RISO - TT - extração de expressões temporais em textos.

Abstract

The necessity of managing the large amount of digital existing documents nowadays, associated to the human inability to analyze ali this information in a fast manner, led to a growth of research in the area of system development for automation of the information management process. Nevertheless, this is not a trivial task. Most of the available documents do not have a standardized structure, hindering the development of computational schemes that can automate the analysis of information, thus requiring jobs of information conversion from natural language to structured information. For such, syntactic, temporal and spatial pattern recognition tasks are needed. Concerning the present study, the main objective is to create an advanced temporal pattern recognition mechanism. We created, heurístically, a rules dictionary of temporal patterns, developing a module in an extendable and flexible architecture for retrieval and marking. This module, called RISO-TT, implements this pattern recognition mechanism and is part of the RISO project (Retrieval of Semantic Information from Textual Objects). Two experiments were carried out in order to evaluate the efficiency of this approach. The first one was intended to verify the extendability and flexibility of the RISO-TT architecture and the second one to analyze the efficiency of the proposed approach, based on a comparison between the developed module and two Consolidated tools in the academic community (Heideltime and SuTime). RISO-TT outperformed the rivais in the temporal expression marking process, which was proved through statistical tests.