OntoQuery: Ontology-based Querying of Texts

This paper addresses techniques for extracting conceptual descriptions from text sources and queries based on an ontology. The taxonomic ontology comprises featurestructured conceptual descriptions, and retrieval is based on description similarity relative to the ontology. Introduction The aim of the OntoQuery project (Andreasen, Nilsson, and Thomsen 2000; OntoQuery) is to contribute to the development of general solutions to the querying of text databases and to the extraction of conceptual descriptions from such databases through limited computational natural language understanding. More precisely, the project addresses content-based retrieval and access to text sources, such as online document databases and encyclopaedias. Methodological Aims The key methodological goal of the project is to create a coherent framework for ontological representation of domains, ontological semantics for pertinent natural language phrases, and ontology-based search in text databases. Coherence is established through a formal and computational language of conceptual descriptions uniting 1. formal ontologies 2. lexicons and lexical semantics 3. ontology-structured semantic domains for nominal phrases 4. querying and search For purposes of validation and demonstration, the theoretical results are examined in prototypes with accompanying tools and resources on selected real world domains. Formal Ontology and Semantic Lexicon Algebraic lattices have been chosen as our theoretical basis, serving as skeleton ontologies shaped by the conceptual inclusion relation. These ontological structures, spanned by general categories such as physical objects, substances, and events, transcend hierarchical classifications by admission of crosscategories. They can be visualized by diagrams in a wellknown fashion. In the logical representation language OntoLog (Nilsson 2001a), the nodes of such skeleton ontologies are adorned with attributes giving rise to compound conceptual descriptions intended for the representation of the conceptual content of noun phrases (NP’s). Descriptions reflect a compositional, phrase-structured semantics where connection is established with lexical semantics as applied in the SIMPLE European initiative, see e.g. (Lenci et al. 2000), (Pedersen and Keson 1999) and (Pedersen and Nimb 2000), which adopts principles from the generative lexicon approach (Pustejovsky 1995). The lexical entries comprise a four-dimensional inheritance structure using the qualia roles: formal, agentive, telic, and constitutive. From: AAAI Technical Report SS-02-09. Compilation copyright © 2002, AAAI (www.aaai.org). All rights reserved.