Sun tackles privacy, speech recognition | CNET News.com
Both parts of this announcement are very interesting. While presence awareness software from Sun is not that surprising (they already have an instant messaging server and an identity management server), I’m surprised to see them getting involved in speech recognition.
Although the article doesn’t mention it, I assume Virsona is built on JXTA. This is an interesting approach to giving people more control over their presence info, as opposed to having it tracked on a server over which they have no control. I hinted at a blog-based approach in my notes from a recent Weblog panel at Cal. I think a combination of these two approaches would be very powerful.
I’m very excited to hear that Sun is getting serious about working with the CMU Sphinx project to create a high quality open source speech recognition engine in Java. Currently, it has only a 1,000 word vocabulary and will be speaker dependent, i.e, each speaker will have to go through a training period before the recognition level will be acceptable. However, this should be sufficient for at least a first stage of auto-transcription for PhoneBlogger.
First, I want to add support for simple titles to audio posts. Right now, the title and text for an audio post is exactly the same. Unless you later go back and edit the post once you have Internet connectivity from a text entry device, a reader of your blog can’t tell what the post is about without listening to it.
Of course, it wouldn’t hurt if I stopped reading news and blogging until past midnight and got working on at least adding support for Movable Type categories.