The comparisons of OCR tools: a conversion case in the Malaysian Hansard Corpus development / Anis Nadiah Che Abdul Rahman ...[et al.]

Optical Character Recognition (OCR) is a tool in computational technology that allows a recognition of printed characters by manipulating photoelectric devices and computer software. It runs by converting images or texts that are scanned beforehand into machine-readable and editable texts. There are a various numbers of OCR tools in the market for commercial and research use, which are obtainable for free or restrained with purchases. An OCR tool is able to enhance the accuracy of the results which as well relies on pre-processing and subdivision of algorithms. This study intends to investigate the performances of OCR tools in converting the Parliamentary Reports of Hansard Malaysia for developing the Malaysian Hansard Corpus (MHC). By comparing four OCR tools, the study has converted ten reports of Parliamentary Reports which contains a number of 62 pages to see the conversion accuracy and error rate of each conversion tool. In this study, all of the tools are manipulated to convert Adobe Portable Document Format (PDF) files into Plain Text File (txt). The objective of this study is to give an overview based on accuracy and error rate of how each OCR tools essentially works and how it can be utilized to provide assistance towards corpus building. The study indicates that each tool possesses a variety of accuracy and error rates to convert the whole documents from PDF into txt or plain text files. The study proposes that a step of corpus building can be made easier and manageable when a researcher understands the way an OCR tool works in order to choose the best OCR tool prior to the outset of the corpus development.