Variations on 3-word locations
Published in
2 min readDec 30, 2017
Recently Emily Fuhrman(@_E_F on Twitter) tweeted about a new project of hers that combines a few types of data I love into something beautiful.
In her project GPS Traces in Lexical Space she builds 3D elevation profiles from a process that involves what3words.
Her process(by her explanation) is something like this:
- GPX data converted to sequences of w3w trigrams
- sequential calculation of coordinate-wise Levenshtein distances
- the elevation plot aggregates the running Lev dist total in each dimension as elevation at that point.
The output is like this beautiful image:
Some Ideas:
- I think it would be natural and nice to make this into a web-app that takes GPX file and generates the output. Maybe even the string of word-triples would be interesting
- There is already elevation for each GPX data point, you could use that data for coloration. Further in this direction you can continue to push more GPX data into the visualization for different purposes.
- I like the notion of combining this idea with another technique I love shown by Nadieh Bremer to visualize the digits of pi; in particular, a tri-axis where each step is the triple of Lev Distances in that axis direction. There’s also just the 3D walk version of this where your axes are orthogonal.
- I also like using different distance metrics, Lev Distance is pretty natural, but some others—like cosign similarity based on some corpus—would be fun. It’d be curious to know what the total word set is like for w3w but I haven’t looked much.
- It’d be interesting to run the output word sets through anagram generators to convert paths to more curious words.
- Global heatmap of adjacent w3w tiles and Lev distance.
More off the wall
- What GPX routes generate a particular lexical elevation map?
- What GPX route of specified length generates the highest scrabble score when adding all the letters in it’s w3w representation? (Assume it must be all on land.)
- What do letter/word shuffles of the w3w representations look like on the globe?
This feels like a really rich idea full of fun diversions! Thanks so much to Emily for discussion on Twitter and the original work.
