From time to time, I encounter something we call negative testing. This is usually performed during quality assurance testing or user acceptance testing. Negative testing is usually performed to try to make the system fail.
Text-to-speech (TTS) technology has come a long way. For old timers, gone are the days of the drunk-sounding TTS playbacks. Nowadays, the technology has improved to the point that, when done well, it is difficult to tell the difference between TTS and a recorded prompt. This technology is fantastic for cost effectiveness and speed-to-market, but depending on what you need to play back, some tweaking may be in order to ensure success.
Here are some of the issues that I have addressed when reviewing various TTS implementations:
Misspellings can be done by agents, customers entering items incorrectly over the Web, or intentionally by a speech technologist to try to get a word to sound a certain way.
Submitted by Amy Goodwin ¿Mande? That might be what someone from Mexico would say when they listen to a Spanish IVR. It means “Pardon me?” in Mexico but not necessarily in other Spanish countries. If you look up “mande” in …
Submitted by Amy Goodwin Anytime someone dabbles in the telephony speech world they hear the word “grammar” bantered around. But what is a grammar, and why should you care? A grammar begins as a text file that defines the range …