3 ms·
When a student is paying for an education in a federally-funded institution, it's reasonable to expect video-captioning, braille, text-scalable HTML (not PDF) w
by westurner 10y ago
When a student is paying for an education in a federally-funded institution, it's reasonable to expect video-captioning, braille, text-scalable HTML (not PDF) wherever feasible. What about Sign Language interpretation? Simple English?
It would be great if everyone could afford to offer accessible content.
Maybe, instead of paying instructors, all lectures should be typed verbarim - in advance - and delivered by Text-to-Speech software (with gestural scripting and intonation). All in the same voice.
- Ahead-of-time lecture scripts could be used to help improve automated speech recognition accuracy.
- Provide additional support for paid captioning
-- Tools
-- Labor
- Provide support for crowdsourced captioning services
-- Feature: Upvote to prioritize
-- Feature: Flag as garbled
- Develop video-platform-agnostic transcription software (and make it available for free)
-- Desktop offline Speech-to-Text
-- Mobile offline Speech-to-Text
-- Speaker-specific language model training
- Require use of a video-platform with support for automated transcription
-- YouTube
--
- Companies with research in this space:
-- Speech Recognition, [Automated] Transcription, Autocomplete hinting for [Crowd-sourced] captioning
-- IBM
-- Google
--- YouTube has automated transcription
--- Google Voice supports transcription corrections, but AFAIU it's not speaker-specific
-- Baidu
-- Nuance (Dragon,)
--
... Textual lecture transcriptions are useful for everyone; because Ctrl-F to search.
- Label (with RDFa structured data) accessible content to make it easy to find
-- Schema.org accessibility structured data (for Places, Events)
--- https://github.com/schemaorg/schemaorg/issues/254 https://github.com/schemaorg/schemaorg/issues/254
--- http://schema.org/accessibilityFeature http://schema.org/accessibilityFeature
--- http://schema.org/accessibilityPhysicalFeature http://schema.org/accessibilityPhysicalFeature and/or
--- http://schema.org/amenityFeature http://schema.org/amenityFeature
-- http://schema.org/Course http://schema.org/Course
--- https://github.com/schemaorg/schemaorg/issues/254 https://github.com/schemaorg/schemaorg/issues/254
- Challenges
-- Funding
-- CPU Time
-- Error Rate
-- Mitigating spam and vandalism
-- Human-verified crowdsourced corrections can/could be used to train recognizing and generative speaker-specific models
-- In the film A.I. (2001), there's a scene where they're asking questions of Robin Williams and the intonation/inflection inadvertantly wastes one of their 3 wishes / requests. https://en.wikipedia.org/wiki/A.I._Artificial_Intelligence https://en.wikipedia.org/wiki/A.I._Artificial_Intelligence
- bobwaycott 10y ago> When a student is paying for an education in a federally-funded institution... That's not what we are talking about.
- westurner 10y ago> [Evasive legalism (Obligations of accepting federal funding, definition of service, funding differentiation, policy harms)] [Solutions for solving the problem (providing transcripts of lectures) most cost-efficiently]
- westurner 10y ago> - Companies with research in this space: - Microsoft -- "conversational speech recognition" - Apple