Advanced Media Corporation (Headquarters: Toshima-ku, Tokyo; President and CEO: Kiyoyuki Suzuki, hereinafter referred to as Advanced Media) has added an End-to-End speech recognition engine to its "AmiVoice Communication Suite," the No. 1 AI voice recognition solution for contact centers in Japan (※1). This engine can be selected as the speech recognition engine for the customer side, allowing it to be used selectively depending on the communication environment, in addition to the conventional hybrid engine.
Speech recognition technology is widely used in the contact center industry as an important foundation supporting improved service quality and operational efficiency. However, customer communication environments and speech characteristics are diverse. In addition to situations prone to the effects of environmental noise, such as calls from outside locations like train stations or streets, or from inside vehicles, there are many factors that affect recognition accuracy, including degradation of voice quality due to smartphone communication status and unclear speech. Therefore, there is a growing need for speech recognition technology that can achieve stable and high recognition accuracy even under these various conditions.
Amidst this situation, "AmiVoice Communication Suite" has newly added an End-to-End type to its customer-side speech recognition engine, in addition to the conventional hybrid speech recognition engine. The End-to-End type is expected to achieve high speech recognition accuracy even in noisy environments or with unclear speech. In internal verification, recognition error rates were significantly improved compared to the conventional hybrid speech recognition engine, with an average improvement of 30.4% and a maximum of 37.8%. (※2) It is possible to select between the hybrid type and the End-to-End type according to the customer's communication environment, or to apply the End-to-End type to the customer and the hybrid type to the operator, allowing for flexible configurations that leverage their respective characteristics.
Furthermore, the improvement in customer-side speech recognition accuracy is expected to lead to improved quality and operational efficiency in subsequent processes, such as response summarization using generative AI. Additionally, it is equipped with a word registration function, enabling highly accurate text conversion of specialized terms and proper nouns.
The End-to-End type is a type of speech recognition engine mechanism. Compared to the conventional hybrid engine, which performs speech recognition using a "language model," "acoustic model," and "pronunciation dictionary," it has a simpler structure that performs speech recognition with a single neural network. Since speech recognition is performed consistently within the End-to-End model, a significant improvement in recognition accuracy is expected.
Advanced Media will continue to work on further advancing speech recognition technology and improving its convenience to meet the diversifying needs of contact centers, thereby contributing to operational efficiency and improved service quality.
About "AmiVoice Communication Suite"
"AmiVoice Communication Suite" is the No. 1 solution for contact centers in Japan (※1) equipped with the AI voice recognition technology AmiVoice. In addition to full-text conversion of call content, it offers numerous features such as emotion analysis, topic extraction, simultaneous monitoring of multiple calls by supervisors, and operator support, aiding in the visualization of call content and improvement of service quality.
Available in cloud and on-premise versions, with real-time and batch recognition processing options, it allows for flexible operation regardless of the number of seats or scale of the contact center.
https://www.advanced-media.co.jp/lp/communication-suite/
※1 Source: ITR "ITR Market View: Image and Voice Recognition Market 2024" Voice Recognition Market - Vendor Sales Revenue Share for Contact Center Operations (FY2024 Forecast)
※2 Our evaluation. This is the result of an internal evaluation using 12 types of evaluation data for contact centers, compared to the conventional hybrid speech recognition engine. Results may vary depending on evaluation conditions.
[Inquiries regarding this matter]
Advanced Media Corporation
CTI Business Division
https://www.advanced-media.co.jp/products/contact/callcenter/?contactchk=AmiVoice%20Communication%20Suite
FACT BOX
- Source: PR TIMES
- Category: 製品リリース
- Organizations: ITR