Desarrollo de un corpus paralelo de texto e imágenes en el idioma japonés para futuros trabajos de investigación en procesamiento de lenguaje natural japonés obtenido a través de un corpus referencial y un corpus ocr: generación de un corpus paralelo en el idioma japonés obtenido a través de un corpus referencial y un corpus ocr.

Abstract

The present work consists of the investigation on the process to generate a parallel corpus given that data were received in plain text files in Japanese language, the project consists of the cleaning of the data to generate a referential corpus, after that it is sought to pass the data through an OCR tool, in order to generate misspellings due to the process that the OCR tool implements internally. Once the two files, the reference files and those generated by the OCR tool with spelling errors, are normalized both files in order to compare them with the diff tool to determine the similarity that exists between each reference file and its respective OCR file. When the expected similarity percentage is reached, the next part is to align the reference files and the OCR files at line level and this will form the result which is the parallel corpus.