Applications that utilize speech recognition are capable of highly interactive dialogs, relegating “press or say” alternatives an interface of the past. There are 2 basic categories of speech applications: Directed Dialog and Natural Language.

Directed Dialog – Employs a user interface where a caller is guided to respond with words or phrases from a given range of selections. Designers of speech recognition applications must anticipate what callers will say and how they will say it in order to build a supporting grammar.

Example: You can say “slow internet”, “no internet connection”, etc.

Natural Language Understanding (NLU) – NLU is able to recognize what callers are saying as if they are speaking to another human. More open ended NLU dialogs are able to be understood by “training” a grammar. Tens to hundreds of thousands of utterances are transcribed and assigned meaning. A Statistical Language Model (SLM) is created to understand what callers mean.

Example: “Please tell me your problem.”

Does the freeform nature of Natural Language Understanding always make it the best choice? Interestingly enough, the answer in no. Directed Dialog can provide the direction necessary to more quickly accomplish a task for a caller when it makes sense. In some cases the best alternative is a hybrid model. Use NLU when callers have difficulty describing an issue, switch to Directed Dialog when a call resolution path becomes apparent.