Discovering the semantics of user keywords An application to improve Ontology Matching Jorge Gracia
This event took place on 5th September 2007 at 11:30am (10:30 GMT)
Knowledge Media Institute, Berrill Building, The Open University, Milton Keynes, United Kingdom, MK7 6AA
Nowadays the Web is an information resource with an enormous potential. However this potential is not fully exploited by traditional search methods which not consider explicit semantics. In this talk, a system that discovers the intended meaning of a set of user keywords will be described. Firstly, the system discovers the semantics of the user keywords at run-time by harvesting the Semantic Web, obtaining a list of possible senses for each keyword. Secondly, it removes possible redundancies by using a synonymy probability measure. Finally, a disambiguation method is applied to select the most probable intended sense of each keyword according to the context. For example, it is expected that, for the keyword set "life of film stars", the meaning of "star" as "a famous actor" arises instead of its astronomical meaning. The output of this step can be used to build formal queries which represent the initial user query in a knowledge representation language.
In this new paradigm of applications that exploit the huge amount of formally specified information available on the Web, we can find another important example in the Ontology Matching field. A new method has been proposed to derive mappings from an exploration of multiple and heterogeneous online ontologies. In this talk I will describe how this method can be improved by using some techniques from the above mentioned system, in combination with the PowerMap system.
Nowadays the Web is an information resource with an enormous potential. However this potential is not fully exploited by traditional search methods which not consider explicit semantics. In this talk, a system that discovers the intended meaning of a set of user keywords will be described. Firstly, the system discovers the semantics of the user keywords at run-time by harvesting the Semantic Web, obtaining a list of possible senses for each keyword. Secondly, it removes possible redundancies by using a synonymy probability measure. Finally, a disambiguation method is applied to select the most probable intended sense of each keyword according to the context. For example, it is expected that, for the keyword set "life of film stars", the meaning of "star" as "a famous actor" arises instead of its astronomical meaning. The output of this step can be used to build formal queries which represent the initial user query in a knowledge representation language.
Click below to play the event (48 minutes)