Human’s mind is unpredictable and different from each other. Users might not always follow your designed flow. They might choose to disrupt it by asking a random question which is totally unrelated to the previous messages. A well-trained chatbot will handle more edge cases.
So Keyword is the feature that increases bot's ability to understand user’s inputs even with different types of wording.
The more set and phrases you add to your bot, the more it becomes smarter and Keyword will allow you to create a reply base on users' message, which can be F.A.Q or small talks. Thus it can trigger a block no matter where you are in the flow.
When you don't want the conversation between users and bot to stop with no response because users type some questions or sentences which disrupt your flow. At that time we will use Keyword feature to prevent that situation happen.
Basically, you will need to create a Keyword Set by click "New Keyword Set". In order to manage bot's Keyword, you can create more than one Keyword Set base on what users need. Then in that Keyword Set, you will create Keywords which have messages you predict users may type and responses you think users should receive according to the messages.
To make it easier to understand, we will guide you step by step to create a Keyword. There are five basic steps to create a Keyword, please follow the instructions as shown below:
Go to Buid -> Keyword.
Click on "New Keyword Set" to create a new Keyword Set.
We will name this Keyword Set is "Goodbye"
Select the type of Keyword you want to create.
After creating a New Keyword Set, you can choose Smart Keyword or Keyword to create.
The different between Keyword and Smart Keyword is that Smart Keyword get support from the third-party provider so that bot can understand much better to the same meaning inputs from users and provide the correct answers which weren’t designed in the flow. Keyword will provide the correct answer according to the rules you have trained to your bot.
We choose Keyword in this example. You will see there are three important areas that should be noticed.
Active (1): Active allows you to disable the Keywords for a period of time. Then enable it again when you want.
Rule (2): Rule includes possible messages that you predict users may type during the conversation with bot. The conditions in Keyword that allow you to train your bot to become smarter are message is, message is one of, message contains, message begins with, message is sticker. The conditions in Smart Keyword is message matches.
Response (3): Response allows you to control the answer that you want to give out to users base on the rule above you have set up.
Set up rule and response.
We think users will mostly type the message "Bye bye" and we set response to this situation is "Thank you for visiting our house, hope you have a nice time with us. See you again."
Here is the result:
Define new Keyword Set
set up new Keyword Sets
Keyword Set is a group of Keywords that have a common content or purpose. We recommend you to create set per topic/domain to increase the accuracy and bot’s learning speed.
Head up to the + button next to the language you want to create Keyword set or click on the New Keyword Set. There are two cases:
You only use Keyword and don't need to set up NLP Engine.
You use both Keyword and Smart Keyword so you need to set up NLP Engine.
Now fill up the following information:
Name: Give a name for Keyword Set.
Language: The language is used in the set. If your bot is multilingual, you should create the same sets in the languages are being used in your bot.
Minimum Confidence: The minimum ratio of confidence for analyzed results. There are three levels of minimum confidence: Low, Medium, Strict.
Choose Low option in case you expect that your bot can automatically give out the designated answer, even when user input's wording is different and matches just a little meaning with what you have trained your bot.
Choose Medium or Strict option if you want your bot to understand user input at a high level than low option among what you previously trained your bot. We recommend you to choose Medium for the best experience.
NLP Engine: This is the AI technology from the third-party that you choose to integrate with your bot. For now, BotStar supports you to integrate with Wit.ai. But in the future, you can integrate with another AI technology more than Wit.ai.
Create a new Keyword
Keyword helps you identify users messages then automatically send them a suitable answer according to the rule you have set up.
Set up Rule
There are five rules that you can consider to train your bot.
Message is (1): This rule will be activated only after the users type exactly the message you have sep up in this rule.
For example, the message must be typed exactly "hello" to be able to trigger the block welcome, bot can also recognize the insensitive case. But this rule only allows one message, you are not allowed to add another message to this rule.
Message is one of (2): This rule has the same function as Message is but the difference is that you are allowed to add-in more than one message to this rule.
For example, you can add more messages like "hi", "good day" beside "hello" and they will be triggered the response It is nice to meet you.
Message contains....and... (3): Bot can recognize more than one keyword in short/long sentences, but this Keyword will only be activated when users' message have all the keywords in this rule.
For example, users type "I want to have the menu about cake" and because the message contain two keywords menu and cake so they will be trigger the block amazing.
Message begins with (4): The message begins with the keyword you want will be triggered.
For example, users' message is "contact with admin" and because the message begins with "contact" so they will be triggered to the block contact.
Message is sticker (5): Currently, BotStar supports you send an automatic response when users send a thumbs up sticker offered by Messenger.
In this example, the response users will get is what do you want to ask us? when they use thumps up sticker.
You have to be careful with short words such as I, you because they can unintentionally match with Keyword which have those words and users may receive wrong answer base on the message.
Rule message is one of and message begins with allow more than one message in one rule and each message is separated by a comma(,). But Message is, Message contains...and... and message is sticker only allow one message in one rule.
Set up Response
To set up response you can choose one of two options which are Select a block or Write a response.
Select a block: you can set up a response by selecting an existing block for the bot to take the user back to the flow, then users can continue the conversation from that block.
Click on Select a block.
Choose the existing block in your bot flow.
Write a response: In the case of training small talks or you want to give users the response about some information that don't exist in the flow base on the message then stop. You can write manually response, it will make your bot more natural.
Click on write a response.
Write the answer you want your users to receive.
Create a new Smart Keyword
Smart Keyword get support from the third-party provider so that bot can understand much better to the same meaning inputs from users and provide the correct answers.
Set up Smart Keyword
To train your bot usng Smart Keyword, firstly, you have to enable chatbot integration with Natural Language Processing (NLP) services such as Wit.ai in Connect > Integrations.
Once your bot is connected with an NLP service, you can start training your bot at Build > Keyword.
Note: For more information about NLP services that we support, please read at Wit.ai
Set up Rule
Create Smart Keyword
After enabling Wit.ai support, write down the message you want. In Smart Keyword, we use the support of Wit.ai to understand the meaning of the message so we have only one rule is Message matches. Users can type the same meaning message and still can receive the correct answer.
Set up Response
The way you set up Response in Smart Keyword and Keyword is the same.
Prioritize Keyword or Keyword Set
Prioritizing has a greate effect on Keyword feature. By prioritizing Keywords or Keyword Sets in the list, bot can understand which Keyword or Keyword Sets you want bot to response first in the flow.
To have a clear thought about this, you can follow this example for more understanding. In this example, you set up two Keywords:
The first Keyword with the message "order" will receive an answer is a block "order?" in the flow which contains "what do you want to order?".
the second Keyword with the message "take-out" will receive an answer which contains "Thank you for using our service. you can go to the cashier to order and wait for your take-out".
If users accidentally type these two Keywords in one message, then users will receive the response from the first Keyword. In this example, the Keyword with the message "order" is above the Keyword with the message "take-out" and then message user will type is "I'd like to order take-out". Here is the result:
When you change the Keyword with message "take-out" from the bottom to the top like this:
Users will receive the response from the Keyword with message "take-out" after the change.
Prioritize Keyword Set
The logic in prioritizing between Keyword and Keyword Set is the same. If users accidentally type more than two Keywords in one message and those Keywords come from different Keyword Set, bot will match the Keyword in the first Keyword Set. You can always change the Keyword Set's order to satisfy your need anytime. Prioritize Keyword set
Note: Remember to Publish your chatbot again if you make any changes on your chatbot orelse nothing will be change in your facebook page or website.