Fine tuning
From Dell's One2One blog, a great example of speech application tuning:
In late May, we made changes to our IVR (Interactive voice recognition) tool to enhance the usability by simplifying menus and improving the verbiage. In the menu selection process, we eliminated unnecessary prompts and have reduced the average time customers spend in the IVR by a full minute. We listened in to the process and learned that we often use Dell-speak rather than language that is more intuitive for you—we think PowerEdge, you think servers. We are tuning our speech recognition software so that it recognizes a broader range of words in order to route you more effectively. For example, our Inspiron and Latitude product lines can now be recognized as laptop, notebook, portable, etc..., this process never really ends and we will keep looking for new approaches.
Hats off to the people who found and fixed these problems. Especially the "Dell-speak" issue - we think PowerEdge, you think servers - that's a classic symptom of a user interface built off a data model rather than the user perspective. You see it all the time, not just in voice UIs.
My favorite bit: this process never really ends. I'll drink to that. Tuning is forever...
