Development of an English to Maninka Machine Translation System

This study investigated the process underlying the translation of English text to Maninka text; designed a computational model for the process; implemented the designed model; and evaluated the developed system. These efforts aimed to develop an English to Maninka Machine Translation system. Guided by the theories of syntax, information on Maninka syntax was gathered from textbooks and journals. Maninka lexical items and their parts of speech were collected from a Maninka glossary named Malidaba. The collected data were digitized and stored in SQLite. Based on the collected information and data, a computational model was formulated using context-free grammar tools in Automata Theory. The system was implemented using Python programming language version 3.5 and Natural Language Toolkit (NLTK). The developed system was evaluated using the mean opinion score, where word orthography accuracy, syntax correctness, and word correctness were measured. The result showed that the system was able to translate English to Maninka with 100% accuracy relative to a Maninka expert translation. However, only 17.3% of the respondents got the orthography accuracy correct, while 91.5% and 97.3% of the respondents got the word correctness and syntax correctness, respectively. This indicated that some Maninka speakers were not careful enough about the writing styles of the Maninka language, especially with the application of tone marks.