News 1

Learn Korean with an App while
Talking to the ETRI AI

- ‘KOKOA’, a special helper of learning Korean for more than 10 thousand foreigners
- Transfer of the AI source technology which can process the evaluation of pronunciation and
talking freely on various topics

The Electronics and Telecommunications Research Institute (ETRI) announced that it had developed a source technology of the interactive education system which freely talks with students and evaluates their pronunciation and expression and transferred it to numerous content developers of language learning.

The interactive education system for learning a foreign language developed by the ETRI adopted the method of having a conversation in either English or Korean on topics in various circumstances. By providing a detailed evaluation of pronunciation and stress as well as recognizing the expression accurately, it helps studying a foreign language by oneself effectively.

IIRTech Inc. received the technology developed by the research team and launched an education service app of the Korean language (KOKOA¹⁾) in July. The app showed a great impact so that more than 10, 000 foreigners downloaded it within 2 months after its release.

The technology of the research team was also applied to the AI App of practicing the conversation in Korean, “KSI Korean AI Tutor²⁾” released by the King Sejong Institute Foundation this March. Also, it is an app developed by IIRTech Inc. in line with the purpose of establishing the King Sejong Institute, which is disseminating the education of the Korean language to foreigners.

Yoon-geun Lee, Senior Vice President of AI Research Laboratory at ETRI expressed, “We have developed an AI source technology that makes it easy for multicultural families and foreign workers in Korea, and also overseas learners who are interested in the Korean Wave to learn Korean. Especially, I hope that it will be helpful for solving the problem of non-face-to-face classes and shortage of Korean language teachers due to the COVID-19 pandemics.”

In the future, the research team is planning to develop the ‘Technology of recognizing the voices of non-native speakers’ and ‘Dialog processing technology’ to recognize the voices of foreigners of various nationalities more accurately and to make them be able to talk on various topics naturally with small amount of the data.

This accomplishment was processed as part of the “Development of semi-supervised learning language intelligence technology and Korean tutoring services for foreigners” of the Ministry of Science and ICT. Then, it could produce the accomplishment of 23 patents, 24 papers, 15 technology transfers, and one case of being selected as the international standard proposal.

1) Knocked on the new door of learning Korean

2) King Sejong Institute Korean AI Tutor

News 2

ETRI Developed the Medical Intelligence, ‘Dr.AI’

- It predicts the health conditions of the patients by collaborating diagnosises with the medical intelligence of institutions and accomplishes the accuracy of 90%
- To be prepared for the “With Corona” era, it established the non-face-to-face
cooperation system using the medical AI

The Electronics and Telecommunications Research Institute (ETRI) developed the Dr. AI, which analyzes the patient’s current condition precisely by integrating medical intelligence established in multiple hospitals and anticipates health in the future.
Dr. AI, which was developed by the ETRI, helps diagnosis by the method of simultaneously utilizing EMR-based medical intelligence of each hospital (ensemble) instead of integrating the EMR. In other words, it has the effect of jointly utilizing medical data from other institutions without directly accessing sensitive information. It means that medical information by each institution has been turned into big data indirectly.
When the patient’s current data is entered into the Dr. AI, the medical intelligence of each institution is analyzed independently. Then, the optimal prediction is selected by adjusting errors after integrating the results. It showed a higher accuracy by about 10% compared to the case where a single institution’s medical intelligence is used, and it is because each medical intelligence has different patient group data depending on hospital characteristics. The research team improved the accuracy by the method of collaborating with medical intelligence, which studied data that differs by institution.
The core technologies of the Dr. AI are ▲Ensemble Medical Intelligence (analysis of forecast trends and errors by institutions) ▲Time-series EMR medical intelligence (analysis of prediction basis and health status) ▲Multi-modal medical intelligence (learning medical data), and more. The Ensemble Medical intelligence identifies the patient’s future health condition based on the most appropriate medical data established by the Dr. AI from hospitals nationwide, no matter which hospital the patient has visited. The research team made it possible to predict more accurately by designing the analysis weight and concentration, such as the frequency of hospital visits and examination items, differently, while utilizing time-series EMR medical intelligence. The multi-modal medical intelligence improves the accuracy of predicting cardiovascular disease and also provides help for the customized treatment for a patient as it studies and utilizes cardiac CT image data as well as the EMR data¹⁾.
ETRI is planning to improve its accuracy by studying a large number of cases with deep learning which humans cannot handle. It announced that the purpose of the research is guaranteeing the healthy lives of the citizens by preemptively diagnosing major diseases while utilizing even similar cases actively for prediction.
Jaehun Choi, who is in charge of developing the technology of Dr. AI as a principal researcher, said, “Tertiary hospital, as well as primary and secondary hospitals, where patient data is not relatively sufficient, can utilize the medical intelligence of other hospitals simultaneously to derive the effects which is like cooperation. It is expected to enable an upward leveling of the medical standard.”
Currently, the research team establishes the joint venture company, MDHi Inc. to commercialize Doctor AI with DAEA Inc. and ETRI. For its core technologies, technology transfers and commercialization are in progress separately. Moreover, this technology has been posted in the world-class journal published by the Institute of Electrical and Electronics Engineers (IEEE).

1) 570 thousand people in Asan Medical Center in 2000~2017, 110 thousand people in Ulsan University Hospital in 1997~2018, and 60 thousand people in Chungnam University Hospital in 2000~2017

News 3

ETRI Developed the Blending Platform of AI and Network for the First time in the World.

- A combination of the AI technology of facial recognition / context-awareness and the network technology such as VPN / bandwidth/ QoS control.
- By linking the KOREN between Daejeon and Pangyo ETRI has demonstrated AI-network convergence platform for realization of human-centered hyper-connected society

The Electronics and Telecommunications Research Institute (ETRI) announced that it has succeeded in developing the core technology of AI-based intelligent edge networking through the research project entrusted by the Ministry of Science and ICT(MSIT). By linking it with the KOREN¹⁾ on the network of a nationwide scale, demonstration and technology verification in connection with 4 application services were also completed.
The Intelligent Edge Networking platform developed by the ETRI is the next-generation core networking technology combining AI technology and network technology. It recognizes the user’s surrounding information through the AI technology of facial recognition/context-awareness automatically, and provides necessary service at a location close to the user (Edge) quickly and safely.
By linking the KOREN between Daejeon and Pangyo, ETRI has succeeded in demonstrating and technically verifying a total of 4 services, such as ▲Transmission path control service of context awareness-based video data ▲Context-awareness-based intelligent secure networking service ▲Context-awareness-based Network Bandwidth control Service ▲Context-awareness based customized application service.
Kim Sunmi, the head of ETRI Network Research Division, said, “Thanks to the encounter of the Intelligent Edge Technology and Networking Technology, it is now easy to create new business as delivering information quickly to where it is needed and responding became possible. In particular, it is expected to be highly useful as a core technology of providing a new experience in the new normal era.”
The research team announced that they were planning to provide an Innovative next-generation distributed intelligence infrastructure environment to expand AI services that companies can develop and use conveniently for free in the future.
This accomplishment could be achieved through joint research with Netvision Telecom Co., K-Bell Co., FISYS Co., and Korea Association of Network Industries as part of the “development of AI-based intelligent Edge Networking technology” project of the Ministry of Science and ICT.
Through this project, ETRI has made an accomplishment of publishing more than 7 SCI-level papers and applying to more than 15 domestic and foreign patents, and it is continuously promoting the industrialization of the technologies, such as 6 technology transfers to the companies related to cloud and security.

1) KOREN(Korea Advanced Research Network)
Leading test network of future networks. It is a test network for technical research related to the network