Parloa Features

#1 Use Contextual Listening

Your users will differ in their speaking rate and pause placement (e.g., when naming large numbers). In order to have the lowest possible error rate here, we invented "Continue Listening". Use this block to check the relevance of the input already at the first End of Speech detection, without users noticing. This helps e.g. to check if the spoken number matches an expected format or to check in the background if the input matches a database.

Example

Bot: “What is your customer number?”

User: “It’s 876...”

→ End of Speech is triggered but “Continue Listening” awaits 8 digits

User: “...5181”

→ End of Speech is triggered again and “Continue Listening” combines both inputs

Bot: “Let me repeat: 8765181. Is that correct?”

#2 Use DTMF as supplement in fallbacks & reprompts

Thanks to DTMF (dual-tone multi-frequency), your users are able to make use of the digits on their phones when it comes to submitting numbers. Of course you can also use this function as a fallback for simple yes/no queries or assign state-specific content to it.

To learn more about adding DTMF intent & slot to your dialog, please have a look at our documentation.

Recommendation: Actively point out the possibility of DTMF input in the reprompt or fallback and either present it as an additional option or even explicitly require it to minimize misunderstandings (although you should of course still always allow voice input!).

Example

Bot: “What is your customer number?”

User: “It’s 8765181”

→ Service received no match of the customer number due to wrong Speech-To-Text: “ey 765181

Bot: "To play it safe: Can you enter your zip code again using the keys on your phone, please?”

User enters: “8765181”

Idea

Ever heard of vanity numbers? They are the reason for the groups of three letters under the digits of most phones in this world. This refers to phone numbers that can be better used for advertising, for example, with words instead of numbers. So 0800 PARLOA would then be 0800 727562. Maybe your use case allows you to use this feature?...

#3 Use Speech-To-Text hints

With Parloa you have the possibility to prime the Speech-To-Text technology state-specific. This is very useful, for example, to read the customer name from the database in the course of authentication on the basis of a recognized telephone number or a spoken customer number and to expect it in the next state when asked. This minimizes misunderstandings many times over! Here you can find an example service for using Speech To Text Hints.

Example

Bot: “What is your postcode?”

User: “It’s 10178”

→ Based on an external address service, Speech-To-Text-Hints are served with all street names in this area

Bot: "10178, got it. And in which street do you live and what is the house number?”

User: “Dircksenstraße 5”

→ Speech-To-Text transcribes the street correctly

#4 Scribble your dialog with subdialog blocks

We invented reusable Subdialogs to help you keep track of your graph and avoid maintaining dialogs and logics twice.

But they also have another important aspect:

You can use Subdialogs in Parloa to completely illustrate your dialog structure in a short amount of time without going into detail yet!

Simply create Subdialogs that are as distinct as possible, define all associated Entrances and Exits, but do not connect any other blocks within the Subdialogs.

This will give you a very good overview of what is needed for your dialog and which scenarios are all conceivable in a very short time.

Of course, the dialog is not yet ready for use, but you create work packages and can, for example, go through the dialog with colleagues and superiors, at least visually.

Example: