1 00:06:24,891 --> 00:06:28,891 the real-time captioning 2 00:09:22,695 --> 00:09:26,695 hello it has 3 00:09:37,582 --> 00:09:41,582 hello test 4 00:10:34,893 --> 00:10:38,893 Good morning. 5 00:10:42,448 --> 00:10:46,448 You can follow me inside the center 6 00:10:53,865 --> 00:10:57,865 are some of you may already know the caption thanks Lois is the service that help people with hearing disability understand the audio content of your content 7 00:11:00,575 --> 00:11:04,575 by reading the text has returned instead of listening to the audio 8 00:11:09,138 --> 00:11:12,639 C & D screen is a demo of the service 9 00:11:12,639 --> 00:11:16,639 why does real-time caption 10 00:11:28,835 --> 00:11:31,269 and approved padding are the caption of pure carbon content so what is the challenging these are real transcription thing is that we have to provide a chicken soup kitchen in Stanley in 5 to 10 seconds after the speaker is speaking 11 00:11:31,269 --> 00:11:34,136 and we also need the 12 00:11:34,136 --> 00:11:36,931 accuracy of the caption to be high enough out 13 00:11:36,931 --> 00:11:39,983 people who read it 14 00:11:39,983 --> 00:11:43,983 panare the caption to understand the content 15 00:11:45,452 --> 00:11:49,452 so the accuracy of the caption should be higher than 90%. 16 00:11:55,479 --> 00:11:59,063 for people to understand do content of the speaker the center we are providing service in 2017 17 00:11:59,063 --> 00:12:03,063 6 years ago 18 00:12:06,465 --> 00:12:10,465 use three different techniques to do that the first one because someone has typing 19 00:12:15,923 --> 00:12:19,832 the second one is a respeaking at the first one we use the automatic speech recognition and you will see that a more obvious at that technique 20 00:12:19,832 --> 00:12:23,832 I still work this may explain the first one 21 00:12:26,665 --> 00:12:30,665 where are the first one that be called a simultaneous typing you use has please 22 00:12:34,883 --> 00:12:38,883 20 seconds type is together should how to type or to transcribe the small part of speech 23 00:12:50,565 --> 00:12:53,867 5 Second so the 1st type is viewer listen to the first five second of the speaker and have the second one will listen and taken off of the speaker and so on 24 00:12:53,867 --> 00:12:57,867 they help each other in double-a minor and they have the 25 00:12:59,611 --> 00:13:03,611 commercial. Tech in order to provide a real transaction 26 00:13:04,442 --> 00:13:08,105 let's move on to this true this is the room by 27 00:13:08,105 --> 00:13:12,105 follow Me here 28 00:13:16,424 --> 00:13:20,424 . this one is that there are 29 00:13:20,428 --> 00:13:24,207 ... that'd be used on the salmon tiniest typing this name 30 00:13:24,207 --> 00:13:26,075 I did y'all doing it in time we are servicing 31 00:13:26,075 --> 00:13:28,174 why did captioning service 32 00:13:28,174 --> 00:13:32,174 for the 33 00:13:35,395 --> 00:13:39,395 University we have a partner University in of hand in the 34 00:13:41,167 --> 00:13:45,167 on autism part of Thailand are they have the classroom with a have a 35 00:13:53,757 --> 00:13:57,757 and if student and teachers are like a student I started together and then the the future and all the professors are used to know my voice and be provided 36 00:13:59,860 --> 00:14:03,860 show me do this service for 4 years and then you see the speak 37 00:14:09,498 --> 00:14:13,498 in December this is either you said it's the first of the summer time yes I've been so they're taught in the 38 00:14:15,967 --> 00:14:19,967 the lower box is the text of the audience 39 00:14:21,387 --> 00:14:22,863 and like I need to take each other to provider 40 00:14:22,863 --> 00:14:26,863 41 00:14:26,929 --> 00:14:30,929 the complete caption. ideas have the sharp chees e 42 00:14:32,214 --> 00:14:33,521 on the left hand side and the right hand side to have them type past and we also use the 43 00:14:33,521 --> 00:14:37,521 pics of the ocean 44 00:14:39,073 --> 00:14:42,002 what did you use on your mobile to start on letter and then 45 00:14:42,002 --> 00:14:46,002 you suggest the full word and they can choose 46 00:14:52,621 --> 00:14:56,621 do you set this technique when he started the center because six years old the accuracy of the 47 00:15:00,612 --> 00:15:04,612 automatic speech recognition is not hiding now but accuracy is about 70 to 80% 48 00:15:05,920 --> 00:15:07,249 which is not enough our people will to understand and it's not enough 49 00:15:07,249 --> 00:15:10,102 to correct 50 00:15:10,102 --> 00:15:14,102 actually we have the editor on 51 00:15:16,181 --> 00:15:20,181 right here if you would like to improve the accuracy of the 10th we have another person here 52 00:15:22,739 --> 00:15:26,739 Department iI'm also so I've been very hours interview 53 00:15:31,387 --> 00:15:35,387 I think it's me to do the Texas to 10 but now it's offered accuracy is a party at high automatic speaker 54 00:15:38,802 --> 00:15:41,321 and we reach the accuracy around eighty-five to ninety percent so we move on to 55 00:15:41,321 --> 00:15:45,321 