This page introduces the NTT group's activities and initiatives that utilize corevo®.
Communication Engine, COTOHA
COTOHA is a communication engine that analyzes and understands spoken or written languages with a high degree of accuracy. As it communicates with the end user (our customer), it responds to inquiries, solves problems, and performs business tasks, such as order entries and issuance of invoices. It solves a wide variety of problems that our customers face, such as labor shortages resulting from a declining workforce and diversification of customer contact tasks.
Robo-Connect provides the following basic functions:
■ Communication function: Robo-Connect can converse with the first-stage communication robot, Sota (from Vstone), using the NTT group's AI technologies, such as a speech recognition/synthesis engine.
■ Photo-taking function: Sota works as a cameraman and takes photos.
■ Remote dialog function: Robo-Connect can control Sota located at a remote site and converse with a person at the remote site through Sota. Furthermore, Sota can be used to provide an optional service for those who provide nursing care.
Natural dialog engine
For corporate customers, we provide an engine capable of conducting a natural dialog. It is based on the language analysis technology used in Shabette Concier (talking concierge).
Although the input Japanese expression can take a variety of forms, the engine can interpret the expression from the context and conducts an appropriate dialog. It can also obtain content information, such as news and weather forecasts, and use it in its dialog.
A dialog can be conducted not only by smartphone applications but also through a variety of user interfaces (UIs), such as a robot, home appliance and Web browser.
Cloud robotics platform that interconnects IoT devices, such as robots and sensors, in an organic way to create new services
The cloud robotics platform is a mechanism that interconnects a variety of IoT devices, such as robots and sensors, and uses the state-of-the-art AI technologies of corevo®. Thus, this mechanism offers advanced intelligent processing useful for creating new services. To date, we have carried out feasibility tests for a wide variety of industries using the communication robot, Sota®. In these tests, we have applied the platform to new forms of over-the-counter services and guided marketing to verify the effectiveness of the platform. Looking forward, we will press on to create innovative services in various industries in collaboration with user corporations, device vendors and technology partners.
Technology for determining the degree of impropriety of an image
This technology numerically determines the degree of impropriety of image content using criteria similar to those exercised by human judgement. Since it gives a numerical result unlike ordinary filters, which make only a yes-or-no judgment on the propriety of an image, it allows an image filtering threshold to be adjusted flexibly depending on the purpose of judgment or the user's age group. It is especially effective for cases where it is difficult to check impropriety by direct visual inspection within a limited time frame because an enormous number of images are updated frequently as is the case with social media and online shops, or for cases where it is hard to keep the threshold level used by different visual inspectors uniform. Furthermore, incorporation of additional learning for a specific purpose can improve judgment accuracy.
Automatic road defect detection system
This system makes it possible to detect road defects at low cost using off-the-shelf video cameras and vehicles that the customer ordinarily drives. In recent years, the deterioration of road surfaces due to aging has become a social issue. Since the conventional road defect detection method requires expensive dedicated vehicles, it has been unaffordable for many municipalities. The present system takes a video of road surfaces while a vehicle is moving. The video is analyzed using deep learning, and road defects are detected automatically. Since the detected defects are plotted on a map, the user can check the status of each problem area while examining the relevant part of the video.
Real-time human detection based on deep learning
This technology automatically detects people in a frame extracted from a video taken by a camera. Capable of real-time detection irrespective of the conditions under which the video was taken (daytime/nighttime) and the type of background, it can relieve staff of the workload associated with continuous monitoring. Since it can detect multiple persons simultaneously, it can be used for marketing analysis that requires detection of the movements of people in a shop. This technology is hardly affected by the posture and size (whether far from or near the camera) of each person, and by the position and angle of the camera. It achieves highly accurate detection for a variety of cameras: cameras with different resolutions, color cameras and gray-sale cameras. Furthermore, incorporation of additional learning for a specific purpose to adapt to a specific picture-taking environment can improve accuracy.
All songs reporting service for broadcasters
This service assumes a scheme in which broadcasters need to report all the songs they used to the copyright management organization. It identifies songs played during broadcast programs using robust media search (RMS), a technology NTT has developed on its own. Since its launch in 2007, its market share has been rising steadily. It won the Minister of Internal Affairs and Communications Award and the Overall Grand Prix Award in the ASPIC Cloud Award 2015.
Monone® - an abnormal sound detection solution that enhances the quality of maintenance tasks, reduces the cost, and improves system availability using device operating sound analysis
One of the environmental data sensing solutions provided by NTT Data is Monone®: an abnormal sound detection solution. It applies abnormal sound detection technology, which is one element of the NTT group's corevo® AI technologies, to the operating sound of a device in order to provide sound visualization and perform analysis. The predictive sign detection thus made possible enhances the quality of maintenance tasks, reduces the cost, and improves system availability. Not only does the operating sound analysis and abnormal sound detection functions adopted in this solution determine the target operating sound correctly but the sound feature analysis technology, which is based on machine learning, can detect unknown abnormal sounds with a high degree of accuracy.
Video analysis reporting service - advanced AI-based video analysis
This service uses two AI-based technologies: technology for detecting the behavior and features of persons from a video and technology for finding a human image that matches a specific person from a large volume of video data*. The service uses these technologies to analyze the customer's video data and reports the analysis result to the customer. We can provide further optimization, such as carrying out additional learning and customizing the learning model used.
*: This technology uses the technology of Preferred Networks, Inc.