When creating a chatbot, different teams of a company must learn to work together, to coordinate but above all to communicate. The key to good communication and good understanding between those teams is to use a specific vocabulary, which is adapted to chatbots and understandable by all. For people outside the areas of expertise required to create a chatbot, the terms used are not necessarily clear. Indeed, talking about “knowledge bases” or “free input” to an ordinary user can sometimes be akin to an extraterrestrial language. Let’s decipher together some terms specific to the chatbots universe!



Term resulting from the combination of the English terms “chat” and “robot”, “chatbot” refers to conversational agents. These conversational agents, also known as “conversational robots” and sometimes even “virtual assistants”, are usually available online in the form of instant messaging. Instead of conversing with a human being, users converse with robots whose answers have been prepared by human beings and grouped in knowledge bases.


A “Chatbot-as-a-service” is a chatbot offered as a service such as Hubi. The chatbot comes with an online platform that allows users to build and customize the chatbot that best suits their needs. The biggest advantage of chatbot-as-a-service is that it does not require any programming knowledge: there is no need to code because the platform allows to manage, deploy and customize the chatbot on a simple and intuitive interface.


The engine of a chatbot is what allows the chatbot to respond to users. This “engine” is actually an algorithm, i.e. a computer program consisting of sets of mathematical and linguistic rules. Most current chatbots combine “machine learning” or “deep learning” bricks with Natural Language Processing (NLP) and/or Natural Language Understanding (NLU) bricks to process messages entered by users. Mathematical rules, on the other hand, make it possible to process user responses by following a specific reasoning that could be the one of a scenario. Chatbots engines are therefore all different since each association of various computer tools and linguistic rules, NLP and/or NLU is unique.
To find out more about NLP and NLU bricks, read How to choose between NLP and NLU for a chatbot?


Scenarios are a hierarchical organization whose graphical representation is reminiscent of a tree and its branches. The succession of the different branches of a scenario is assimilated to the various paths a conversation can take. Scenarios cover all possible turns of a conversation where answers are only “quick replies”. Scenarios are a series of questions and quick replies that allow a process to be carried out automatically. If you are familiar with the concept of “workflow”, the operation of scenarios is similar to that of a workflow.

Machine learning & deep learning

Machine learning or “automatic learning” in French is a technique specific to artificial intelligence that allows machines to learn through mathematical and statistical methods on the basis of rules predetermined by human beings. To do this, the machines need to process huge amounts of data.

Deep learning is a sub-branch of automatic learning, which allows machines to learn by themselves without rules predetermined by humans. Like machine learning, deep learning requires huge amounts of data on which the machine will base all its learning.

Knowledge base

A knowledge base or database is like a large library that gathers all the possible answers of a chatbot to a multitude of questions. All questions pointing to the same answer are grouped together and called “alternative questions”, and are associated with a single answer. Usually, a knowledge base is similar to a FAQ and deals with a particular field or topic. The questions and answers contained in a knowledge base have been prepared by human beings.

Open and closed questions

Open questions are questions that users can answer freely. Unlike closed questions, it is not possible to answer an open-ended question with “yes” or “no”, “true” or “false”. Closed questions also correspond to questions with “quick answers”, i.e. button answers.

Quick replies

The “quick replies” or “quick answers” result in answers proposed by the chatbot in the form of buttons. The user has no choice but to choose among the answers proposed by the bot and click on one of the buttons if he wants to continue the conversation.


Adaptive cards or “richcards” are another possible response format in a chatbot. Richcards are a formatting of the text in the form of a card that can be accompanied by media such as images, gifs or videos, but also links, etc.


An utterance is a message freely written by a user to answer or question a chatbot. It is a free input which, contrary to the “quick replies”, uses the NLP and/or NLU bricks of the chatbot engine to process the information in natural language contained in the user’s message. The statements contain strong elements called “intent” and “entities”.

Intent and entities

Intent and entities are the strong elements contained in a user’s utterance and use NLP and NLU techniques to be processed and understood by the machine. Intent corresponds to the overall meaning of the statement, such as the action the user wishes to perform. Entities are the other strong elements contained in the more specific statement. To learn more, please read the following articles: The Named Entity Challenge and How to choose between NLP and NLU for a chatbot?


A module is a container to which one or more knowledge bases and/or scenarios are associated. Modules allow the chatbot to access the information contained in the knowledge bases and/or scenarios connected to it. The modules thus allow to deploy the knowledge of a bot on different channels.


A channel can refer to all the applications and sites on which a user can deploy or connect the chatbot. It can be a SharePoint site or a Microsoft Teams chat channel for example.

What about Hubi?

Hubi is a chatbot-as-a-service that combines machine learning with NLP and NLU bricks. Our chatbot is currently available on SharePoint and Microsoft Teams channels. Thanks to continuous improvement, Hubi.ai teams are perfecting Hubi’s automation scenarios as well as its numerous knowledge bases to offer you the best chatbot experience.

Camille is a computational linguist by training. Following two experiences in Parisian start-ups on named entity recognition and callbots, she recently joined the Hubi.ai team at Hub Collab as a chatbot scriptwriter.