The “myCMS and the Web of Data” – IKS Community Workshop have just ended and I’d like to thank IKS for inviting us as an attendee to the workshop. The two days have really given us an insight of what CMS vendors is working on and looking for regarding providing semantic capabilities to end users. I must say that a lot of focus where on products. I will here post a summarized version. Slides can be found here.
A lot of focus at the workshop have been on entity tagging and linking entities to each other using rdf and resource databases. When talking about semantic technologies in content management systems I feel there are good solutions that has been overlooked. If that’s because these solutions haven’t been explained properly or if the tools for using them haven’t been around for long enough I don’t know.
As the Internet gets flooded by all this information we need a way of filtering and relating documents and texts to each other. Saplo has something called related texts which is a semantic way of relate the contextual meaning of texts to each other. Using this feature it’s possible to create a graph structure between content without using any entity extraction or any resource databases at all. For example it’s a great way of structuring all pages that exist in a wiki. It can be an internal wiki where you automatically want to link to or between content regardless of what department you are working for or what space it’s published in. A person working within human resources can find related content from the finance department and reversed. Our current customers (primarily publishers) is using this to provide site visitors with links to related texts within their archive. In this way editors don’t have to search for those articles manually.
One of the best parts about this technology is that it’s totally language independent. All that is needed is a bunch of texts to learn from.
For those interested in related texts who is currently using Drupal we have built a module that automatically finds and presents related texts from your archive. It will soon be committed to the Drupal community, we just have some last tests left to run. If you are interested in the module just signup and we will notify you when ever it’s released.
Personalization is hot! How can we personalize the content? Just because we are friends and we share the social graph, that doesn’t mean that what is relevant to you necessarily have to be relevant to me.
We have created a way of using contextual recognition to be able to recommend or filter content based on a specific contextual profile.
It basically works like this. You create a group of texts which represents your profile and you can then compare texts to that group (or even compare a group to another group) to find out the relation between them.
Here I demonstrated Saplo Stream, a way of creating personalized filters and recommendations using Saplo Text Analysis API. I have made a screencast including sound that briefly explains how Saplo Stream works.
The reason I was demonstrating this is because I really think a CMS is also about showing the most relevant content to the user. Imagine a front page at the New York Times where the articles at the top is more relevant for you to read than the articles at the bottom. Or use this feature to actually relate people to each other based on their personal interest rather than what friends they have in common. It would be a great way of getting to know if there are anyone else working on the same problem within a large organization just by analyzing what they have posted on the company intranet.
I also wanted to give a heads-up regarding a service that we have been working on for some time and that we are planning to release for beta users this autumn (if everything goes as planned).
It’s a prediction engine where you can create your own prediction model. The model is basically several texts with an associated value e.g. number of re-tweets, credit score of a company or something else and based on the trained model it is possible to predict an outcome for another text. Balancing on the edge of overselling this, but wouldn’t it be cool to be able to predict the future?
Oh I almost forgot…we do have entity tagging as well. We currently support English and Swedish and return tags categorized into persons, locations or organizations and we also give tags a relevance value for the given text. Entity tagging will also be available in our soon-to-come text analysis module for Drupal.
How to get started
That’s it for me! Hope to get invited back ;)