Learning Ontologies by Processing Natural Language Text
This event took place on Wednesday 03 November 2004 at 12:30
Dileep Damle KMi, The Open University
Ontology construction is a costly and slow process requiring ontological engineering skills as well as domain expertise. The semantic web will be greatly facilitated if domain ontologies could be built quickly and cheaply without human expertise of either kind.
One approach to automatic ontology creation is to utilize existing knowledge resources such as database schema. These can be very useful, but are rare and natural language text is more likely to be available for many domains. But complete, unambiguous and accurate interpretation of natural language texts by computers is currently a very major challenge This work is concerned with extracting ontological elements such as concepts, their properties and inter-relationships from natural language corpora in order to grow an ontology for the domain in a semi-automatic way. The hypothesis is that it is not necessary to fully, and unambiguously interpret all sentences in text, but inferences drawn from parts of sentences may be sufficient for the purpose if enough text is available.
The presentation will cover some early results in two of the main areas of this research.
1. Identification of the domain relevant terms in the corpus
2. Some early examples of semantic extractors
This event took place on Wednesday 03 November 2004 at 12:30
Dileep Damle KMi, The Open University
Ontology construction is a costly and slow process requiring ontological engineering skills as well as domain expertise. The semantic web will be greatly facilitated if domain ontologies could be built quickly and cheaply without human expertise of either kind.
One approach to automatic ontology creation is to utilize existing knowledge resources such as database schema. These can be very useful, but are rare and natural language text is more likely to be available for many domains. But complete, unambiguous and accurate interpretation of natural language texts by computers is currently a very major challenge This work is concerned with extracting ontological elements such as concepts, their properties and inter-relationships from natural language corpora in order to grow an ontology for the domain in a semi-automatic way. The hypothesis is that it is not necessary to fully, and unambiguously interpret all sentences in text, but inferences drawn from parts of sentences may be sufficient for the purpose if enough text is available.
The presentation will cover some early results in two of the main areas of this research.
1. Identification of the domain relevant terms in the corpus
2. Some early examples of semantic extractors
