The Skills tab allows you to view and edit the content of your bot. The page is composed of 4 sections:
- The Skills List (with its search field)
- Folders (which allow you to group skills)
- The buttons to add a skill or to import content from an Excel file
- On the right, there is an area that allows you to test each skill
# The Skills List
Displays all the skills of your chatbot. Each of them has the following elements:
A colored border located on the left edge that allows you to quickly recognize the type of skill:
- Yellow: Scenario type skill
- Blue: Action type skill
- Pink: FAQ type skill
A handle allowing you to click and drop the skill to change its position in the list.
The central part, with the title (which can be modified directly here), the number of Resources that make up the skill, the date of the last edition and finally the description of the skill (which can be modified from the Settings tab within the skill).
On the right hand side, a group of buttons for (ℹ️ the use of these buttons is detailed below):
- Go to the skill's edit page
- Choose a folder for this skill
- Instantly send the skill to all users of the bot
- Launch the test of the skill
# Actions that can be performed from an item in the Skills list
# ✏️ Editing the skill
To edit the details of a skill (its resources, parameters, ...). See the dedicated section below for more details on this.
# 📁 Adding the skill to a file
Opens the contextual menu to place the skill in a folder In the input field, you can either enter a previously created folder, or enter the name of a new folder to be created, and then press "Enter". The skill is then added to this folder. You may place the skill in several folders by choosing/adding as many as you want in the input field.
To remove the skill from a folder, click on the cross to the right of the folder name in the input field.
The list of folders is displayed above the search field at the top of the page.
By clicking on one of the folders in the list, the list of skills it contains is displayed.
The button ➕ placed at the end of the list of folders allows to create a new folder.
When you right-click on one of the folders, a contextual menu is displayed allowing you to :
- Rename: to change the name of a folder
- Move: to move the folder to another folder (to obtain a hierarchy of folders and subfolders). ⚠️ ➿ Thus, a folder can belong to several parent folders ➿
- Delete: delete the folder. ℹ️ If the folder is the last parent of one or more folders, then these folders become "independent" folders, i.e. without a parent, i.e. they will appear directly in the folder section when you arrive on this Skills page.
⏪ You can quickly go up in the hierarchy of your folders and subfolders via the breadcrumb trail provided for this purpose :
# ✈️ Sending a skill to all bot respondents
Pushes the skill to all bot respondents (i.e. all people who have already interacted with the bot).
After validating the operation via the window that appears, the first resource (or at least the relevant resource, if you have set up filters or other logic on this skill) will be sent to all respondents regardless of the platform they use (web, Teams, Slack, ...).
ℹ️ Respondents in Live mode will be switched back to Bot mode. ⚠️ On the other hand, the status (attribute) isWaitingForLiveChat will not be modified.
# ▶️ Skills Test
Start the skills test in the test area (on the right of the skills list); it is up to you to respond to your bot to see if the skill in question meets your expectations.
# Types of skills
At Vizir, skills have a type. These types allow us to prepare a certain number of elements and to simplify your life! There are 2 main settings that we vary :
Can the logic be automatic or manual by default?
Should we ask the question again if we already have the answer to the question?
Remember to choose the type of skill when creating the skill. This will make your life much easier. Bugs may occur if you change the skill type after creating it. So read this article carefully 😉
A difficult choice awaits you! :
# The scenario type skill
They can be identified by their yellow border.
Here the logic is manual. And we ask the question again even if we already have the answer. So it is a free skill mode. We use it for example for all the skills of introduction, menu, incomprehension etc... It is also used to propose step by step resolution guides (selfcare).
# The action type skill
They can be identified by their blue border. Here the logic is automatic. You don't ask the question again if you already have the answer.
Why? Simply because action-type skills are used to:
- collect information
- to send them by API,
- and return a result to the user (execution confirmation, search result,...)
# The FAQ type skill
They can be identified by their pink border. Here the logic is automatic. We ask the question again if we already have the answer.
As its name indicates, a FAQ skill is used to answer questions. So the skill is organized by content nested one in the other. To make your editing life easier, the links between the resources are made automatically according to the organization you have chosen.
