I have a Voice-Over-IP (VOIP) phone system in my home in addition to my cell. This is of course one of the clearest indicators that I am indeed “Old”. The VOIP number really doesn’t get a whole lot of use. Every year around this time, I’ll noodle on whether I really need to carry this line item forward into next year’s budget. The chief reason I keep it is for the infrequent occasions where (for whatever reasons) I need to take work-related conference calls at home. I think I have worked from home a grand total of 1.5 days in 2012, so the cost/benefit equation is probably heavily skewed toward the cost side. The VOIP line is provided via Vonage.
One unintended side benefit I have derived from the VOIP line over the last few months is the (somewhat juvenile) pleasure I get from reading the various butchered translations of political advertisements left as messages and converted to text by Vonage’s Visual Voicemail software. Of course this pleasure will all come to an end on Wednesday. As with most of the VOIP systems, Vonage can be configured to send email to you when messages are left by callers. The email will include an attachment that includes the encoded audio file for the message (to which you can listen) and also the text of the message (output from the Vonage Visual Voicemail speech recognition software).
Now in no way am I throwing stones at the Visual Voicemail software. There are a lot of challenges in speech recognition software based on size of the vocabulary (small vs large), whether the system is used by an arbitrarily large number users or a small set of finite users (whereby the system can be “trained” for those users), the quality of the sound wave captured for analysis (background noise) and so on. The current systems work well when used for a “small vocabulary/large number of users” (for example, calling in an Internet outage to Time Warner Cable or requesting the current balance of your bank account) or “large vocabulary/small number of users” (where the system can be tailored to the individual user’s speech patterns). Vonage Visual Voicemail, on the other hand, must deal with a large vocabulary and a large number of users (that don’t get to train the system). It also has to support continuous speech, dialects and accents, etc. It is a pretty tough nut to crack.
Eventually, Moore’s Law will no doubt provide relief here. In the meantime, I have compiled a couple of interesting (to me at least) translations of political messages left on my Vonage line over the last couple of months. In particular, the Vonage software seems to be struggling a bit in translating Governor Romney’s name correctly.
Who is Justin? You talking to me? Invoking President “Ronnie” Reagan probably a good move to get out the base. “We gonna do” – sounds like some Dock Ellis there. Interesting how the three occurrences of the phrase “We don’t have to settle” were translated. What in the heck is this “shuttle for unemployment”? I thought we got rid of that wasteful NASA crap. “I’ll get her incomes up” – I’ll bet you say that to all the girls Governor. “I service governor and she’ll like to get them done”. Hmmm, does Ann know about that?
