CTC Launches SmartBERT, a Next-gen AI Natural Language Processing Engine from Intumit in Taiwan

In the future, CTC will collaborate with Intumit to continue using SmartBERT in a range of applications such as predicting customer needs and multilingual translation. This partnership aims to further enhance service quality and drive digital transformation (DX) for customers.

Improving the correct answer rate of AI chatbot services using big data

Intumit Inc.

ITOCHU Techno Solutions Corporation

ITOCHU Techno-Solutions Corporation (Ichiro Tsuge, President & CEO; headquartered in Chiyoda-ku, Tokyo; hereinafter “CTC”) today commences sales of SmartBERT, an AI natural language processing engine utilizing the deep learning technology of Intumit Inc. (JD Chiou, CEO; headquartered in New Taipei City, Taiwan; hereinafter “Intumit”), a startup in Taiwan that operates AI and robotics businesses. It is expected that the combination of this engine with SmartRobot, an AI chatbot service that Intumit has provided for some time, will improve the correct answer rates of the automatic replies of AI chatbot services using big data. Regarding the menu of services, customers can select from two options, the on-premises version (24 million yen or more) and the SaaS version (400 thousand yen or more per month).

As part of efficiency improvements in business associated with work style reforms and digital transformation, companies are increasingly introducing chatbots to respond to the inquiries of employees and customers. Because of the increase in the number of inquiries, however, maintenance also increases, such as the expansion of scope of operations and the addition of new FAQ data. Automation and efficiency are necessary for chatbot operations.

SmartBERT, which CTC begins selling at this time, is a next-gen AI natural language processing engine, implemented by Intumit using the BERT* algorithm, a natural language processing technology used by Google for its search engine. SmartBERT automatically learns to correlate sentences from the FAQ data provided to the chatbot and history data from past responses. This improves the accuracy of its understanding of the intent of inquiries. The use of SmartBERT reduces the workload for operations associated with the increase in inquiries and FAQ data, while maintaining a high correct answer rate, enabling automation and efficiency in chatbot operations. In Taiwan, a large AI chatbot service utilizing the processing engine has sustained a high correct answer rate of approx. 90% for automatic replies with approx. 3,000 FAQ data items stored. It has also enabled a 50% reduction in the work hours necessary to maintain the data, when compared with conventional AI chatbot engines.

From standard help desk answers to operations and troubleshooting regarding complex equipment, it allows automatic responses to be provided for inquiries that are becoming more complex and diverse, leading to improvements in customer satisfaction and service quality.

In localizing SmartBERT into Japanese, CTC has worked on the verification and evaluation of performance and functionality as well as the ensuring of quality from the prototyping phase. Moreover, CTC has been selling SmartRobot since 2019. In addition to selling SmartBERT, CTC will provide a total service, including system construction, security measures, training, and maintenance support, which all accompany the introduction of SmartBERT.

Moving forward, in cooperation with Intumit, CTC will continue using SmartBERT in an array of applications, such as predicting the potential needs of customers and translating into multiple languages, to help customers further improve service quality and advance DX.

* BERT: Bidirectional Encoder Representations from Transformers, a method of natural language processing to improve the accuracy of understanding context by reading sentences from both head and tail.

Contact Information

Intumit Inc.
E-mail: sales@intumit.com
https://www.intumit.com/

ITOCHU Techno Solutions Corporation, Corporate Communications Dept.
E-mail: press@ctc-g.co.jp
https://www.ctc-g.co.jp/