Hybrid Spoken/Written Documents

Integrating Podcasts into the Web

Ryan Shaw & Dan Perkel

UC Berkeley SIMS

What makes audio documents good?

Can be consumed while doing something else

Working with headphones on

What makes audio documents good?

Simultaneous group consumption: listening together

Kids listening to the radio

What makes audio documents good?

What makes text documents good?

What makes web documents good?

Hyperlinking

Glowing spiderweb

What makes web documents good?

Keyword search

Man with magnifying glass

Cross-breeding audio and text for the web

Cross-breeding audio and text for the web

Mockup of proposed interface

Motivation: the podcast explosion

The current podcasting model

Drawbacks of the current model

Proposed solution

Use case

Use case: The Leaky Cauldron PotterCast

Sue produces the news
Walk through the production of Leaky Cauldron, starting with the news segment.
  1. Records her voice
  2. Gets uploaded to server (recording to Flash Communication Server, or uploading an audio file using a webform, or FTP, or even calling in to some place).
  3. speech-text processing (for example: WizzScribe or Sphinx) - repeat point about not requiring great speech-text... that's the whole point - returns an xml file transcription synchronized with audio (we'll reveal more later).
  4. File is styled to present a visual editing interface in a browser, reminiscent of a text editor. 2 classes of editing - things that just affect the textual representation and things that affect the composite document as a whole.
  5. She corrects text, maybe adds some punctuation or extra paragraph marks.
  6. She adds the audio clip about the news to introduce the thinger.
  7. She makes it public (which could mean a few different things, but not relevant here).

Use case: The Leaky Cauldron PotterCast

Collaborative process
Sue's document is then included in the whole episode of PotterCast:
  1. Person makes file in similar manner
  2. Includes previous file
  3. Makes public

Milestones

Data model requirements

  • Point about needing to consider interaction and presentation when coming up with the model.
  • None of the existing standards seem to meet these requirements.

Styling requirements

Interaction requirements

Current status