Journalist-bots, data and connected everything

Robot Journalist

As a part of exploring how our research-platform for interconnected objects (see "the Social Web of Things") could enable new user experience concepts, we looked into what we could do with the log of data generated by the connected gadgets, sensors and services. This data is both used to maintain an updated snapshot of the current state of everything, and also for detecting historical patterns, which can be used for automating the set up of rules that decide how the devices interact with each other.

We had these two different left-over ideas lying around from earlier work, which we now realised that could fit together in an interesting way: First we had been interested in how the data-log contained the 'keyframes' to a chronicle describing the everyday life in the community of our connected things. There are narratives hiding within that data, about network-ness and the invisible interrelationships between everything connected. Second, when we investigated natural language generating technologies for the conversational interface of the Social Web of Things concept, we assessed a number of companies specialising in technology that could generate readable stories from basic fact-box data. These services are using machine learning and data mining algorithms to compose written articles, and a number of news websites (for example Forbes) already use "robot journalists" to generate articles on certain kinds of topics.

The same way the Social Web of Things concept makes the networks of relations and interactions between the connected things more legible to people, we could do something similar by feeding selected datasets from the data-log into a journalist-robot to generate a readable narrative about what our things had been doing during a given period of time. In stead of rendering the activities in real time as a social network newsfeed, we could use the same platform to make something like a newspaper.

We made a quick first prototype using flipboard as a viewer, to test the experience. Format-wise we chose the short fact based "breaking news" format since news reporting is very conform and follows clear patterns within each story-category. The data describe the technical context (sensor readings, system status, activities) but also potential actions and relevant functions or services that could be suggested to give desired outcomes in certain situations. The narratives was based on patterns in news stories found on BBC News. Images were added to each article by retrieving creative commons licensed images from flickr's API using relevant keywords from each article. The articles was published in an RSS feed and we created a number of twitter accounts which we scripted to listen for different article categories in the RSS feed and then tweet only the links to new articles. Finally one twitter account which was following all the others was set up in Flipboard (we did this twitter-detour since the flipboard-app didn't treat RSS feeds the way we wanted it to). The result: an illustrated magazine with stories from our connected things.

Read news in tab

It was amusing to read the stories for a while, but to be frank, we realise that most people are probably not geeky enough to sustain a reading habit based on computer-generated stories about their gadgets. So we did another take on the same concept. A TV news program.

In addition to the setup described above, did the following: send the text files to a web-based avatar service that supports text-to-speech and lip-sync. We used SitePal which has ok APIs and provides links to separate flash-files for each rendered scene. Then the flash scenes was embedded into a TV news program production-template we created in flash, and voilà! A virtual TV News Program, generated and compiled automatically on-demand for a target audience of one:

We think it is important to explore ideas by making them into something concrete that can be experienced and discussed. It might be something slightly absurd about news stories from the daily lives of inanimate objects, written by algorithmic robot-journalist being read by a virtual news anchor, but the point is that this is possible with available technology, today. However, we see the technology as less interesting. We are rather interested in generating questions and finding different ways of thinking about user interaction in an Internet of Things. We use work like this to reveal, explain and discuss the invisible relations and hidden possibilities in an all connected world.

Or perhaps it could be a commercial service after all? The Self Aware ROOBMA has over 16K followers as I write this, so who knows. People find themselves entertained by so many things.

The Ericsson Blog

Like what you’re reading? Please sign up for email updates on your favorite topics.

Subscribe now

At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.