Advanced Media Inc. (Headquarters: Toshima-ku, Tokyo; Chairman and CEO: Kiyoyuki Suzuki; hereinafter, "Advanced Media") has announced the launch of AmiVoice Cloud Platform, the No. 1 voice tech platform for developers in Japan.*We have added a "word registration" function to the end-to-end speech recognition engine provided by "AmiVoice API," a speech recognition API that utilizes AI speech recognition AmiVoice.
Previously, the "word registration" function was only available with a hybrid speech recognition engine. Now, by adding support for an end-to-end speech recognition engine, we have improved convenience and made it more user-friendly for a wider range of users.
In speech recognition, word registration is a crucial function that greatly contributes to improving the accuracy of recognizing technical terms and proper nouns, and it has actually been requested by many customers. With the addition of this function, users can register the technical terms and proper nouns they want to be recognized in advance, enabling more accurate text conversion.
Currently, our end-to-end speech recognition engine is widely used by many customers, particularly for applications such as transcribing conversations in contact centers and recognizing numbers and short utterances in voice bots. With the addition of the "word registration" function, it has become easier to adopt our engine in applications where its handling of proper nouns and technical terms had previously been a reason for hesitation.
The end-to-end speech recognition engine provided by "AmiVoice API" is a general-purpose speech recognition engine that supports Japanese, Chinese, and multiple languages (recognizing Japanese, English, and Chinese with a single engine). The "word registration" function is available for all of these speech recognition engines.
List of End-to-End speech recognition engines with added "word registration" functionality
| Engine | Recognition method | Message |
|---|---|---|
| Japanese_general purpose | real time Batch | An engine suitable for transcribing conversations such as conference calls, contact center calls, interviews, broadcasts, and speeches. |
| Chinese_General purpose (Simplified Chinese, Mainland China) | real time Batch | A speech recognition engine suitable for Chinese transcription and voice input. |
| Multi-Language Support (Japanese, English, Chinese) | real time Batch | A single engine recognizes three languages: Japanese, Chinese, and English. Recognizes the speaker's language and converts it into text in the appropriate language. |
End-to-End is a type of speech recognition engine architecture. Compared to conventional hybrid engines that utilize a "language model," "acoustic model," and "pronunciation dictionary" for speech recognition, it has a simpler structure that performs speech recognition using a single neural network. Because speech recognition is performed consistently within the End-to-End model, improvements in recognition accuracy and faster processing speeds are expected.
Advanced Media will continue to work on improving the convenience and practicality of its voice recognition services by enhancing features that meet the business challenges and usage scenarios of our customers.
About the AmiVoice Cloud Platform
"AmiVoice Cloud Platform" has the No. 1 domestic market share*AmiVoice is a voice tech platform for developers that provides AI voice recognition, AmiVoice. Depending on the system and security requirements of the service you are developing, you can choose from three plans: AmiVoice API, a voice recognition API that can be used in a cloud environment; AmiVoice API Private, which can be used in a customer-dedicated environment; and AmiVoice SDK, which can be used standalone.
https://acp.amivoice.com/
*Source: ecarlate LLC “Speech Recognition Market Trends 2025” Speech Recognition Software/Cloud Service Market
Inquiries regarding this matter
PF D&O Department
Inquiries about this product
Management Promotion Headquarters Public Relations Team
Inquiries regarding the contents of this report
