Wir sind die Roboter – Why and how to build a chatbot

The air is a breeze with mega-trends and next-big-things about the upcoming year 2017. If one is to believe the shamans of digital evolution, one of those next big things are chatbots. Reason enough to do a bit of digging and answer the 3 fundamental questions in regards to chatbots: Why to build a chatbot? How to build a chatbot? And what to beware of?

Why to build a chatbot?

Users are deserting social networks and moving on to ever more powerful messaging apps. Therefore businesses face an interesting challenge: How to keep in touch with clients that are retreating into the privacy of personal chats, group conversations and hang outs? Chatbots, programs specializing into conversational interaction with humans.

Through their ability to interact one on one, have a (more or less) natural conversation with a human and to learn from these episodes on an individual and a collective level, chatbots are believed to be able to change fundamentally, how we search (for a cosy restaurant nearby), how we order stuff (pizza, clothes, transportation) and even how (and by whom) we receive advice in legal, medicial or professional fields.

For a more comprehensive write up of the business side of chatbots read Gareth James – Complete Guide To Chatbots. If you would like to get an idea on how the journey is going in the long rung, check Venture Beats Bot Channel.

How to build a chatbot?

Before we cover the how, let’s take a look at why again. As building a chatbot involves mastering aspects of three big technologically challenging fields – Artificial Intelligence, Natural Language Processing and (Big) Data – it pays to think which of these challenges we really need to tackle. Chatbot Magazines So you want to build a chatbot provides a great sparring in this regards.

If after all you still need to build a chat bot and if you are aiming to build the bot (your understanding) from scratch, these simple, hands-on tutorials on building a chatbot in Python, JavaScript or PHP will serve as a great starting point. Having gotten a general feel for the subject, these deep dives into NLP for chatbots, into chatbot architecture and into deep learning might prove well worth your time.

If at any point during your chatbot journey, you long for some tooling, Chatbots Magazine (again) got a neat overview of some of the most popular tools and their use cases.

What should you be aware of?

Let’s close this intro with a three challenges you might not have thought about (enough):

  1. A chatbot needs to scale. Massively, most platforms are (not yet) built for this.
  2. A chatbot needs to chat. No technology imaginable will be able to make up for a poor or lack of use cases and intents. Beware from reducing chatbots to a technology problem.
  3. A chatbot will not work (sometimes). This  makes collecting analytics and feedback even more valuable. Again, there is (not yet) a standard solution for this.

Head over to infermedica blog for a more in depth glance on these problems and their solution by the infermedica team.

Image: Emoji art supplied by EmojiOne