Domain Ontology Learning Enhanced by Optimized Relation Instance in DBpedia

Domain Ontology Learning Enhanced by Optimized Relation Instance in DBpedia Liumingjing Xiao1, Chong Ruan1, An Yang2, Junhao Zhang1, Junfeng Hu1, * 1 Key Laboratory of Computational Linguistics, Ministry of Education, Peking University, Beijing, P. R. China. 2 School of Life Sciences, Peking University, Beijing, P. R. China. {xlmj, pkurc, yangan, junhao.zhang, hujf}@pku.edu.cn Abstract Ontologies are powerful to support semantic based applications and intelligent systems. While ontology learning are challenging due to its bottleneck in handcrafting structured knowledge sources and training data. To address this difficulty, many researchers turn to ontology enrichment and population using external knowledge sources such as DBpedia. In this paper, we propose a method using DBpedia in a different manner. We utilize relation instances in DBpedia to supervise the ontology learning procedure from unstructured text, rather than populate the ontology structure as a post-processing step. We construct three language resources in areas of computer science: enriched Wikipedia concept tree, domain ontology, and gold standard from NSFC taxonomy. Experiment shows that the result of ontology learning from corpus of computer science can be improved via the relation instances extracted from DBpedia in the same field. Furthermore, making distinction between the relation instances and applying a proper weighting scheme in the learning procedure lead to even better result. Keywords: Ontology Learning; Text Mining; DBpedia * To whom all correspondence should be addressed. 1. Introduction Ontologies are one of the most important parts of seman- tic webs and attract more and more attention. Domain ontology describes knowledge of a certain domain and has been used in many applications such as QA systems. However, constructing ontologies by hand is a tedious and time-consuming work. In recent years, many ontology learning algorithms have been proposed to address this problem. These algorithms are designed to mine ontologies from unstructured, semi-structured or structured texts using statistical or linguistic methods. Clustering algorithms, which are based on the assumption that similar words have similar contexts, are typically used in statistical approach. Linden and Piitulainen (2004) used clustering to find synonyms and achieved good performance. Term subsumption is another useful technique, which employs conditional probabilities of the occurrence of terms to decide what extent the generality of a term is and thus discovers term hierarchy. Fotzo and Gallinari’s (2004) work is an exem- plar of this approach. In the linguistics settings, syntactic structure analysis and dependency analysis are utilized to discover hierarchical relations between terms. The idea of using lexico-syntactic patterns was proposed by Hearst (1998), who used some predefined patterns such as “NP such as {NP, NP … , (and | or)} NP” to obtain hypernyms. Many researchers followed this idea. To name just a few: Hippisley et al. (2005) used head-modifier principle to discover hypernym-hyponym pairs, and Sombatsrisom- boon et al. (2003) exploited frames such as “X is a/an (kind of) Y” to finish the same task. Besides these learning-from-free-text methods, some researchers start from a small, core ontology and enrich it with external data sources. Gavankar et al. (2012) pre- sented work done towards …

Content truncated to 500 words