Translation system for academic lectures created by Japanese researchers

Nara Institute of Science and Technology

Scientists at the Nara Institute of Science and Technology (NAIST), Japan have developed a new machine learning that will help students to listen live to lectures by Nobel Prize Laureates. Moreover, they can even earn credits from some of the most reputable universities. Scientists found that although a number of students have access to online academic information, a major barrier is the language of the videos which is not comprehensible to foreign students.

Advancements in communication technology have made it easy to locate an address in a foreign city. These commands are usually precise- about a sentence or two. Hence the translation systems are also competent. However, this is not the case when it comes to the translation of academic lectures, which last for at least an hour. Here the communication becomes less coherent.

“NAIST has 20% foreign students and, while the number of English classes is expanding, the options these students have are limited by their Japanese ability,” reports NAIST Professor Satoshi Nakamura, who is also one of the lead authors of the study.

The research group led by Nakamura produced a learning-based system which translated Japanese lecture speeches into English. This they achieved by archiving 46.5 hours of lecture videos from NAIST, together with their transcriptions and English translations. The end result is presented with subtitles in English as well as in Japanese, which were synched with the lecturer’s speech.

According to reports, the team has used archived videos with subtitles, this helps with better processing time and accurate translation, in comparison to simultaneous streaming of translation with live presentations. The lectures that were achieved and used for the study were based on robotics, software engineering, and speech processing. The team reports of various error rates while working on the project. One of the major error rates was the length of time speaking without pause. The team adds that the corpus of the project could be further improved for efficiency.

NAIST presented the study at the 240th meeting of the Special Interest Group of Natural Language Processing, Information Processing Society of Japan (IPSJ SIG-NL). “Japan wants to increase its international students and NAIST has a great opportunity to be a leader in this endeavor. Our project will not only improve machine translation, it will also bring bright minds to the country,” Nakamura concluded.