Speech recognition is a critical tool to optimize self-service channels, but has the opposite affect when customers just wind up lost and aggravated. If the platform asks the caller to say what they would like to do, it should understand the response. The .NET (News - Alert) VoIP Platform Voice Elements from Inventive Labs offers a better alternative.

According to this Inventive Labs report, a few rules should be a priority to ensure speech recognition works well. First, the .NET VoIP platform should include better instructions. It should provide users with clear instructions to better understand exactly what needs to happen. Open ended questions are a good option, as long as there is guidance. 

Longer words and phrases are also important and can be included in the user directory. It’s common for these systems to ask the customer to say the name of the individual with whom they wish to speak. The problem with this approach is the wealth of names that sound very similar. Think Larry, Marry, Harry….you get the idea.

Likewise, there may be multiple individuals at a company with the same first name. When limited information is provided, it can affect the performance of the platform, as well as the satisfaction of the user. If the platform is provided more source data in the interaction, such as a last name, there is a reduced chance for error.

The number of words in the grammar file should also be reduced. When there are too many options from which to choose, it can be more difficult for the system to understand what is being said. An example would be the speech recognition platform asking the caller to say where they would like to go on vacation. Such a broad open ended question can make it more difficult for the system to determine exactly what was said.

While it’s common for users to want to include as many grammar options as possible, starting simple is actually a better approach. And, users can easily generate files on the fly using the Microsoft (News - Alert) Speech Platform or with Lumenvox.

Finally, a number of users tend to be confused as to why they can’t call a company and expect it to work like Apple’s (News - Alert) Siri. While it would be great if .NET VoIP could operate in this manner, yet telephone systems have inherent limitations and don’t offer the high quality microphone built into the iPhone (News - Alert).

When used correctly, speech recognition can reduce the cost of customer service and improve the overall experience. Check out the guidelines above to be sure the design of your speech recognition platform performs as expected.