Introducing FollowThis

Quick Pitch

Challenge: A user would like to follow a specific story for updates and related items.

Typical Use Case

1. A user arrives at a story that they are interested in. They’d like to keep up with it, or at least some aspects of it.

2. The user clicks the FollowThis button (or bookmarklet). An overlay appears which will offer some site-specific options, like whether they’d like to follow on this site only or across the web, and the frequency of the notifications.

3. The user provides their email (if not already logged in) and submits.

4. The user will receive a verification email to ensure we aren’t spamming random email addresses.

5. From then on, the user will receive email notifications if that specific story is changed and updates containing links to related articles and videos. They will be able to modify or unsubscribe to the updates at any time.


Click to visit the IPTC site and view the rNews Domain Model

The system will store metadata about articles in a specific ontology in an RDF database. Most likely this will be rNews, since it shows great promise in getting adopted and will lend itself well to inter-operation between this and other systems. The system will make SPARQL queries available through a REST API that will return new and related items based on the user’s subscription metadata.

When a user chooses to follow the story, the subscription is stored along with entities that comprise the topics that story covered. These are more specific than the level we would get with tags. For example, the subscription might follow stories that deal with Java the island, but not the programming language, or coffee.

We’ll need feedback from the users as to whether they want to dive down into these sub-topics at the point of follow or whether it would better be left to refine at a later point or some amount of both options. Perhaps and “advanced follow” link. Feedback from users will be key toward polishing the interface.

We’d also like to work collaboratively with the Editors and Producers of the sites. A good amount of metadata can be set up upon initial launch but natural language processing isn’t always as good as humans at some more complex entity relationships. We’d like to have editors aid in the production of this metadata, as well as users, without the process becoming a burden.

Any news organization using the software will be highly encouraged to start adding Semantics like hNews or rNews to their presentation layer. While we will make the software work without it, it will be much more effective if we start with a solid base of metadata.

Since hNews currently has a much higher adoption, an hNews to rNews converter will be one of the first components needed. We will release this to the community as a separate standalone library since it could be helpful for other applications.

Many organizations already have the required metadata within their existing editorial backends, it’s just that they aren’t presenting it to the browser. Implementing one of these specs is not more than a few day long project.

Secondly, and also optional, we’d like to encourage collaboration across news organizations. In other words, a user choosing to follow a story would also be submitting that story to a commons of semantically categorized news articles that other sites could present as notifications to their users. Sites could provide RDF dumps to each other to create a distributed wire system, in a similar fashion to the way that Usenet newsgroups work.

An added benefit to this collaboration would be driving inbound traffic from other collaborating organizations and also offering the end users the ability to choose to follow the topics from one specific site or from the web at large.


Most traditional news organizations don’t usually think in terms of collaborating with their competition the way the tech industry does. This is why we make some of the features are optional with the hope of later showing how value can be gained from unorthodox strategies like sending users and content to competitors.

Another issue would be the inability of the organization to provide the metadata needed for these relationships. Natural language processing can be used to extract entities, like Open Calais does. This would be a challenge to build ourselves and it isn’t clear that there is already an open source alternative. More research into some of these related open source projects will be necessary. NLTK to RDF seems to have potential.

Why should a business adopt FollowThis?

The most precious resource a news organization has is an interested reader. Keeping that user engaged should be the primary goal. FollowThis allows your users to stay on top of the stories they are most interested in, by notifying them of updates or related items. By keeping these users engaged, the user benefits and the organization gets its content to the right audience, and drives more traffic. Aggregators like Google News have begun to personalize their offerings. News organizations must do the same and do so while they have their users at the “point of sale.” The metadata that powers the service is already available in the CMS of most organizations, but is being under-utilized. A by-product of this project for any news organization would be a database that could easily be used  for other areas, like ad targeting. Implementing FollowThis will make for happier users and a healthier business.

About Me

My name is Matt Terenzio and I’ve been building websites for news organizations for almost ten years. I’m interested in pursuing how we can use some of the existing and emerging metadata stored in these organizations to help the organizations themselves and help their users get a better news experience. Contact me on Twitter @mterenzio or mterenzio at gmail Keep an eye on FollowThis