AISIN: Migrating the backend system of its speech recognition app to the cloud to significantly streamline operations
About AISIN
AISIN CORPORATION, is a global supplier of automotive systems. Upholding the new corporate principle of "Inspiring 'movement', creating tomorrow" we will contribute to the evolution of mobility, while providing the global population with inspirational "movement" experiences. Aisin will strive to achieve carbon neutrality by 2050 and help resolve social and environmental issues by offering concrete solutions. Further, we will bring together the technological and manufacturing capabilities perfected in our mobility and energy-related businesses with the objective of becoming a solution company that contributes to a healthier, happier and more inspired society.
Tell us your challenge. We're here to help.
Contact usBy migrating the back-end system of its voice recognition application, YYProbe, to Google Cloud, Aisin adopts a serverless architecture with Cloud Run, resulting in an environment that is highly scalable and easy to maintain.
Google Cloud results
- Increases accuracy and reliability in its speech recognition system with Speech-to-Text
- Simplifies server management work and significantly reduces maintenance load with Cloud Run
- Improves development workflows by introducing CI/CD with GitLab and Cloud Build
Opens up new possibilities for future growth with Google Cloud migration
YYProbe is a smartphone application developed by Aisin that can recognize speech in real time, convert it to text, and display the contents of the conversation on the screen.
It was originally developed to collect and analyze audio data from meetings and other events, to accumulate knowledge about operations and improve productivity. It is equipped with a unique algorithm that can quickly and accurately identify voices even in noisy environments such as a factory. YYProbe has a high processing speed and speech recognition accuracy.
Currently, Aisin is utilizing its high-precision speech recognition technology to promote the use of the app as a communication support app for the hearing impaired, and have begun offering a paid premium plan.
YYProbe uses Speech-to-Text as its voice recognition engine. Junki Yamakawa, Assistant Project Manager, YYSystem Project, 5TH DEVELOPMENT SECTION, ADVANCED DEVELOPMENT DEPARTMENT, shares the reasons behind adopting Speech-to-Text, "When we compared several speech recognition engines at the beginning of development, we found Speech-to-Text to be overwhelmingly superior in terms of speech recognition accuracy and real-time response, and decided to adopt it. At this point, the backend system itself was built on another cloud vendor service, and only the speech recognition engine was using Google Cloud."
Rearchitecting while maintaining speech recognition accuracy, and building a mechanism that scales
In this project, YYProbe's backend system infrastructure was also migrated to Google Cloud, and at the same time, it was rebuilt into a more modern system configuration using managed services. Aisin needed to improve scalability due to growth in the number of users, and the increased operational load and costs associated with the increase in the number of servers.
"At the time of planning the migration project, we re-examined the accuracy and response of the speech recognition engine, and confirmed that Speech-to-Text still has the advantage."
—Junki Yamakawa, Assistant Project Manager, YYSystem Project, 5TH DEVELOPMENT SECTION, ADVANCED DEVELOPMENT DEPARTMENT, AISIN"The number of users has grown due to the increase in awareness both internally and externally. As such, this resulted in the operational load and servers increasing, and adjusting resources to respond to the increase in access," shares Yamakawa. "In addition, since we started offering paid plans, the impact of system failures also increased, and it became necessary to review the architecture itself in order to ensure higher stability and scalability. To meet the needs of our customers, we are also developing analytics and API services other than YYProbe within the system for development and operation, which is something we always had the intention to build."
Yamakawa goes on to add that while the system needed to be modified, they continued using Speech-to-Text for the speech recognition engine. "At the time of planning the migration project, we re-examined the accuracy and response of the speech recognition engine, and confirmed that Speech-to-Text still has the advantage. Since it is essential to be a speech recognition engine, switching to another engine was not feasible. Given that the existing system already has a unique algorithm that is suitable for Speech-to-Text, we wanted to leverage it."
He also explains the rationale behind the decision for infrastructure migration, "First, to continue using Speech-to-Text, it would have been more advantageous in terms of cost to unify other parts with Google Cloud. The fact that it was a managed service and had all the necessary functions was also a deciding factor in our final decision. We were concerned that our team did not have enough experience using Google Cloud, or knowledge about serverless architecture, but the Google Cloud team actively supported us."
