Bahasa Kita is a startup under UMG Idealab focused on developing voice processing and natural language technology product – known as Natural Language Processing (NLP). They’ve recently introduced new innovation, a technology to transcribe automatically.
Bahasa Kita, using its own NLP technology, has examined the latest innovation to transcribe conversation of the last presidential candidate debate in real-time. They also present data analysis from the transcribed conversation.
Ardhi Ma’arik, Bahasa Kita’s VP Product explained to DailySocial about the transcription process, starts from voice recording, then the voice will be sent directly to Bahasa Kita transcription engine. The machine will process it and produce the text, to be sent back to the recording device in real-time.
“In general, a real-time transcription process is similar to any speech engine, what makes it different is the accuracy, speed, and domain covered,” he added.
Speaking (with voice) is the most natural engine people use to communicate. Voice technology is in the front row to acquire information. When voice becomes text, it’ll be easier to extract information to know the meaning and objective of the speech on the computer.
In addition, we can identify one’s personal information from the voice data. For example, the origin (through dialects), age prediction (without asking), health analysis (from the voice color), and many more.
Bahasa Kita’s strategy in 2019
Aside from launching “minutes of meeting” with ability to transcribe speaker’s voice into text, Bahasa Kita has also released a mobile-based transcription tool this year. Bahasa Kita aims to be the best company in charge of voice detection technology in Indonesia.
“We tried some procedures, such as engine accuracy improvement, Indonesian local language covers, and others. We also want to develop various kinds of voice products and its extensions, for example, smart speaker including the synthesis of text into voice,” he said.
–
Original article is in Indonesian, translated by Kristin Siagian
