Some fun automated versions of classic wikipedia games
I'm using Python 3.6. Dependencies lxml,nltk - but I think these come with Anaconda anyway
If you use Python 2.X you'll have to change to using urllib2, but the rest may just work.
Sarab S Sethi (http://www.imperial.ac.uk/people/s.sethi16)
start_end_game.py
My go at automating the Wikipedia game (https://en.wikipedia.org/wiki/Wikipedia:Wiki_Game)
Given a starting wikipedia article, just by following links try to get to an unrelated end article
We use a wordnet embedding to derive a similarity score between the title of the end page and the title of each link in a Wikipedia page, then jumps to the best page.
Turns out the most fiddly bit was dealing with cycles and pages with no links or in non-standard formats Otherwise surprisingly simple and seems to work well!
I think one of the (many!) major limitations is matching is done on single words there's no concept of phrases or titles made up of multiple words.
RANDOM_MODE generates random start and end points. This is pretty tough and most
of the time the game ends on failure. Set the links manually to give the game a
better chance of finishing. Recommended value: False
STRICT_MODE defines whether you have to land on exactly the correct end page or
whether something "close enough" is fine. Recommended value: True
Let me know how you get on - the algorithm was able to get from "Financial engineering" to "Reality" in an impressive 7 steps
philosophy_game.py
Apparently if you click the first link in any Wikipedia article and do the same on each subsequent page you end up on the page for philosophy 95% of the time (https://en.wikipedia.org/wiki/Wikipedia:Getting_to_Philosophy)
Big claim, easy to test
This may have previously been true, but now there are lots more cycles, notably from the article for Science which comes up quite often. If we purposely jump out of loops this works quite nicely