October 24th, 2016
We recently made a structural change to the way that we store information about which webpages Ultralinks are found on. Previously, we only stored a record that would indicate that an Ultralink had been found in a specific text fragment. Now we not only store that association, but we also store the exact string or link that caused the match.
This new change allows us to do some really neat stuff like see exactly when a new celebrity nickname started taking off or examine the most prevalant ways a certain place or thing is referred to. The granularity with which we can perform really interesting analysis has sharply increased, now that we can track things on a per word level.
For example, check out the current Mainline Database match breakdown (as of Oct 24, 2016) for the two major US presidential candidates:Donald Trump
This is a graph of the top 16 matching references by occurrence count. You can see that the Republican candidate is most often referred to by his last name "Trump" and his first and last name "Donald Trump". While he has a number of other monikers, those two are the only ones of general significance.Hillary Clinton
The Democratic candidate is less often referred to by only her last name "Clinton" (which makes intuitive sense, but the data bears this out) and is most often referenced by both her first and last name "Hillary Clinton". She is also significantly referred to simply by her first name "Hillary".
Explore the Mainline!
Every Ultralink has it's own webpage where you can see these interesting visualizations. The graphs above only examine the number of occurrences of each match, but on the actual pages you can break the visualization down by unique webpages and unique websites too. Feel free to explore the Mainline database and see what other interesting tidbits you can discover!
We're working on a lot more interesting data analysis and visualization projects that we can hopefully share with you soon.