Much has been made of whether podcasting can reach out to users with accessibility issues. We have come to accept the medium as being very helpful to those with visual impairments but of no benefit to users with hearing impairments. Fortunately, methods like transcripts and subtitling are beginning to make the medium much more accessible to these alienated users.
Podzinger and speech recognition
There is huge interest in the potential of speech recognition technologies for users with hearing impairments. While those of us with an eye on accessibility are familiar with screen-readers that generate speech from text, these new technologies provide a text transcript from audio recordings. Some governments currently use these technologies to monitor international, particularly Middle Eastern, radio broadcasts for threatening news.
In late 2005, BBN technologies, a company that specialises in speech recognition technology launched a beta version of Podzinger, a podcast search engine with a difference.
Instead of the traditional methods of simply crawling through the description and title contained in the xml file, Podzinger uses speech recognition to prepare text transcripts of podcasts registered on the site. It then searches through the xml file and the transcripts to provide more relevant search results. When using the Podzinger search engine, you get an excerpt from the transcript and can click on any word and the podcast will play from there.
Linguist Mark Liberman discusses Pozindger’s “lack of appropriate punctuation and capitalisation” but commends the recent developments in the field of speech-to-text (STT) programming. At present the systems tend to rely on a “language model”, which creates statistical algorithms to try to predict what words and phrases will be used by analysing existing podcasts.
Traditional issues such as poor quality audio recordings and local dialects or accents still provide the biggest barrier to fully verified podcast transcripts. At present for copyright and protection purposes, Podzinger does not allow you to access the full podcast transcript.
While systems like Podzinger aren’t of direct benefit right now to those with access issues, it hints that we are very close to a beneficial workable system. This system could allow a user to view full text transcripts of any podcast of his/her choosing. If a process were ever made available to the public that could provide accurate STT, then it truly would bring accessible podcasting to the hearing impaired.
Manual transcribing
At present there are a small number of dedicated people who manually provide transcripts for their podcasts and offer them side by side with the link to the audio version. Accessibility blogger Andy Carvin feels that the notion of typing up a transcript of your podcast is “counter to the ethos of podcasting” and personal multimedia publishing. He notes even “a detailed summary is better than nothing. Otherwise, podcasting will be yet another media juggernaut that will zoom by the lives of millions of people without giving any of them a chance to benefit from it as well.”
Unfortunately most providers of podcasts are not aware of the varied benefits of providing a transcript to their podcast content. The accessibility advantages have been covered but it would also allow people, be they users with slow connections or simply those who prefer to read than listen, to have their own text printout of the podcast.
Subtitling and Captions
While taking the audio aspect of the podcast out of the equation entirely is one option, sometimes those with partial hearing may still want to try to listen to the podcast on their player. Here they could use subtitling as an aid to the listening experience. SMIL is one method of synchronising a captions or subtitles file with an mp3 file. This of course would only be compatible on a computer using the relevant software, and would be incompatible with all current portable players.
One advantage of the recent players like the Zen Vision: M and the iPod G5 is their ability to show a visual description of the audio. The Vision: M’s default software will automatically create styled motion jpegs from the information in the xml file. Captions with headings and summaries of each episode are created.
If more detailed meta data was to be used and if a transcript could be provided, then a user with access difficulties could have a live textual aid to the podcast. Hopefully, with the advances in STT technologies, we could soon have a program process that would go through the downloaded podcast, prepare a text transcript and then synchronise the two for presentation on the portable player in audio and text format.
As more and more podcast directories are offered to users, including those looking at the medium through an Irish perspective like podcastingireland.ie, lets hope developers consider the benefits of keeping accessibility in mind when offering podcasts. Advances in technology and awareness mean that hopefully those with hearing impairments will soon be able to fully reap the benefits of podcasting just as those with visual impairments have.
Tags: accessibility, captions, podcast, subtitles, transcription
Your article somewhat misinterprets what I was saying about podcasting and accessibility. I personally do not believe that transcripts run against the ethos of podcasting. If you read what I wrote, you will see I was attributing this attitude to other podcasters who ignore accessibility issues or aren’t familiar with accessibility. Quoting from my transcript:
As you can see here, I’m suggesting that this is a view of of many podcasters, some of whom I go on to cite in my article. The point of this was to suggest that transcripts should not go against the ethos of podcasting, not the other way around. While transcripts aren’t always possible for near-real-time podcasts, it makes sense for producers of regularly scheduled podcasts to provide a transcript, or at least a detailed summary.
[...] Theres an interesting blog entry over at Accessibility.ie that mentions this very subject. Posted by Sean Nicholls in Podcasting [...]