This is what I learned after teaching chatbots to journalists: 3 takeaways for newsrooms

In a guest post for OJB Maria Crosas points out three main takeaways that newsrooms should consider when aiming for a complete chatbot experience. 

Over the past year I’ve been frequently invited to share ideas around how bots can help newsrooms to deliver news, and advice on how to build an engaging chatbot experiences. And throughout these classes, I’ve also had challenging questions on how these technologies are pushing the boundaries of ethics, artificial intelligence and storytelling.

I’ve boiled down these experiences into 3 takeaways for newsrooms that want to begin the chatbot journey. Here they are…

1. Teams need to integrate different skillsets

Screen Shot 2016-12-07 at 20.06.44

Source: Bootcamp 2016

Artificial Intelligence engineering teams are no longer composed only of developers or data scientists, but also writers, designers and sociologists that put the user in the centre.

As an example, Google is hiring storytellers to work on Google Home’s personality.

The same applies to the media industry. Relying only on developers to create the commands and journalists to push the content will not be enough to make a new channel successful and engaging.

What would newsrooms need if they want to build a text or voice bot for their audience? Below are some key roles to consider:

  • Community manager to provide insights on the demand and feedback from the audience
  • Writer to deliver the content
  • Designer to create a unique user experience
  • Data analyst to analyse the bot once it is live, and identify how to improve it
  • Developer to not just rely on third parties for the Natural Language Processing (NLP) component

2. Manual training can help journalists to find stories

A chatbot project doesn’t end when it’s launched to the public. Users are going to interact with it, ask about everything they come across (random and specific questions), and provide (directly or indirectly) feedback.

Most of the Natural Language Processing (NLP) platforms out there have a training tab that shows the conversations that the bot is having with the users.

It’s easy to see if specific sentences have matched existing intents (something that was already within the flow) or existing content. Below, for example, is an image of what this training tab looks like in Dialogflow:

Screen Shot 2018-05-19 at 15.41.05

This manual training can help journalists in two ways:

  1. Make the bot more intelligent by increasing the number of sentences for each intent. Based on the conversations that the bot has with the users, these platforms record the matched and unmatched intents in each dialogue. This will help the bot to be ‘smarter’ and avoid a default ‘fallback’ answer.
  2. Discover stories or leads from users’ conversations when a large amount of people are asking about topics that haven’t been tackled yet, or when citizens delivering specific information for follow ups.

3. Identifying the right channel is as important as the content

Delivering good content through the wrong channel is useless and might lead to failure. So it’s important to identify where your audience is before elaborating on the content.

Currently WhatsApp, Facebook Messenger, Viber, Line and WeChat are the most popular messaging apps. Even though more than 100,000 millions bots have been approved by Facebook by September 2017, it might be that your audience is not there yet.

An interesting approach is that of the BBC News Labs and BBC Visual Journalism teams, who have developed a custom bot-builder application that makes easier for reporters to build chatbots and insert them into their stories.

This will help users to grasp the basics of a complex story:

Screen Shot 2018-05-19 at 16.00.01

In this ocean of artificial intelligence, chatbots and voice assistants, it can be helpful to go back to basics. The definition of chatbots by Matt Schlicht, CEO of Octane AI and Chatbots Magazine​, is the one that I find the most clear:

A chatbot is a service, powered by rules and sometimes artificial intelligence, that you interact with via a chat interface. The service could be any number of things, ranging from functional to fun, and it could live in any major chat product (Facebook Messenger, Slack, Telegram, Text Messages, etc.).”

Do you have any tips about building chatbots for news? Please share them in the comments or via Twitter @mcrosasb

An earlier version of this article was first posted on Maria’s blog Dinfografia. Maria is a graduate of my MA in Online Journalism at Birmingham City University, now the MA in Data Journalism and the MA in Multiplatform and Mobile Journalism.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.