Conversational Interfaces

Defining conversational interfaces

Defining conversational interfaces is a great point to start from if you are considering starting a project or even a career in conversational interfaces. A lot has been spoken about conversational interfaces. As we already saw in this blog, conversational interfaces are not only super popular and present in our culture through things like movies that feature talking machines, they are also not a new thing and they have had a long history that started in the Middle Ages. Conversational interfaces are being pushed by all large companies and implemented in different use cases with more or less success. In this blog post I want to go back to the very first question we need to answer: what, exactly, are conversational interfaces?

What is an interface?

Before we define conversational interfaces lets look at interfaces in a more general sense. An interface is defined as a shared boundary used by elements of a system to exchange information. Interfaces consist on hardware or software that humans interact with in different ways:

In computing we define interface as the features of a computer system which allows the user to interact with it, and it is composed of both hardware and software features. There are five main categories of user interfaces:

  • command line (cli)
  • graphical user interface (GUI)
  • menu driven (mdi)
  • form based (fbi)
  • natural language (nli)

What are conversational user interfaces?

Conversational interfaces are interfaces that use conversational means or elements to let humans and computers interact. Conversational interfaces let humans and machines interact using natural language, sentences in languages like English, Korean or Polish with meaning that can be matched with software and hardware functionality.

In a world of Digital Transformation where machines are working together with humans or replacing them at repetitive tasks, facilitating the interactions between these two groups in the workforce – the ones made of flesh and bones and the ones made of hardware and software – is key.

Conversational user interfaces may seem very modern but they have actually been part of our imaginary since the Middle Ages (check out details in this blog post about the history of conversational interfaces) and in movies since the early days of cinema (more info in this blog post about movies that changed the way we see AI).

What types of conversational user interfaces are there?

Defining conversational interfaces typically leads to talking about the different types of conversational interfaces we can interact with so far.

Conversational form filling

I consider user experiences where form labels are written as conversational questions the most basic implementation of a conversational experience. When we write “What is your name?” or “What shall we call you?” instead of “First Name” we are facilitating our users a more conversational interaction. Facebook and twitter used conversational prompts brilliantly as they added messages like “What are you thinking?” as placeholders in our empty posts.

Button-based chatbots

These bots typically work inside of mainstream messaging apps like Messenger or Telegram and they use buttons to guide the user through a tree structure. These chatbots are cheaper to build and good for early phases of a project. They are also the best option for certain users or workflows. Do not underestimate the power of buttons! It may be as simple as using a third party service like Fobi.io to transform your Google Form in a conversational experience.

Interactive voice response (IVR)

This technology has been around for ages and you have probably interacted with it at some point in the past. These call center robots can understand keywords and are used to triage phone calls before they are dispatched to the corresponding department. Most IVRs do not automate resolution.

Keyword based chatbots

Some button based chatbots imitate IVR and can recognize certain keywords. They are a first step towards full NLP and usually suffice for the most frequent natural language queries related to the business during the first period of operation.

NLP chatbots

These chatbots can recognise complex natural language requests using AI technology. Humans still need to program the different intents or features of the chatbot including a few of the different sentences or expressions known to be frequent for that request as well as the answer or answers the chatbot must give. NLP chatbots can’t react to intentions that were not defined by humans (no, they don’t learn new features on their own, sorry) so they have a default answer they give as a fallback.

Voice apps and skills

These are domain-specific conversations with NLP that are integrated with voice-to-text and text-to-voice technologies. In this way, using a microphone, we can capture the expressions from the users. We can then transform them to text to process them with NLP, and convert our NLP response to voice using synthetic voices. This audio is played back to the users through the speakers, all in a matter of milliseconds. These domain-specific apps are usually integrated into larger platforms like Amazon Alexa or Google Assistant.

Voice-based conversational assistants

General purpose assistants that have an understanding about a broad amount of topics and that can integrate with different data sources and systems. Alexa, Google Assistant or Siri are conversational assistants but there are others like Mozilla Voice. Alexa and Google Assistant are also platforms for voice apps and skills.

Did you enjoy reading this post defining conversational interfaces? Continue reading about conversational interfaces integrated in every day objects or in movies that have changed our expectations about them.

One comment

Leave a Reply

%d bloggers like this: