Parallel Annotation and Population: A Cross-Language Experience
Bahareh Sarrafzadeh, Mehrnoush Shamsfard · 2009
In recent years automatic ontology population (OP) from texts has emerged as a new field of application for knowledge acquisition techniques. In OP, the instances of an ontology classes will be extracted from text and added under the ontology concepts. On the other hand, semantic annotation which is a key task in moving toward semantic Web tries to tag instance data in a text by their corresponding ontology classes; so the ontology population activity accompanies generating semantic annotations usually. In this paper we introduce a cross-lingual population/annotation system called POPTA which annotates Persian texts according to an English lexicalized ontology and populates the English ontology according to the input Persian texts. It exploits a hybrid approach, a combination of statistical and pattern-based methods as well as techniques founded on the Web and search engines and a novel method of resolving translation ambiguities. POPTA also uses Wikipedia as a vast natural language encyclopedia to extract new instances to populate the input ontology.