A Supervisioned Approach for Semantic Resolution in Portuguese with Neighborhood Analysis Integrated to the Falibras Translator

Abstract

Deafness makes social interaction considerably more difficult since it inhibits the individual from communicating through the oral-auditory pathway. To facilitate communication between deaf people and listeners, machine translation tools to sign languages can be used. However, according to reports in the literature, about 75% of the deaf community feels dissatisfied with the translation of the existing tools. Among the main causes of this dissatisfaction is the use of inappropriate signs for words with semantic ambiguity. In this work, we propose the addition of a new semantic module to the translation process of the Falibras System, with the objective of improving the quality of the translation with respect to semantic disambiguation of the input text (in Portuguese). The main objectives of the proposed approach are: (1) to know the state of the art regarding the resolution of ambiguities in Portuguese; and (2) to combine existing techniques to develop an ambiguity resolution module for Falibras. The evaluation of the results showed a gain in accuracy, in a comparison with existing work in the literature. The evaluation was conducted based on the Goal-Question-Metric model.