Transforming Semi-Structured Indigenous Dictionary into Machine-Readable Dictionary

Abstract

Creating a machine-readable dictionary for an indigenous language is not an easy process and thus, transforming an existing indigenous dictionary into a machinereadable dictionary is one approach to speed up the process. This paper presents the sequential transformation of two bilingual Sarawak indigenous dictionaries, Melanau-MukahMalay and Iban-Malay dictionaries, from their initial semistructured form into their structured representation. The transformation makes use of an OCR to convert the original PDF format of the dictionaries into HTML files, which is then analysed by the Python HTMLParser to extract only the content of the dictionaries. The extracted content is saved in plain text file. To understand the original structure of each dictionary, the textual units in the plain text file are converted into generic symbols. The observation of the collocations of the generic symbols yields to the writing of regular expressions that can delimit each dictionary element. The result is a machinereadable dictionary stored in comma-separated values format. The inspection of each column in the comma-separated values file indicates that the written regular expressions offer a good coverage of the different dictionary elements present in the studied dictionaries. Therefore, the proposed sequential transformation is efficient in accomplishing the conversion of a semi-structured indigenous dictionary into a structured machine-readable dictionary.