Automatic Generation with Manual Post-Editing of Shallow-Transfer Rule for English-to-Indonesian

Machine translation from English to Indonesian is required in line with the flood of document written in English. In order to develop shallow-transfer machine translation for English-to-Indonesian, we need shallow-transfer rule to change the structure of English into Indonesian. However, it cost high expense to build it manually from the scratch as it needs much of linguistic knowledge to cover most of the translation rules. In this research, we combine the automatic and manual approach to generate the shallow-transfer rule for English-to-Indonesian. Automatic generated rule is manually post-edited to improve its quality. We compare the performance of system when applying no rule (word-for-word translation), automatic rule and post-edited rule. The result shows that the automatic rule can significantly improve word-for-word translation and post-edited rule also fairly improve the quality of automatic rule. Overall, post-edited rule can significantly improve word-for-word translation by 40.45%.