# Create an FAQ skill
This guide is only valid for the "New FAQ". Make sure that the "New FAQ" option is activated as shown in the image below.
# FAQ concepts
The FAQs can be visualized as a mindmap. That is, we have "nodes" with children. Let's take the example below to explain this logic
In the example below of a very simple FAQ on some applications, we can see the advantages of this mindmap structure:
- The resource
sharepoint passwordis "accessible" in several ways. This means that the user can reset their password through the
FAQ > Sharepoint problems > Password(First Case) or directly
reset password > Sharepoint(Second Case)
- 1 resource deals with 1 topic. It is positioned in the different nodes that represent the different possible logics of conversation.
- Mix the power of NLU (intents/entities) and the power of Elastic Search to always bring a reliable answer to your users (we'll get into the details later 😊 )
Let's take these two conversations to understand the two case possibilities.
# First Case
# Second Case
# Add Resources
You can now add your resources. You will then see an interface very similar to the other skills but with additional buttons as below:
To organize the resources as we saw in the previous step you can "arrange" them in nodes. In our example we get a structure like below:
For each resource add the keywords that will be used to determine the best resource to return when the user asks a question.
# Setting up the NLU so that the FAQ type will be effective
You've added all your FAQ topics, now let's see how to make the chatbot's response consistent with your users' requests.
NLU works very well when extracting known entities (like application names here) and intentions when the meaning (intention) of the sentence is very clear (like "change my password").
Now, we will use these two advantages ()of the FAQ type (entity/intent pairs) to create the NLU.
Two intentions are created as below:
That we train the intentions with one of the sentences in which we highlight the entity (application):
Then we create two configurations as below:
ResetPassword configuration will return us to the
reset password node
And the configuration
Logic without Intent- Entity (application) will return 3 different nodes depending on the value of the entity, as can be seen below:
Leaving the default logic empty allows not to use this configuration if the entity found does not have an expected value (first red frame in the image below)
Here, we organize a configuration priority for the entities 👉 Learn more here
# Examples of User Requests
Let's look at some examples of user requests to understand why this configuration priority structure works:
|1||I want to change my Slack password||ResetPassword||Slack (application)||Logic without intention + Entity application = Slack|
|2||I want to create a Sharepoint team||Other||Sharepoint (application)||Logic without intention + Entity application = Sharepoint|
|3||I want to change my SAP password||ResetPassword||SAP (application)||Logic is |
# Case 1
This is a very simple example where everything happens as we expected. The chatbot will search the Slack resource node for the best resource for passwords.
We are not going to use Slack as a keyword because we consider with the entity-based NLU logic that this part of the query has already had an impact on the logic.
# Case 2
Case 2 is very similar to the first case. Here we simply want to highlight that there is no need for intention for the NLU to work well.
In the case of creating a team it can be very difficult to train an intention on the subject because the training sentences for
creating a team will be very close to the training sentences for
creating a contact in the CRM or
creating a support ticket even though the 3 subjects are very different. It is therefore better, in this case, to use entity-based logic.
# Case 3
This time, we do not use the logic according to the value of the entities because we have not defined a logic for SAP. So we use the intention logic that exists in this case.
The chatbot will then search in the set of resources of the node
Change password for the best resource to answer this request.
In this case, we have 3 resources in the node
Change password which are:
- Change Slack password
- Change Sharepoint password
- Change Teams password
To make sure that the chatbot does not return any of these 3 resources, once we have done the keyword search of the node resources, we compare the scores and the retrieved keywords.
In our case, we realize that the keywords are very similar (within 1 word) and that all scores are ranked similarly.
Therefore, we estimate that none of the child nodes' resources is better than the parent node resource. Therefore, the outcome is the parent node resource
# Launch a workflow before a skill
Sometimes you will want to retrieve information before the start of the skill.
For example in a chatbot, it is possible to retrieve dynamically the IT ticket categories directly in your tool (Easy Vista, Service Now, Github, ...) so that this information is always up-to-date.
To do this you can follow the guide here