PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Inna Novalija and Dunja Mladenic
In: SiKDD 2008, 17 Oct 2008, Ljubljana, Slovenia.


Ontologies are commonly used for annotating textual data mainly based on human language technologies [1]. This research focuses on manual extensions of ontologies to support the annotation of business news. Experiments were conducted on a well known Cyc ontology and using Cyc annotator on two business news datasets. We show that the proposed extensions of ontology results in annotation with better coverage of terms that are relevant for the business domain. The results of identifying financial terms in business news using the original Cyc ontology show the average precision of 56% and recall of 41% in case of Reuters news and the average precision of 69% and the recall of 57% in case of Yahoo financial news. Using the proposed extension results with increased performance, the average precision of 82% and average recall of 73% for Yahoo financial news and average precision of 84% and average recall of 63% for Reuters news.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Natural Language Processing
ID Code:4973
Deposited By:Jan Rupnik
Deposited On:24 March 2009