Dear Santa,
I know that you have a lot of things to do right now and you do not have the time to read yet another letter besides all these billions of letters from kids all other the world: This kid wants a blue bike with red stripes, that kid wants a red bike with blue stripes and do not forget little Timmy’s one of a kind model train. It must be incredible complex to sort all these information and align them in a logical and manageable way. I feel for you. So hear me out, maybe I can help you with your problem and ease the information overflow, you are experiencing right know.
All these letters have something in common: they are written in plain text and, despite a kid’s basic knowledge of grammar, lack common structure. You have to read the whole letter to find out what the kid wants for Christmas. Would technology help, I hear you say? Well, machines have problem understanding data which has no clear structure. So, just buying a new computer with some software and hope for the best will unfortunately not work for you. And just hiring more elves to read all these letters for you will make your costs explode.
But do not worry, help is on the way. Ontology Learning might be a solution for your problem and could help you to extract and sort only the essential information from the letters. The concept creates a so called “taxonomy” which means it creates groups and hierarchical structures out of unstructured information sources when such groups share specific characteristics and dependencies. SAP is already using Ontologies as a tool within its Service Oriented Architecture to bring structure to a complex domain. Moreover, Ontology Learning also recognizes relations between different terms by using statistical measures. Bikes can be a sub-group of sports article and the color would be another sub-group below bike. And all this happens (semi-) automatically!
Still not convinced? Just think about how much time you could save if the content of all these letters would be pre-organized and could be assigned to the right elves in the right departments. You may ask why you cannot just use a simple search engine to look for keywords. The reason why Ontology Learning is more suitable to fulfill the task is the potential ambiguity of words in language. Ontology learning does not simply look for the term, moreover it sets the term into the used and correct context. This makes sure that the kid gets a baseball bat and not a hairy, winged animal for Christmas if it just states “bat” in the letter. This avoids “surprises” on Chistmas Eve.
So long and all the best,
Jens
