NLP and Quick Replies
You can talk to chatbots in two basic ways: NLP (Natural Language Processing) or with static button replies.
With NLP processing the chatbot engine tries to process the text typed by the user, extracting and processing the words inside the same text to decode the supposed “user intent”. The intent, in the case of Tiledesk native bot, is the answer for what the user was searching for. NLP is a powerful technology, because provides the user with the freedom to write as he was chatting with a real human (while there is a machine on the other end). But sometimes this freedom is not the best solution.
Quick Replies (aka Buttons) are an alternative way to help end users to reply to a message. A button, in Tiledesk, simply sends the button-text to the chatbot on the other side of the conversation. So, if the button contains the text “Talk to agent”, after pressing the button this same text will be sent and shown in the conversation as if you typed it manually.
To better understand when and how buttons are a better solution than using NLP, we can start with a practical example.
For this example we can start from our Charlie chatbot in the first tutorial. Just consider, but we’ll return on this later, that the way described in this article to render buttons works on all Tiledesk chatbots, e.g. Native chatbots, External and Dialogflow chatbots.
Now suppose that the user wants to ask something about the type of accounts supported by Quantabank. He can, for example, ask something like this: “what type of accounts do you provide?”. Let’s create the answer to this question.
Fill the questions-answer form with the following text:
what types of accounts do you provide?
__Quantabank__ actually provides the following account types: - Basic Checking Accounts. - Savings Accounts. - Interest-Bearing Checking Accounts. - Money Market Accounts. - Brokerage Accounts. What of this accounts are you interested in? (i.e _Savings account_)
Now, to fulfill the user knowledge needs, we should provide to the user additional replies for each of the proposed options. We the purposes of this example can only provide an answer for the “Savings account” question. To accomplish this easy task we must create a new answer. Presse the + New answer button in the chatbot Charlie toolbar (as before). Fill the fields with the following data:
A __Savings account__ is an interest-bearing deposit account held at a Quantabank. Though these accounts you pay a modest interest rate while their safety and reliability make them a great option for parking cash you want available for short-term needs. [more info](https://quantabank/accounts/savingsaccount)
Click on CREATE ANSWER. Now we can test all the questions-answers workflow in the SIMULATE VISITOR view (press SIMULATE VISITOR in the Request panel). Click on the widget and press “New conversation”. After the chatbot greetings us with “Hello”, we can type our first question:
The chatbot will reply with a list of options in the message body with the final suggestion in the message tail to type the name of one of this options to get more information. We can type “Savings account”, getting the expected result:
Very good as result, the user got his response and he is probably satisfied with the reply. But he had to type by hand “savings account”. This is an error prone task, because typing often introduce errors and, moreover, typing is tedious for the user. Why writing something when there is a better alternative? Render the accounts as a sort of options menu: the list of available account types. It’s better in this case to give up on the power of NLP and switch to buttons instead! How to get buttons for the list of accounts?
Are you also looking to learn how to render buttons for quick replies?
With Tiledesk open source chatbot builder it’s very easy. Read our how to render buttons for quick replies quick guide and enjoy Tiledesk Quick replies!
Please feel free to send feedback about this tutorial to email@example.com. Thanks!