USEN CORPORATION which provides USEN Omotenashi Cast, an application that delivers multilingual announcements in commercial facilities, tourist destinations, airports, and other public spaces. We spoke with Mr. Kitazawa, Director of the Universal Design Tech Lab., Group Alliance Promotion Division, U-NEXT HOLDINGS Co., Ltd., about how they utilize speech recognition, the reasons for switching from Siri to AmiVoice, and the effects it has had.

Issues/background

An app for creating and broadcasting multilingual announcements

"USEN Omotenashi Cast" is an app that is used in facilities that are particularly facing labor shortages, places that need to accommodate inbound customers, commercial facilities, hotels, airports, etc. These facilities need to provide quick and accurate information to customers visiting from both Japan and abroad, so multilingual support is extremely important.

This app not only allows you to create in-house announcements and instantly broadcast them in multiple languages, but also allows you to translate and display instructions for foreign customers in their native language on tablet devices, and display text for the hearing impaired.When creating announcements or using tablets to show text to customers, there is a strong demand from users who want to be able to input information quickly by voice rather than typing on a keyboard, so we have added a voice input function.

Key Drivers of the Adoption

Switched from Siri due to its ability to recognize proper nouns with high accuracy

Although Siri was originally used for voice input, there was a need for higher speech recognition accuracy. In places like commercial facilities, tourist attractions, and airports, it is particularly important to be able to handle proper nouns and technical terms, to recognize phrases accurately, and to respond quickly to new proper nouns.
Siri often has difficulty recognizing proper nouns and technical terms, so we decided to consider switching to another engine.
AmiVoice API has high accuracy in recognizing common proper nouns and has a wide variety of variations. The voice recognition engine is constantly updated to the latest version, so it can recognize new words that have recently emerged, and we appreciate its fast response time.

The difficulty of implementation

The process went smoothly without any particular issues. The manuals, technical documentation, FAQs, and other content provided were easy to understand and provided great support during the implementation process.

Impact of the Implementation and Future Outlook

Reduced misrecognition allows for smoother guidance

The introduction of the AmiVoice API has resulted in significant benefits, such as fewer misrecognitions even for longer audio and an improved accuracy rate for kanji conversion. As a result, it has become possible to provide a more comfortable and smoother service on-site.

In the future, we expect further enhancements to the translation features of "USEN Omotenashi Cast", especially support for simultaneous speech recognition in Japanese and Korean (currently, AmiVoice API supports Korean Engine).
We have also released "USEN Mobile Interphone," a remote customer service tool that can handle not only multilingual announcements and face-to-face translation, but also multilingual translation remotely.
We hope to continue to improve and strengthen the accuracy of speech recognition, handling of proper nouns, and accuracy rate of kanji conversion, and we are considering expanding our use of the AmiVoice API in the future.