SEARCH

검색창 닫기
  • BTC
  • ETH
  • XRP
  • BCH
bithumb제공 bithumb제공
  • BTC
  • ETH
  • XRP
  • BCH
bithumb제공 bithumb제공

Seoul City to take minutes through artificial intelligence

Bidding for AI minute support system due in April


There are 28 conference rooms in the Seoul Metropolitan Government. More than 8,000 meetings take place in these rooms a year. Of Seoul City’s 195 committees, 168 committees have to make public minutes. Seoul City’s officials in charge must work additionally to take minutes after meetings. The time needed for this work is nearly double the time for meetings. So it’s impossible to confirm or adjust what was discussed in the meeting right after it was finished. One has to confirm separately through such means as email.

Seoul City has embarked on a project to “establish an artificial intelligence minute support system” to resolve these work burdens and inefficiencies. Bidding proposals for this project will be received April 19-23. Seoul City plans to use the project to realize an intelligent e-Government using AI and other new technologies.

The core of the project is the accuracy of voice recognition concerning the speaker’s syllables. It’s necessary to measure how many syllables can be recognized “properly.” The goal is more than 90%.

The Seoul municipal government is drawing up a road map in which more than 50 speakers can be recognized in real time and such functions as distinguishing speakers and analyzing and sorting out the speaker’s keywords are possible.

“When multiple speakers converse, the recognition accuracy can be much lower than when one speaks,” said a voice recognition industry official, adding that “if we can classify and record who says what accurately in conferences attended by many people, we can diminish our efforts to take minutes.”

This project will be carried out in cooperation with the Seoul City Data Center, which already has hardware for the AI minute support system.

The test operation of the system will begin in the third quarter of this year. The system will be test-operated after data is collected and this process will optimize the voice recognition engine. /shim@decenter.kr

사동석 기자
sahds@dcenter.kr
< 저작권자 ⓒ 디센터, 무단 전재 및 재배포 금지 >

이메일보내기