WikiTranscript: Difference between revisions

From OLPC
Jump to navigation Jump to search
mNo edit summary
mNo edit summary
Line 1: Line 1:



__NOTOC__
__NOTOC__
Line 15: Line 14:


== Project description ==
== Project description ==
{{draft}}


Note that this is a very technology-heavy idea, and will need some tech-hackers to work on the code implementation. In the meantime, it should be possible to create the wiki and start adding transcripts by hand.
There is no set format for this section, and it can be as long or as short as you want. What do you want to do, why do you want to do it, and how will it happen? What kinds of people would be good at working on this? How will you know if you've succeeded? What similar things have been done before? You can supplement (and should) your application with external links to helpful resources, but a reader should be able to understand the project you are proposing even if they only read the abstract you've written here.

The problem: Audio on the web is not accessible to people with hearing disabilities or those who speak a language other than the one recorded.

The idea: A website where you can type in the url of any audio or video file and see the transcript for that file as the file itself plays. For instance, you enter the url of a YouTube video which has been filmed in Urdu, but you speak Portuguese; the video plays in one half of your browser and the transcript (translated to Portuguese) is displayed in the other half.

The catch: The "database" is actually a wiki, with each transcript as a separate wikipage. That way, people can edit, add to, and correct transcripts, translate them into different languages, and merge/redirect transcripts of identical files. When someone enters a file url into the WikiTranscript site, a few things happen:

* The site checks to see if the transcript is already in its database. If so, it directs the person to that transcript and finishes. If not...
* The site checks whether the file is in an audio or video format that is readable to WikiTranscript, and that the audio quality is sufficient to start a machine translation. If so, it passes the audio through a speech recognition engine and outputs the results on a wikipage, then directs the person to that transcript. If not, it displays a "sorry, I can't understand this file either" message and offers the option to try doing a manual transcription, or to request someone (from a pool of volunteers) to do a manual transcription.

The technologies: Probably using Mediawiki for the wiki engine. [[User:Mchua|Mel Chua]] is looking for people to experiment with various open source speech recognition engines to see if the transcription quality is adequate.


== Interested interns ==
== Interested interns ==
Line 26: Line 37:
== Interested mentors ==
== Interested mentors ==


=== Mentor name ===
=== Mel Chua ===


Contact info on [[User:Mchua|userpage]]. I've got a personal interest in this project, since I'm hearing-impaired and frustrated by the amount of content on the web (podcasts, videos) that's utterly inaccessible to me. Have experimented a little with open-source speech recognition engines in the past, can help with technological aspects of the project as well as community-building.
Contact information, why you'd be good for the job, any specific plans, variants, or details you would personally like to implement and why


== Other interested parties ==
== Other interested parties ==

Revision as of 01:03, 23 July 2007

 

  • Interns - If you are interested in this project, add your name to the Interested interns section below along with a brief description of why you're interested and why you'd be a good mentor for this project, along with any specific ideas for execution you might have beyond the project description.
  • Mentors - If you are interested in this project, add your name to the Interested mentors section below along with a brief description of why you're interested and why you'd be a good mentor for this project, along with any specific ideas for execution you might have beyond the project description.
  • Others - If you are interested in this project in a role other than that of potential mentor or potential intern (example: you are an organization, a potential end-user/tester, may have helpful resources, or want to be notified if the project is chosen), add your name to the Other interested parties section below with contact information and details.
  • Everyone - Contribute to the project description on this page, or discuss this project on the associated talk page.

The deadline for editing this proposal or adding yourself to the list is 11:59pm EST (GMT-5) on August 6, 2007.

Project description

Pencil.png NOTE: The contents of this page are not set in stone, and are subject to change!

This page is a draft in active flux ...
Please leave suggestions on the talk page.

Pencil.png

Note that this is a very technology-heavy idea, and will need some tech-hackers to work on the code implementation. In the meantime, it should be possible to create the wiki and start adding transcripts by hand.

The problem: Audio on the web is not accessible to people with hearing disabilities or those who speak a language other than the one recorded.

The idea: A website where you can type in the url of any audio or video file and see the transcript for that file as the file itself plays. For instance, you enter the url of a YouTube video which has been filmed in Urdu, but you speak Portuguese; the video plays in one half of your browser and the transcript (translated to Portuguese) is displayed in the other half.

The catch: The "database" is actually a wiki, with each transcript as a separate wikipage. That way, people can edit, add to, and correct transcripts, translate them into different languages, and merge/redirect transcripts of identical files. When someone enters a file url into the WikiTranscript site, a few things happen:

  • The site checks to see if the transcript is already in its database. If so, it directs the person to that transcript and finishes. If not...
  • The site checks whether the file is in an audio or video format that is readable to WikiTranscript, and that the audio quality is sufficient to start a machine translation. If so, it passes the audio through a speech recognition engine and outputs the results on a wikipage, then directs the person to that transcript. If not, it displays a "sorry, I can't understand this file either" message and offers the option to try doing a manual transcription, or to request someone (from a pool of volunteers) to do a manual transcription.

The technologies: Probably using Mediawiki for the wiki engine. Mel Chua is looking for people to experiment with various open source speech recognition engines to see if the transcription quality is adequate.

Interested interns

Intern name

Contact information, why you'd be good for the job, any specific plans, variants, or details you would personally like to implement and why

Interested mentors

Mel Chua

Contact info on userpage. I've got a personal interest in this project, since I'm hearing-impaired and frustrated by the amount of content on the web (podcasts, videos) that's utterly inaccessible to me. Have experimented a little with open-source speech recognition engines in the past, can help with technological aspects of the project as well as community-building.

Other interested parties

Name

Contact information, what your interest is