How May I Help You?

Lots of modern speech systems include some way to let the users say anything, and let the computer figure out what to do next. It's not really dictation, which is simply trying to understand and transcribe the exact words that were spoken. Instead, the idea is to try to pick out, based on specific words that the user said,

This is surprisingly effective in lots of interesting scenarios. Take the following examples:

• "what's my bill?"
• "Tell me my bill for last month".
• "Could I speak with Billing please"
• "I'm looking for a way to pay my bill"
• "I'm going to be late with my next payment".

In each case I don't really care the exact content of all the words, but I just want to have the computer transfer my call to the billing department.

MSR has been working on this problem for a while too and I just saw a demo. Pretty cool!