another technique 56 00:15:49,835 --> 00:15:53,428 . Let's look at this one on the left hand side and just bring the technique that we use for the real time 57 00:15:53,428 --> 00:15:57,428 what I'm speaking right now so we you the speech recogni 58 00:16:03,606 --> 00:16:04,988 tion in English and then because my speaking may not still in a few and I'll get other and their speech recognition 59 00:16:04,988 --> 00:16:08,988 create 60 00:16:10,731 --> 00:16:13,267 the error or miss spelling and then he have a character here 61 00:16:13,267 --> 00:16:17,267 to correct the error 62 00:16:21,345 --> 00:16:24,838 to character to help each other to correct to make it be able to do it in real time because we need 63 00:16:24,838 --> 00:16:28,838 both speed and accuracy for real-time captioning 64 00:16:30,155 --> 00:16:32,138 we also use part-time also but beat me do it in English our staymore 65 00:16:32,138 --> 00:16:36,138 is it a lot of videos 66 00:16:37,635 --> 00:16:39,535 are you still technique more and more right now because of the accuracy of the 67 00:16:39,535 --> 00:16:43,535 speaker condition is 68 00:16:48,743 --> 00:16:50,806 another technique that we use before this one is to be called on let's move on to 69 00:16:50,806 --> 00:16:53,755 another part of our Center 70 00:16:53,755 --> 00:16:57,162 let me show you we have 71 00:16:57,162 --> 00:17:00,950 72 00:17:00,950 --> 00:17:04,950 revoice and help this kind of projects 73 00:17:15,056 --> 00:17:19,056 yes so be be used in many parts so that you you see the what's the suggestion in Thai so that 74 00:17:23,772 --> 00:17:27,772 can you type up later and you suggest then the next one. yes I do the suggestion that we use the 75 00:17:33,095 --> 00:17:37,095 to select the word that I should be the one because she couldn't of people and chola 76 00:17:38,101 --> 00:17:42,101 a hundred but that's type two or three letters we have to select a sophomore. 77 00:17:42,196 --> 00:17:46,196 the camost likely to be the one that stays 78 00:17:59,180 --> 00:18:01,401 me to use you use what's that is state of our general type and also for the specific domains because the cake for my education to use different model or different 79 00:18:01,401 --> 00:18:05,401 the text that we have 80 00:18:10,910 --> 00:18:14,910 have to to type and then also be used to the automatic speed of commission so now we reduce the number of a 81 00:18:18,322 --> 00:18:21,802 instead of use our focus on one person video statue character to do this tack 82 00:18:21,802 --> 00:18:25,802 let me show you that I'm not. 83 00:18:27,609 --> 00:18:28,747 we have a full Studio yeah that we do the Re-voice so this is one this YouTube 84 00:18:28,747 --> 00:18:30,346 improve. 85 00:18:30,346 --> 00:18:34,346 the quality of this piece 86 00:18:37,372 --> 00:18:38,899 because we know that if the voicspeed is not good enough that there's a wife that is it nice 87 00:18:38,899 --> 00:18:41,524 88 00:18:41,524 --> 00:18:45,524 such as the noise at the area from automatic speaker 89 00:18:54,454 --> 00:18:58,454 so now we have the what does be called the verify why are we speak but they hear and a speaker 90 00:19:01,172 --> 00:19:03,381 clear manner in a quiet room to improve the quality of the speed before we send it to the spirit of 91 00:19:03,381 --> 00:19:07,381 today more for you here 92 00:19:07,994 --> 00:19:11,994 actually if you would like to try you can Skype also 93 00:19:12,712 --> 00:19:16,712 so you will listen to the radio audio and bedroom speaker 94 00:19:43,189 --> 00:19:46,230 so sometimes we have to choose the caption for the c. 95 00:19:46,230 --> 00:19:50,017 llooking at this line I can tell 96 00:19:50,017 --> 00:19:52,562 you and that's fine what was 97 00:19:52,562 --> 00:19:56,562 the text vocabulary that 98 00:19:58,598 --> 00:20:02,598 yeah talking about sometime they have a technical term or the name so you have made 99 00:20:02,765 --> 00:20:06,765 sure that is similar to the one that she was on the side 100 00:20:08,670 --> 00:20:12,670 and did you also see the text that returned from the speaker 101 00:20:15,782 --> 00:20:19,486 if you can notice that some of their viwer you may not pronounced correctly and then you did you try it again in the next 102 00:20:19,486 --> 00:20:21,646 I trying to say can you try to improve your quality of speaking 103 00:20:21,646 --> 00:20:25,646 anyone would like to try 104 00:20:29,872 --> 00:20:33,872 . we have four studios in here this one we used to 105 00:20:34,973 --> 00:20:38,973 1 Re-voice agent and 102 character depend on undertale 106 00:20:40,785 --> 00:20:44,785 okay. any questions. 107 00:20:49,553 --> 00:20:53,553 so we provided the service f 108 00:21:00,986 --> 00:21:02,776 or classroom television on the news and then before the coronation ceremony and also the news during the public 109 00:21:02,776 --> 00:21:06,776 the news during the COVID 110 00:21:06,846 --> 00:21:10,846 at the same time on five different channel in Thailand 111 00:21:12,821 --> 00:21:16,821 Thank you thank you for visiting us