a few assorted data bugs

[jmhessel]

Hi there!

Thanks for sharing this data, and for your work at nextml!

I wanted to mention a few discrepancies I found with this data, in case they are helpful to know about:

• I think contest 644 has the wrong image associated with it because https://github.com/nextml/caption-contest-data/blob/master/contests/info/644/644_captions.csv do not match https://github.com/nextml/caption-contest-data/blob/master/contests/info/644/644.jpg
• 665 and 666, while matching the shark picture, appear to be near duplicates
• I think https://github.com/nextml/caption-contest-data/blob/master/contests/info/607/607.jpg shouldn't be a duplicate of 605.jpg , I think it should be a picture of death in a coffin.
• https://github.com/nextml/caption-contest-data/blob/master/contests/summaries/688_summary_KLUCB.csv contains the wrong contest id in each row.
• I've encountered some odd behavior within the csvs when the captions themselves contain line breaks. I don't have an example on hand (can find one if it comes up again), but while this should be escape-able with a reasonable csv parser, I pandas' was breaking for me.
• contest 656 (cowboy therapy) has captions for contest 655.
• contest 657 has the wrong captions.

I may encounter more as I play around with the data, and can keep sharing if helpful

Thanks!
Jack

[stsievert]

Thanks for the comment! I'll leave this issue open to track these. In the meantime, you might be interested in checking out http://nextml.github.io/caption-contest-data2, which has contests 660–763 (and is being updated more frequently).