12/23/2019
By
MJV Team

Essay on the Taxonomy of Chatbots

Chatbots are a strong trend in the market for corporations from all areas.

And the Taxonomy of Chatbots are consolidating because they do not necessarily need the IT professionals involved to become a reality. So MJV has brought a series of blog posts with what you need to know to develop a bot in your company.

But are you really prepared to enter the world of bots? Do you already know all the concepts used when it comes to conversational robots? Or do you get lost in the middle of so many terms and feel like giving the default response of the chatbots “unfortunately I’m still not ready to talk about it”?

If the last answer was positive, do not worry. Continuing our series, we’ll talk about the taxonomy of chatbots. We separated some more important expressions of the area so that you leave speaking chatbotnese fluently by the end of these lines. Read closely

First steps

Chatbot

They are conversational interfaces – systems programmed to interact and respond to messages from humans automatically. They can be integrated into various media and channels, activated by voice or text.

Entities

It is about what the intention refers to, that is, they are the important parameters for the actions that the user wants to execute. Entities complete the meaning of intentions. In the context of chatbots, entities are usually characterized by nouns, adjectives, services, and products that are part of the customer’s business.
Example: In the sentence “I want to download an ebook”, the entity would be the word “ebook”.

Intentions

It’s what the user means by the message he sent to the bot, the main idea of a phrase. It is through intentions that we can understand the actions that the client wants to perform. In the context of chatbots, intentions are characterized by verbs.

Example: In the sentence “I want to download an ebook”, the intention would be the word “download”.

Persona

It’s the personality of the bot. Here are several factors, such as what kind of language it will use; whether it will be more formal or more relaxed; make jokes, use memes and emojis; what subjects he will be able to answer; whether it will be a young bot; what is the gender, etc. The ideal is to marry the brand persona of the company with the persona of the chatbot.

Avatar

It’s the chatbot image. It doesn’t necessarily have to have an avatar. But if you choose to have it, it needs to be aligned with the persona of the bot, with the brand persona and the persona of the client. Depending on the profile, it is possible to think of a person, a robot or even an animal as an avatar.

It is important to remember that the chatbot avatar is not the same as the company mascot.

NLP

The acronym means “Neuro-linguistic programming” (NLP). This is a subarea of artificial intelligence that studies the ability and limitations of a machine to understand the language of man. The goal of NLP is to give the machine a basis so that it has the ability to understand and compose texts. That is what enables the interaction between humans and machines.

Platform

It is the tool for creating the interface of your chatbot. We currently have many options available in the market and most are free (charge only for an upgrade if you think there is a need) and very simple to use (even for laymen, if it is No Code – and we already talked about it here!).

Channel

Where the chatbot of the company will stay. You can set a channel or several, such as Facebook Messenger, WhatsApp, Telegram, company website etc. It is important that the choice of which channel the bot will act on is also in line with the persona of the bot, the brand persona and the customer persona.

Concepts related to Artificial Intelligence

Artificial Intelligence

It allows machines to learn through experience. This area of computing uses computational methods and devices to make the machine “learn” to use the rational capacity of the human being.

AI Model

Based on intentions and entities, an AI model is built, which is the representation of a reality. In the case of NLP, the AI Model is a conceptual design of what is believed that users will want to talk to the robot.

AI Engine

The AI Engine is the technical solution that uses the AI Model and Manually Assigned Examples to understand and classify new messages automatically. With this, the AI Engine is able to automatically classify new messages from the users and understand the intention to which it refers with a certain degree of confidence. Examples of AI Engines: Watson, from IBM; Luis, from Microsoft; Google Dialogflow; WIIT.ai, from Facebook.

Ranking

When the AI Engine receives a new message from a user, it compares with the AI Model it has (with all intentions and entities) and can recognize the intent to which that message refers, with a level of Confidence.

Trust

Probability given by the AI Engine so that a message is sorted according to a particular intent. The messages can have different percentages of confidence according to the manually sorted intent. The application that is using the AI Engine is what determines the correct answer.

Extraction of Entities

In addition to classifying intentions, the AI Engine also extracts the entities. That is, it identifies in the message that a given entity exists.

Machine Learning

It is a method of data analysis and standards that automates the construction of improved models.

General Definitions

API

The acronym comes from the English term “Application Programming Interface”. An API consists of a set of instructions, routines, and programming patterns used to access an Internet-based software application or platform.

Integrations

For the bot to be able to work beyond the decision trees, some integrations can be made to improve the user experience and / or service. At this point we decide if we will have integrations with some provider of NLP, database, CRM, OCR, etc. The integrations are diverse, and, through them, we will make the bot more attractive and intelligent.

Fall back response

It is the default response that a bot gives when it does not understand or is not yet ready to respond to a user interaction. It can also be configured for other situations, such as inappropriate comments or issues that the company does not want the chatbot to speak about or disclose.

Human transfer

Human transfer occurs when a chatbot forwards the call to a human, who is online in the chat or responds later, so that they resolve issues that the bot is not able to resolve. Practical examples would be to provide a budget, close a contract, hire an event or answer very specific questions, such as the status of a particular policy.

Fluency in the language of bots

In any industry – especially the more technical and IT-related – it is normal for them to have a logic and language of their own. And it is natural that, in a first contact, these terms seem confusing. But the more we immerse ourselves in the area, the more the communication becomes an organic process.
As bots are becoming increasingly popular and present in corporations, it is only natural that new functions and concepts emerge all the time. If you missed something we did not list here, leave your comment!

To better understand how a chatbot can help your business, discover our series of chatbot content. Start with the Chatbots: The definitive guide to implement them in your business ebook and start to dig deeper into the subject!

Back