Guess the Poster CCC-WordCloud Competition Part 5

[chrisw]

Guess the Poster
Hopefully, this is now the final version of everything (famous last words).
Previously guessed posters now have new WordClouds and will need to be re-guessed, ordering is randomised.

28 posters included, one of them might be you !!
(plus another 20 for CTF)

CCC Postings WordClouds
1-4
https://funkyimg.com/i/2XG5x.png
5-8
https://funkyimg.com/i/2XG5A.png
9-12
https://funkyimg.com/i/2XG5E.png
13-16
https://funkyimg.com/i/2XG5L.png
17-20
https://funkyimg.com/i/2XG5N.png
21-24
https://funkyimg.com/i/2XG5P.png
25-28
https://funkyimg.com/i/2XG5Q.png


CTF Postings WordClouds
1-4
https://funkyimg.com/i/2XG5c.png
5-8
https://funkyimg.com/i/2XG5a.png
9-12
https://funkyimg.com/i/2XG59.png
13-16
https://funkyimg.com/i/2XG5f.png
17-20
https://funkyimg.com/i/2XG5d.png

Good luck!!

[Laskos]

I am (7) on CTF, right? What's the matter with "number" being such a common word for everybody on CTF?

[Laskos]

I am (7) on CTF, right? What's the matter with "number" being such a common word for everybody?

And (14) om CCC?

[chrisw]

I am (7) on CTF, right? What's the matter with "number" being such a common word for everybody on CTF?

correct

[chrisw]

I am (7) on CTF, right? What's the matter with "number" being such a common word for everybody?

And (14) om CCC?

correct.

‘number’ was substituted during preprocessing phase for ints, floats, things uninterpretable but with a lot of numbers in, which also took out specifically all chess moves and chess notations. NLP sentiment analysis reasons. Also misspellings and I also found a list of deliberate misspellings of so called ‘bad’ words to use which get converted to the real thing. Ultimately words get vectorised and used as NN inputs, and NN can make more sense of ‘meaning’ if word count can be reduced, mistakes corrected and so on, there are large preprocessed word vector dictionaries, and one preferably wants words that match.

maybe number should be in the stop words list for these wordclouds, but now you know, it could be a clue, I suppose. A large font “number” means poster posts a lot of numbers.

Come to think of it, CTF postings got the preprocessing treatment, whereas CCC postings I was still working on, so the CCC wordclouds may well be less preprocessed that the CTF ones. This was done all well over two months ago when it got sidelined to help Ed with the Similarity Testing, so one tends to forget what one has done without getting a memory jog.

[chrisw]

The CCC list is maybe a bit difficult because there are so many WordClouds. To easy it up a bit, here's the list of the 28 possibilities by talkchess username, all shuffled, so in random order. The puzzle is to link each username to its worldcloud number (1-28)

Gerd Isenberg
Tord Romstad
Laskos
chrisw
Guenther
Terry McCracken
Rebel
Vasik Rajlich
hgm
Dann Corbit
Daniel Shawul
MikeB
fern
mclane
sje
Adam Hair
bob
mcostalba
michiguel
syzygy
Graham Banks
Gian-Carlo Pascutto
mjlef
Don
crem
Rolf
Christopher Conkie
zullil

[Sam Hull]

Interesting to compare the first and second versions of what I'm pretty sure are my CTF contributions:



-Sam-

[Ovyron]

These new clouds look a lot more generic than the old ones. Like, I was able to catch Graham Banks very easily on those, and in these new ones I can't find him.

But it seems add_rate was changed to addrate, so 16 must be chrisw!

[chrisw]

Interesting to compare the first and second versions of what I'm pretty sure are my CTF contributions:



-Sam-

CCC 7 = Laskos

CTF 4 = Sam Hull
CTF 8 = Henk
CTF 14 = Laskos


Correct about 4. in the second version.
You might be 10 in the first, the indexing is now sadly lost.

Version 2, I did kind of finally read the documentation for WordCloud, and instead of giving it an ordered list of the 500 most frequent words and set its word limit to 150, for version 2 it got the entire word list, as a dictionary with frequency information (and a word limit of 150).

Version 1 WordCloud was effectively expecting "free text", ie just a normally constructed piece of written text, so version 1 is what it did with its algorithm on an ordered word list. It's a miracle it worked at all, really.
Version 2, after reading some docs, always to be avoided haha, knows when it gets given a dictionary structure, what to do with that, so I think version 2 is more in line with WordCloud's intentions, so to speak.

[chrisw]

These new clouds look a lot more generic than the old ones. Like, I was able to catch Graham Banks very easily on those, and in these new ones I can't find him.

But it seems add_rate was changed to addrate, so 16 must be chrisw!

Yes, 16 is me. Via addrate?! Crikey.

CCC 7 = Laskos
CCC 16 = chrisw

CTF 4 = Sam Hull
CTF 8 = Henk
CTF 14 = Laskos