Developing Indonesian-English Hybrid Machine Translation system

Rule Based Machine Translation (RBMT) and Statistical Machine translation (SMT) have different approach in performing translation task. RBMT uses linguistic rule between two languages which is built manually by human in general, whereas SMT uses co-occurrence statistic of word in parallel corpora. We combine those different approaches into Indonesian-English Hybrid Machine Translation (HMT) system to get the advantage from both kind of information. Initially, Indonesian text is inputted into RBMT. Then, the output will be edited by SMT to generate the final translation of English text. SMT is capable to do this because on the training process, it uses RBMT's output (English) as source material and real translation (English) as target material. Unavailability of ready to use Indonesian-English RBMT system becomes a challenge to do this research. Our study shows that SMT still outperforms HMT by 8.01% in average.