Intentions

An intention represents the purpose of the end user in their conversation turn.

It consists of the following information:

Training phrases	Training phrases are examples of what end users can write or say (user input). ChatFlow activates the intent when it detects that the user input resembles one of these phrases. It is not necessary to specify all possible variations as ChatFlow's built-in machine learning expands its list with other similar phrases.
Gaps	In training phrases, you can specify placeholders to extract parameters from specific parts of the user's text.

Intention Coincidence

When a user writes or says something, known as input, ChatFlow compares it with the intention training phrases to find the best match. This process is called intention matching. Intention matching can only happen for intentions associated with a route of the active page (see pages).

While searching for a matching intention, ChatFlow assigns a confidence value to the matches, also known as confidence score. These values range from 0.0 (completely unsure) to 1.0 (completely certain). Once the intentions are scored, there are three possible outcomes:

If the highest scoring intention has a confidence score greater than or equal to the set threshold, it is returned as a match.
If no intention reaches the threshold and there is a default route (not associated with any intention), this route will be invoked.
If no intention reaches the threshold and there is no default route, the user will be informed that they have not been understood with a default message.

Training phrases

Training phrases are examples of what end users may write or say (user input).

For each intention, you can create multiple training phrases. When an end user's input resembles one of these phrases, ChatFlow detects the intention.

For instance, the training phrase "I want material" will make ChatFlow recognize similar phrases such as "I need material" or "I want to order material".

You do not need all possible examples, as Dialogflow's built-in machine learning expands your list with other similar phrases. We recommend creating at least 10-20 (depending on the intention's complexity) training phrases so that the bot can recognize a variety of user inputs. For example, if you want to recognize an end user's intention to confirm an action, you can define the following training phrases:

"Yes"
"Perfect"
"Yes, I confirm it"
"Of course, yes"
...

Adding slots in training phrases

You control how user data is extracted by annotating parts of their training phrases and setting up associated entities.

For example, in the material request, we could have a phrase like "I need 3 pairs of gloves". If you want to extract both the quantity of material and the desired product, you should write training phrases similar to:

I need @number pairs of @product
I want @product
I want you to send me @number @product
...

When you annotate parts of a training phrase, Chatflow recognizes that these parts are just placeholders where the actual values provided by end users during the conversation will go. For an end user input like "I need 3 pairs of gloves," Chatflow would extract the number parameter as "3" and the product parameter as "gloves".

To annotate training phrases, simply type @ and a dropdown with all possible entity types will appear.

Cancellation intentions

During a conversation, it is possible that the end user may want to cancel the current conversation topic. For example, the currently active page may be requesting the desired amount of material but the user has ultimately decided not to request it. The end user can say something like "cancel" or "I don't want material". To handle this situation, you can create one or more cancellation intents on the page. You can name these cancellation intents whatever you like, but it is common to include "cancel" in the name. You should associate these cancellation intentions with routes of the page that is active when the user is writing. These routes should transition to a suitable page to handle the cancellation.

Training phrases for cancellation intentions should include both generic and topic-specific phrases. For example:

Cancel
Stop
Changed my mind
Never mind
I don't want material
Cancel material request
Delete request