Friday, April 28, 2006

The E-volution of Voicemail

Remember the time that email used to be just for conveying short messages, to be deleted when done reading? For many of us email has become much more than merely a quick communication mechanism. It is not uncommon to find inboxes with 1000s of email messages going back months, if not years. Free email service providers have had to offer larger and larger amounts of storage space to keep up with this trend. Some even offer unlimited space, so that you "never have to throw anything away". As storing email messages rather than deleting them has become more and more feasible and convenient, email inboxes have become repositories for important, and oftentimes even critical information that is not stored anywhere else.

Search technologies have evolved to accommodate the needs of this newly habituated generation of email users. You find such search technologies embedded in email clients, available as add-ons, or in the "search anything on your machine" applications that index your email messages alongside your files. This has served to further reinforce the use of email as a convenient place for leaving information that can be easily and quickly retrieved when required. Some users even deliberately email information to themselves so that it is available in the same trusty repository, accessible from virtually anywhere.

Enough said about email. What about voicemail? Technically, voicemail has been around for a while, but has remained virtually unchanged in terms of the way people use it. One primary impediment to using voicemail as a repository of information is that most voicemail interfaces still only provide sequential access to the message list, which makes navigation prohibitively time consuming. In addition, information inside a voice message is accessed sequentially, that is, by playing the message from start to end. Some amount of efficiency is gained by hunting for specific information by skipping over irrelevant parts of the message.

With the advent of voicemail in VoIP phone services and software such as Skype, there is a great opportunity for voicemail to progress to the next step in its evolution, so to speak. The ability to search through voicemails based on the subject/content of voice messages would allow people to leave information stored in the form of voice messages.

UmeSkype provides the ability to search through your Skype voice mails by date and by caller. We have tried to make it easy for people to speak the queries by saying phrases like: "find voicemails from john last month" or "find voicemails from john on january fourth two thousand six". We hope that this is just a start to what can be done to realize the full potential of voicemail as a knowledge/information repository.

Monday, April 24, 2006

The Promise of Speech Recognition?

Many of us watched Captain Kirk or Scotty on Star trek ask the ship's computer "computer, where is the ship's engineer?", or Captain Picard tell his food synthesizer to make him an "Earl grey, hot!" How far are we from such a scenario?

Anandtech has a good article on the current state of the art as reflected by two of today's leading speech recognition application suites from the standpoint of accuracy and performance.

Recognizing speech accurately is one aspect of speech recognition, and "natural language understanding" is quite another. While we are still quite a long way from achieving true natural language understanding by computers, there are tremendous opportunities to create intuitive interfaces for humans to interact with computers.

UmeSkype embodies an attempt to identify high level tasks that users perform in specific applications (in this case Skype) and provide a spoken interface to them.

Friday, April 21, 2006

UmeSkype features overview part 1

Calling Skype contacts

When you start up UmeSkype, the contacts tab shows 2 columns:
the voice alias and the skype handle. The first time you run
UmeSkype, the voice aliases will all be identical to the skype
handles. This means that there could be voice aliases that
are difficult for you to say, as well as for the speech recognizer
to understand.



You can change any voice alias in windows explorer style
by "slow double-clicking" on an alias (i.e. click, wait, and click again),
and then typing in the new alias when it becomes editable.


You should use create voice aliases that are both
familiar to you and easy to pronounce for you.

Thursday, April 20, 2006

Welcome to the UmeSkype blog!

UmeVoice is proud to be part of the wave bringing practical
speech recognition applications into the world. Our aim is
to create natural, intuitive language that people can use
to communicate with applications on their computer.
UmeSkype is one product of this effort.

In this blog we intend to share and discuss UmeSkype's features,
future plans, user impressions, user requests, and anything
else that relates to UmeSkype, including speech recognition,
user interfaces, and accessories like noise cancelling headsets.

Hope to generate some interesting discussion!