Constructing shallow-transfer rule for Indonesian-English machine translation

Shallow-transfer rule is the main component of shallow-transfer machine translation that matches the pattern of source language and then performs required action to translate them into the target language. However, high development cost is required to build it because it needs much of linguistic knowledge. This paper focuses on the process to automatically generate the shallow-transfer rule for Indonesian-to-English that is followed with manual post-editing to improve the quality of automatic rule. The result shows that the automatic transfer rule can improve the performance of the word-for-word translation. Furthermore, manual post-editing process was also shown to increase the quality of automatic transfer rule. According to this result, using automatic approach is an effective way to build shallow-transfer rule at starting point that can be further enhanced by performing rule post-editing.