Lexical, structural, and semantic ambiguities in machine translation: explorations in parallel corpora

Abstract

Machine Translation, part of Natural Language Processing, gathers lots of research with the goal of automating translation by computers and enhancing its quality. Against this linguistic and technological backdrop and embracing similar goals, this research work focuses on the investigation of one of the most pervasive problems to be found in machine translation: word sense, structural and semantic ambiguities. To tackle each of these, the analysis of ambiguities of translation tokens from English to Portuguese is split into two complementary theoretical domains. In the linguistic domain, ambiguity patterns in English-Portuguese output pairs of sentence tokens from machine translation systems available on line will be studied with the help of both quality pairs of translation tokens from corpora and theoretical studies on lexical-grammars and lexical-conceptual structures. In the computational-linguistic domain, different machine translation disambiguation strategies will be reviewed to propose disambiguation heuristics.