With the latest update of the Serendipity Machine (www.serendipitymachine.com) we did something awesome! We are now using an Artificial Intelligence algorithm to find connections within tags and people. So check in and find out with whom we are matching you! Why we did this?
Within The Serendipity Machine, our aim is to connect people so that they can develop themselves to their fullest potential. We think that the best way to connect these people is based on their knowledge, or expertise. On The Serendipity Machine, knowledge is represented in so called ‘knowledge tags’ and these tags are used in the first place to visualize the knowledge available per person. Using these tags people can find other people of interest to either chat or meet-up with them in order to create a useful value network.
To assist the user in their quest for creating such a network, the system uses an Artificial Intelligence algorithm to find connections within tags and people. The question in this case is, what kind of people are most useful to connect with? Are these people with exactly the same knowledge and skills, so that you can discuss these? Or are these people with slightly different knowledge, so that you can expand your knowledge, or receive help in problems you cannot conquer.
The Serendipity Machine provides the user with, in the first place, matches within the first degree of knowledge. This is a direct match between the knowledge of two users. Let’s say for example that user A has the knowledge tag “web design” and user B also has the knowledge tag “web design”. They are now a direct match on this knowledge tag.
But the problem with the direct match is that it’s really limited to only the people who have exactly the same knowledge tag(s). People describe their area of knowledge mostly slightly different from each other, which results in just a few direct matches. We found the solution for this with the collaborative filtering algorithm.
We use this collaborative filtering algorithm to find tags that are similar to the tags entered by the user. Based on these similar tags, other meet-up suggestions can be found because the scope in which to search for users is extended. The similar tags which are found by this collaborative filtering algorithm are tags that are not primarily entered by the user, but are close to his/her tag cloud. This gives us the opportunity to suggest people that have slightly different knowledge (similar tags), but are of interest for the user.
Let’s say for example that user A has knowledge about “photography”. We can now look to the tags which are similar to photography and match the user also on these tags. For example, a lot of photographers have filled in “Photoshop” next to the knowledge tag “photography”. We recognise these patterns because it would be very likely that user A has also knowledge about Photoshop because most photographers edit their photos in Photoshop. So now user A is matches with user B, while they doesn’t have the same tags in their tagcloud.
This algorithm is a self-learning, unsupervised algorithm, as it learns from previous connections made in the data, and creates its own rules and connections. When inputted more data, some connections between related tags will grow stronger, some will grow weaker.
The better you describe your knowledge in clear tags, the better suggestions we can give you and the more relevant your matches will be. These knowledge tags are parts of your knowledge represented in at best 1 or 2 keywords. Clear tags are easy to use, easy to understand and easy to interpret by our system!