Shinya Yamamoto

My name is Yamamoto and I am the head of the API development team.

AmiVoice APISentiment analysis is now available as an option for the asynchronous HTTP speech recognition API.The sentiment analysis function is part of our contact center solution.AmiVoice® Communication SuiteHowever, it is included as standard and has been well received.

AmiVoice API is provided by ES JapanESAS (Emotional Signature Analysis Solution)ESAS is based on the latest emotion analysis engine "LVA7" provided by Israeli company Nemesysco.ES Japan has accumulated operational experience at its own call center and tuned the result parameters to make it easier to use in Japan.A sentiment analysis engine.

The unique feature of ESAS is that it analyzes emotions from voice rather than text. It is actually used in the following use cases:

Call center

• Visualizing employee work stress to provide support before employees leave
• Visualization of customer purchasing intent and decision-making

HRM/Recruitment

• Personality assessment during online interviews
• Job matching based on emotion type classification

Other Information

• Robotics, education, games, etc.

As a result of emotion analysis, 20 emotion parameters are obtained at a rate of about once every two seconds. For example, emotion parameters can be expressed as negative emotions, such as:stress,Puzzled,Aggressive indignation, etc. are considered positive emotions.joy,passion,excitement,For details, see the manual.Emotion parameter range and value trendsSee the table below.

First, visualize the time transition of emotion parameters, collect data for a certain period of time, and then determine whether the average, variance, or values ​​of emotion parameters are higher or lower than normal. If you can find correlations with real-world phenomena, we believe you will be able to utilize this technology in a variety of ways beyond the use cases listed above. Please give it a try. We hope it will help you add new value to your applications.

For details on how to use sentiment analysis and the results you can obtain, see the AmiVoice API manual.Emotion analysisPlease refer to.

from now on

Sentiment analysis is currently only available as an option for the asynchronous HTTP speech recognition API, but we plan to make it available for the synchronous HTTP speech recognition API and the WebSocket API in the future.