Masters student Beatrice Chin is working on the project to build an automatic translation system for Iban and English dubbed Tribe.
“The Iban language is taught up to Form Five and students can take it at PMR and SPM levels. Textbooks are available but there is a lack of other teaching resources such as reference books.
“So there is a need to translate reference books from English or Bahasa Melayu to Iban. If this is done manually, it will be very time consuming and costly,” Chin said when giving a presentation about Tribe at Sarawak Development Institute here yesterday.
Tribe is an ongoing research project under the university’s Sarawak Language Technology Research Group (SALT), which aims to preserve the state’s indigenous languages through the use of information and communication technology (ICT).
Chin said the system used example-based machine translation, which utilises previously translated examples of similar sentences to translate a new sentence.
“For example, I want to translate the sentence ‘He bought a storybook’. The system already has translations of ‘He bought a pen’ and ‘He read a storybook’, so it will combine these examples to translate my sentence,” she explained.
She said Tribe currently had about 1,700 English sentences translated into Iban in its database.
“To increase the accuracy of the translation, we need many more examples in the database,” she said.
According to Chin, machine translation was much faster compared to human translation.
However, human editing was still required as the machine translation was only about 60% accurate.
“From our study, we found that machine translation plus human editing were about twice as fast as human translation,” she added.
This article originally appeared in The Star Online (see link below).
