Smart Discovery Tool
We developed our engine in response to a complex challenge in the world of content:
How best to match a film to the subject of a lecture or lesson?
The usual answers up until now have been to ask for recommendations from knowledgeable people, to scan our film review blog, refer to site support for help, ask colleagues, search for videos by keyword, and more.
Well, 'All of the Above are Correct', but this is a rather long and tedious process, which does not always produce the best recommendations, and worst of all, the feedback we receive from these sources is limited.
A popular method is to enter a limited number of movies (or books, or artists or professionals) into a naturally generated virtual selection. The advantage of such selection methods is that they reduce the search time.
On the other hand, they limit the scope of one’s research: Ask some film enthusiasts which are their favorite movies and which they rank highest. Probably, 90% will include Citizen Kane among their choices. Now, while we do not question in any way the genius of Orson Wales, this overwhelming response, recurring systematically for 40 years and more carries a strong odor conformism and fashion. To these influences, we should also add the concern of the speaker for how he or she will appear to her audience. It cannot be helped. This is human nature.
We Believe in Machines!
More specifically, we believe in smart machines – not potato peelers (even though their engineers may be smart).
So, we sat in a closed room with such a machine and taught it everything we know about classifications. Yes, classifications, not movies, because this is the core idea behind any recommendation. The machine's role in life is to know how to recommend on a video, based on the analysis of the input text. In other words, you, the user, paste into our engine some text. The text represents the topic you want to find a video for (e.g - an abstract of an article you want to present to an audience).
Our machine reads that text and tries to understand what it's about.
Then it goes to our pre-classified ever growing movie database to find a matching movie.
This process is known internationally as the algorithm of Content Discovery, and by association, we also gave the name Movie Discovery to our website.
The result is, in fact, a recommendation generated by a robot. It is also known as a relationship-dependent recommendation or a Contextual Recommendation, and the process is often referred to as Text Mining. The ability to understand the emotion expressed in the text and between the lines is called Sentiment Analysis, but this is less relevant for the understanding of academic texts, relating more to the analysis of opinions or surveys.
WHAT WE RECOMMEND ON
We recommend on videos from our own catalog and from other VOD providers, including those which are totally free, like YouTube. For YouTube, we had a huge challenge: how to filter out low quality videos ("quality" in terms of content, not resolution). It's was and still is, a tough task, as YouTube is full of non-academic, unverified content, what is also called Cats on Skateboards.
WHO NEEDS IT
Karl Mehta of TechCrunch, one of the the leading tech blogs, thinks it's crucial to have such a platform, in order to handle and manage the influx of new and existing knowledge. Actually, when we read it, we thought he was tapping our conversations (not), as it totally matches our rationale. We promise you we don't know the guy and never sent him any message or a PR communication.
You are invited to visit the machine in its air-conditioned lab. Challenge it with your texts and let us know what you think of the results.