Desarrollo de un corpus paralelo de texto e imágenes en el idioma japonés para futuros trabajos de investigación en procesamiento de lenguaje natural : creación de un corpus paralelo para el idioma japonés.

Abstract

The presented work aims at creating a parallel corpus of Japanese texts. This corpus starts with a set of files that go through an initial cleaning that produces the reference corpus. Based on this corpus, a parallel corpus is obtained with the insertion of spelling errors. These errors are simulated by what is known as a noisy channel. An OCR system is used to simulate this noisy channel. The errors of this OCR system represent the noisy channel. Following the error production process, an alignment is performed in order to have a localization of the produced errors.