Closed Captions

Closed captions are not subtitles. Subtitles are displayed on the video and cannot be turned off. Some might say they are "burned" or "baked" into a video frame. They cannot be removed.

Subtitles provide information typically every audience member would need to know:

• a translation of a foreign language to that audience, e.g., English speaking audience getting a translation of Spanish dialogue
• a translation of a fictional language, e.g., audience getting a translation of alien dialogue

Subtitles are incredibly useful when the volume cannot be raised. Yet, subtitles are still not visible to people with visual disabilities. They are not text, since they are part of the visual. If subtitles are used, provide a transcript or audio description.

Various video hosting solutions use the term "subtitles" to say captions are present:

• Microsoft PowerPoint
  • Allows presenters to turn on "subtitles" during slideshows
  • They are open captions that the audience cannot turn off during the presentation
• YouTube
  • Provides "subtitles/closed captions" on videos
  • They are closed captions (CC) which can be turned off

Open captions are text alternatives for audio content. Open captions behave just like subtitles in that they cannot be turned off. This term is used more often for real-time captioning.

Open captions benefit everyone, including those with hearing or cognitive disabilities. Yet, open captions can accidentally cover presentation content.

This can be solved by software that allows the captions to appear on a separate window. That separate window may be a challenge, though. Some people may have problems balancing different windows or browsers for viewing the presentation.

Artificial intelligence (AI) engines listen to what is said and try to translate it to text.

The speech translation part is called Automatic Speech Recognition (ASR). It is what runs speech-to-text software.

It is not entirely accurate. That is because ASR alone does not recognize the context of the speech pattern.

Was it supposed to be "2", "to", or "two"?

The context translation part is called Natural Language Understanding (NLU). The AI engine looks at the words near the one that sounds like "2", "to", or "two".

Was it a number? Was it part of a phone number or address? Was it directing someone?

Right now, it is easier for a human to understand context than an AI engine. Manual, or human, quality assessment fixes problems AI cannot perfectly solve yet.

Misunderstood words may accidentally display:

• similar sounding words,
• inappropriate or offensive words, or 
• wrong spellings of words.

Misunderstood context may cause:

• telephone numbers to display as words or
• email addresses to display the words "at" and "dot" instead of punctuation marks "@" and ".".

• missing periods and capitalization to stop and start new thoughts or
• missing commas or other punctuation to organize thoughts.

Additionally, AI does not know the spelling of everyone's name. Names are uniquely spelled, and AI would need the spellings ahead of time. Web conferencing tools solve that problem for people logged in who speak. Make sure people spell their names correctly on their profiles.

Check how the words appear to tell if a video was auto-captioned. If words appear one or two at a time, it is likely auto-captioned. If sections or lines of words appear at a time, it may have been through quality assessment.

It is important to fix auto-captioned videos. If you are using an auto-captioned video, contact the video's owner. Ask if they have a transcript or can create the captions themselves. If they do not respond, you may need to transcribe the video instead. Provide that transcript to students requesting ADA accommodations. This should be done ahead of time, so students can interact and take part in real-time.