Refining non-taxonomic relation labels with external structured data to support ontology learning

Abstract

This paper presents a method to integrate external knowledge sources such as DBpedia and OpenCyc into an ontology learning system that automatically suggests labels for unknown relations in domain ontologies based on large corpora of unstructured text. The method extracts and aggregates verb vectors from semantic relations identified in the corpus. It composes a knowledge base which consists of (i) verb centroids for known relations between domain concepts, (ii) mappings between concept pairs and the types of known relations, and (iii) ontological knowledge retrieved from external sources. Applying semantic inference and validation to this knowledge base yields a refined relation label suggestion. A formal evaluation compares the accuracy and average ranking precision of this hybrid method with the performance of methods that solely rely on corpus data and those that are only based on reasoning and external data sources.

Confidence Management for Learning Ontologies from Dynamic Web Sources

“…The underlying ontology learning components have been thoroughly evaluated and published in (Liu et al, 2005), (Weichselbraun et al, 2010a) and (Weichselbraun et al, 2010b). The framework presented in this paper extends this work by introducing dynamic data structures for confidence management, the source impact vector (SIV) and through mechanisms for its adaptation according to user feedback, as we detail next.…”

“…Finally, relation type detection with techniques to identify taxonomic relations as described in (Liu et al, 2005) and methods to label non-taxonomic relations (Weichselbraun et al, 2010b) conclude the learning process.…”

Computing Semantic Association: Comparing Spreading Activation and Spectral Association for Ontology Learning

“…The original system [14] learns from domain text only and already includes spreading activation as the major building block to integrate evidence. Weichselbraun et al evaluate information from social media as additional evidence source [16] and present novel methods for learning non-taxonomic relations [17]. Finally, Wohlgenannt et al discusses data structures and algorithms for the fine-grained optimization of the system from feedback collected with games with a purpose [19].…”

Knowledge capture from multiple online sources with the extensible web retrieval toolkit (eWRT)

“…Developed as part of the DIVINE research project [18], the structured data acquisition component has been applied to several domains including climate change [13] and tourism [7].…”