Quick Start to Captioning 06-22-2017

Editor's Notes

• #5: Time-synchronized text that can be read while watching a video Usually the CC icon Originated as FCC mandate for broadcast in the 1980s
• #6: 2) Assume the viewer can't hear; convey sound effects, speaker ID, & other non-speech elements relevance: [keys jangling]
• #7: Terminology: Captions assume the viewer can’t hear the audio; time synchronized & include sound effects Subtitles assume the viewer can’t understand the audio; time synchronized & translate the audio Transcripts include a plain text version of the audio; not time synchronized; sufficient for audio-only
• #8: DIY Standards Accuracy ASR errors
• #9: DIY method: YouTube. Recommend using it for timings. Transcribe your video – takes 5-6x real time; include non-speech elements Use YouTube to transcribe & set timings Try not to attempt timings manually! There’s a lot of room for sync errors. Or, download YouTube’s auto captions or edit them in the interface If you need a different format, use a caption format converter.
• #10: Important to follow best practices for caption quality. Spelling should be at least 99% accurate; include grammar for readability Speaker identification should be consistent Include relevant sound effects (keys jangling) Use punctuation to improve readability Verbatim: for broadcast, include every “um,” etc because it is scripted. Clean read: for lectures, live presentations, etc., clean read is usually preferable. CHARACTERS: 1-3 lines; 32 characters per line FONT: non-serif SYNC: should be perfect MIN DUR: 1 second PLACEMENT: shouldn’t obscure SILENCE: should note – can drop off the screen.
• #11: We expect continuous improvements Currently: 80% with our current tech Even 95% isn’t sufficient for accurately conveying complex material. Typical sentence of 8 words: 95% accuracy rate means there will be an error every 3 sentences. We use our large database of human-corrected, near-perfect transcripts to continually improve our recognition accuracy Word-to-word accuracy issues – multiply the chance of each word in a sentence being incorrect and you’ll see how accuracy issues compound.
• #12: Examples of where ASR fails. A lot of the errors make sense acoustically, but not linguistically. A human wouldn’t make those errors. Issues to look out for: Punctuation Hesitation words aren’t removed Speaker changes not captured No Speaker ID Acoustic errors: - Should be “New England Aquarium” vs “new wing of the Koran” - “forester” vs “four story”
• #13: Captions are supported on most devices where you can publish video. One limitation: some social video.
• #14: Originated as a mandate for broadcast TV Now published across all devices and platforms Most players and platforms have caption compatibility; some more advanced User control options on many platforms
• #15: Side Car: like YouTube, where you upload caption file and associate to video Encode: encode caption file onto the video > Use for kiosks or offline video Open Captions: burned in and can't be turned off Integrations make this trivial
• #16: Caption format depends on platform/player. Two common formats here Same exact file; first 3 caption frames. SCC uses hex codes; SRT is much more readable. Consider a caption format converter for something like SCC.
• #17: Many reasons outside of accessibility
• #18: 1) Accessibility: 48 million Americans growing due to: medical advancements 20% of Americans 2) Better Comprehension: 80% of users Accent, difficult content, background noise, ESL 3) Flexibility to view in noise-sensitive environments (office, library, gym) Autoplay: more and more, companies are playing captions by default with no sound – Social video like Yahoo, Facebook, Twitter w/o captions, these are inaccessible to everyone 4) Video Search: MIT survey - 97% of users found they enhanced experience Used to being able to search and go 5) SEO – Google can't watch a video DDN study > increased views by 7.3% 6) Reusability UW = 50% of students repurposing transcripts Create infographics, white papers, case studies, other docs, course materials 7) Translation > translate English transcript to make video accessible on a global scale 8) Required by Law
• #19: Research study with Oregon State University: how & why do students use captions? Prove how much captions help all students learn 98.6% found captions helpful 75% of students who use captions – not just those who are deaf/hard of hearing - used captions as a learning aid #1 reason for using captions is to help students FOCUS on the video content
• #20: 1973 Section 504: anti-discrimination law that requires equal access for individuals with disabilities. Applies to federal & federally funded programs. Section 508: introduced in 1998 to require federal communications and information technology to be accessible. Apples to federal programs, but often applied to federally funded programs through state & organization laws. Closed captioning requirements are written directly into Section 508, and are often applied to Section 504 Section 508 refresh was released in January, which references WCAG 2.0 guidelines.
• #21: 1990 5 sections Title II: public entities; Title III: public accommodations (extends to private sector) "Places of public accommodation" – what constitutes this? Tested against online businesses Refresh coming? NAD vs Netflix   Netflix sued by National Association of the Deaf in 2012 for failing to provide closed captions for most of its "Watch Instantly" movies and television shows streamed on the Internet. First time that Title III of the ADA (place of public accommodation) had been applied to Internet only businesses (before, it had only been applied to physical structures like wheelchair ramps) Court ruled in favor of the National Association of the Deafj. Netflix settlement: Netflix agreed to caption 100% of its streaming content. Profound precedent FedEx was sued for not providing closed captions on training videos Hulu settled with the National Association of the Deaf Amazon settled with the National Association of the Deaf National Association of the Deaf vs. Harvard and MIT   Harvard and MIT were sued by the National Association of the Deaf for providing inaccessible video content that was either not captioned or was inaccurately/unintelligibly captioned The first time that accuracy has been considered in legal ramifications for closed captioning (YT auto captions) NAD argued that educational online videos should be constituted as a public accommodation. In June of 2015, the Department of Justice submitted a statement of interest supporting the Plaintiffs' position that Harvard and MIT's free online courses and lectures discriminate against deaf and hard of hearing individuals by failing to provide equal access in the form of captions Still waiting on a decision Outcome will have huge implications for higher education.
• #22: 2010 Previously aired on television Clips – straight lift & montages as of January
• #23: Because the laws were written before the proliferation of online video, we must turn to the courts to see what they believe about how the ADA and Section 504 should apply to the online sector: Read quotes
• #24: Captioning/transcription/subtitling (soon! Audio description) company Out of MIT in 2007 Still based in Boston
• #25: 2,000+ customers in higher education, government, corporate, & media
• #26: Our goal is to make captioning easy, which is why we offer: Easy to use account system Fast & reliable turnaround options Automated workflows 50+ download formats Caption import; transcript alignment Editing interface Video search tools Integrations
• #27: Integrations are a key piece of making captioning easy Tag video > we’ll process it > play your video with captions Most leading video players and platforms
• #28: Multi-step human review process: More efficient, cost-effective At least 99% accuracy All work done in U.S. Consistent transcription standards
• #29: Our editors are all US-based and come from all different backgrounds. They go through a rigorous certification process.
• #30: https://infinitehistory.mit.edu/video/cynthia-breazeal-ms-93-scd-00