|
Descripción
|
|
|---|---|
| Modern search engines such as Google or Yahoo have long come into our everyday life and we hardly imagine how we could do without them. Nevertheless, however useful these applications may be, they are rightfully reproached for ?not understanding? the texts they are dealing with. They find far too many texts, while the overwhelmingly most part of them has nothing to do with what the user is asking about. On the other hand, if a text conveys the relevant meaning but it is expressed by words different from the ones used in the user?s query, this text will hardly be found at all. For many natural language applications, first of all, for Information Retrieval and Extraction as well as for Question Answering, it is essential that they should be able to discover semantic similarity between the texts if they express the meaning in different ways. Cf. synonymous sentences (1) ? (3): (1) Real Madrid and Barcelona will meet in the semi-finals on Thursday. (2) The semi-final match between Real Madrid and Barcelona will take place on Thursday. (3) The adversary of Real Madrid in the semi-finals on Thursday will be Barcelona. If we wish to extract the meaning from the text irrespective of the way it is conveyed, we should construct a semantic analyzer capable of producing identical semantic structures for sentences (1) ? (3), or at least semantic structures whose equivalence can be demonstrated. To account for the equivalence (1) ? (3), one needs to formalize linguistic knowledge. The problem becomes much more difficult if text understanding includes access to language-external world knowledge. For example, sentence (4) describes the same situation as (1) ? (3) and, ideally, all four sentences should be returned as the answer to the same questions. (4) The semi-finals on Thursday will see the winner of the UEFA Champions League 2010-2011 and the team of José Mourinho. To be able to discover the equivalence (4) <=> (1)-(3), the system should know that it was the football club Barcelona who won the UEFA Champions League in 2010-2011, and that José Mourinho is the coach of Real Madrid. This implies that linguistic knowledge should be linked with language-external information. The creation of a semantic analyzer of this type requires a powerful linguistic processor capable of building coherent semantic structures, a knowledge-extensive lexicon, which contains different types of lexical information, an ontology, which describes objects in the domain and their properties, a repository of ground-level facts (such as ?Coach of Real Madrid in 2012 : José Mourinho?), and an inference engine capable of manipulating all these data. The talk is devoted to one of the ways these problems could be solved. I will show how the text is analyzed, how its meaning is represented and how the interface between linguistic and world knowledge is established. | |
|
Internacional
|
Si |
|
Nombre congreso
|
Grand Seminaire LINA (Univ. Nantes - France) |
|
Entidad organizadora
|
(Univ. Nantes - France) |
|
Nacionalidad Entidad
|
FRANCIA |
|
Lugar/Ciudad de impartición
|
Nantes |
|
Fecha inicio
|
22/11/2012 |
|
Fecha fin
|
22/11/2012 |