuristic expansion phase (expansion to concepts with the same name regardless of the namespace and uppercase and lowercase letters usage,  expansion to the concepts with the same name regardless of formatting, expansion to the concepts which names are prefixes to the names of original concepts,  expansion to the concepts which names include original concepts as prefix, expansion to the concepts which contain a sequence of at least  4 characters from the original concepts, expansion to the classes containing concept names.
  4. Iteration to the step 2 After every iteration linear sorted result set is provided to the user. The user decides which concepts are relevant to him and launches new iteration if needed.

The method has been evaluated on number of ontologies found online with the semantic web search engine Swoogle (swoogle.umbc.edu) but as the primary test ontology an ontology developed within our research project has been used.

The idea of query expansion is suitable for rather complex ontologies with many classes, instances and relations between them. Iterative usage of the expansion (expansion of the expansion) causes an interesting effect of gradual ontology discovery. That way, the user can navigate himself in a complex ontology and can thus find very quickly what is relevant to him.

Sqex architecture

Overall architecture

The figure shows architecture of the method. The user enters a query and parameters through the query interface. Then it is processed by the term separator after which the individual terms could be expanded with the heuristic "Acronym expansion". The resulting terms are then normalized and looked up in the ontology by the user - a fulltext search engine. Initial set of concepts is formed and the above mentioned method is applied.

References

  1. Malecka, Juraj - Rozinajová, Viera: An Approach to Semantic Query Expansion. In: Tools for Acguisition, Organisation and Presenting of Information and Knowledge : Research Project Workshop Bystrá dolina, Nízke Tatry, Slovakia, September 29-30, 2006, Proceedings. - Bratislava : STU, 2006. - ISBN 80-227-2468-8. - pp.148-153