Creation of a class hierarchy (of knowledge)

Since the 60-70s that saw the birth of NLP, one of the largest issues has been grasping the meaning of a document automatically. Most systems intended for these tasks have shown themselves to be ill adapted due to the overly generic vision of automatic understanding. Research has thus been focused more on the issue of information extraction making semantic categories like named entities or time markers the pivot of their work. If spotting these elements in and of itself does not allow a text to be understood, it does, however, allow the primary elements of meaning to be extracted to create conceptual specifications that are very useful for ISs [ZAC 07].

Ontologies are a means of modeling a domain vocabulary [GRU 93] by identifying the terms and relationships that they contain. Applied to Artificial Intelligence (AI), ontologies not only allow for the fine extraction of date, but also for the proposition of systems based on developed knowledge [BUI 05]. For our work, we will situate ourselves in the first step before ontology creation, namely realizing a taxonomy and developing a class hierarchy, without concerning ourselves with the description of the attributes and relationships between these classes [BIE 05]. This aspect will be completed in subsequent steps with the semantic relationships that morphosyntactic structures (S, NP, SP, VP, etc.) can preserve upon being extracted from the contents.

< Prev   CONTENTS   Source   Next >