50 likes | 198 Views
Human Language Technologies. Issue. Corporate data stores contain mostly natural language materials. Knowledge Management systems utilize rich semantic models.
E N D
Issue • Corporate data stores contain mostly natural language materials. • Knowledge Management systems utilize rich semantic models. • It is challenging to link the natural language materials in the data stores to the semantic models in the Knowledge Management systems.
A Pair of Definitions • Semantic annotation • Process of tying semantic models and natural language together • The dynamic creation of bidirectional relationships between ontologies and unstructured/semi-structured documents • Ontology based information extraction (OBIE) • Differs from traditional information extraction through use of an ontology. • Ontology serves as a schema for the output AND as input data
Results • Authors implemented two methods of ontology based information extraction: • ML algorithm to take advantage of hierarchical class structure. • ML techniques targeted at linguistic features identified • Compared to two ML methods without use of ontologies, the OBIE approaches performed better.
CLIE and CLOnE • Authors recognized that the layman would find it difficult to create ontologies to be used for OBIE. • CLIE (Controlled Language Information Extraction) • “an application which will allow users to design, create, and manage information spaces without knowledge of complicated standards… or ontology engineering tools” • CLOnE • Sublanguage of English • Allows for conversion of natural language statements to ontology elements