Adopting a serverless architecture and CI/CD with Cloud Run to significantly reduce maintenance load
The new system adopted a serverless architecture using Cloud Run. This is because it was necessary for Aisin to build a highly scalable system that leverages the value of the cloud, taking into account requirements such as future growth in the number of users, countermeasures for times when load is concentrated, and the addition of new services.
"We wanted to minimize maintenance, so we adopted Cloud Run because it eliminates the need to manage each server individually and allows easy linkage with other services. You can fine-tune the settings with parameters, and deployment is completed with a single command. This greatly reduces the maintenance effort compared to our legacy system, especially when various tasks can be performed with just a browser."
—Junki Yamakawa, Assistant Project Manager, YYSystem Project, 5TH DEVELOPMENT SECTION, ADVANCED DEVELOPMENT DEPARTMENT, AISIN"We wanted to minimize maintenance, so we adopted Cloud Run because it eliminates the need to manage each server individually and allows easy linkage with other services. You can fine-tune the settings with parameters, and deployment is completed with a single command. This greatly reduces the maintenance effort compared to our legacy system, especially when various tasks can be performed with just a browser. In our previous system, just checking the log requires multiple procedures such as downloading the log file and logging in directly to the server. With Cloud Run, everything can be completed on a single screen on the browser," says Yamakawa.
Aisin also took on new challenges in system development workflows when migrating to Google Cloud, with the introduction of a Continuous Integration/Continuous Delivery (CI/CD) mechanism using Cloud Build. According to Yamakawa, the introduction of CI/CD has had a great effect in improving the efficiency of development.
"The new system uses GitLab for source code management, and a push to GitLab triggers the build process by Cloud Build. You can complete the build by simply pushing the source code locally, which is very convenient, and the development workflow has been simplified, leading to improved efficiencies. In the future, we are considering automating deployment to Cloud Run after establishing an operation method," says Yamakawa.
Future plans to migrate other services to Google Cloud
In addition to YYProbe, Aisin provides various services based on YYSystem. For example, conversations recorded using YYProbe can be aggregated and analyzed using YYAnalytics, and the analytical results can be viewed in a browser. In addition, the company is in the midst of developing an API service that can leverage the various functions of YYSystem as an API, which will be made public in the future when completed.
Yamakawa says that following the successful migration of YYProbe to Google Cloud, they are considering migrating infrastructure for these services as well. "By accumulating knowledge on Google Cloud, we are now able to think about new possibilities for systems other than YYProbe. For example, we don't know how many requests we will receive in advance. Cloud Run is useful in this case as it offers high flexibility. We also want to introduce the concept of microservices."
Yamakawa goes on to say that there are still challenges to be tackled with YYProbe. "Since some customers want to use YYProbe on-premises, we are considering how we can achieve this on Google Cloud. We believe that Anthos, a hybrid cloud platform, is able to help us with this. And since we moved to Google Cloud, we are also looking at AI platforms such as Vertex AI."
"Since there are a certain number of customers who want to use YYProbe on-premises, we are considering how we can achieve this on Google Cloud. We believe that Anthos, a hybrid cloud platform, is able to help us with this. And, since we moved to Google Cloud, we are also looking at AI platforms such as Vertex AI."
—Junki Yamakawa, Assistant Project Manager, YYSystem Project, 5TH DEVELOPMENT SECTION, ADVANCED DEVELOPMENT DEPARTMENT, AISINYYProbe is now at a turning point, with the start of its paid premium plan being offered, and Yamakawa is extremely excited about expanding it into various fields in the future. "Currently, we are receiving calls from various customers, including local governments, hospitals, schools for the deaf, and private NPOs, so we would like to actively promote the introduction of the system. If we proceed with this development, I expect that we will be able to start realizing the concept of recording and utilizing voice data, which was the original idea," he concludes.
Tell us your challenge. We're here to help.
Contact us
About AISIN
AISIN CORPORATION, is a global supplier of automotive systems. Upholding the new corporate principle of "Inspiring 'movement', creating tomorrow" we will contribute to the evolution of mobility, while providing the global population with inspirational "movement" experiences. Aisin will strive to achieve carbon neutrality by 2050 and help resolve social and environmental issues by offering concrete solutions. Further, we will bring together the technological and manufacturing capabilities perfected in our mobility and energy-related businesses with the objective of becoming a solution company that contributes to a healthier, happier and more inspired society.