Actually there was:

1. Australia (country)
2. New Zealand (country)
3. Australia/New Zealand (region)

Likewise for all the individual countries that make up regions like South America, West Africa, etc.

The reason Australia doesn’t show up in the charts I made is that I just picked a selection of countries. NZ was interesting because its adoption of aquaculture was unusually high.

I was going to mention that. In fact I used WA to sanity-check the results I was getting. It’s extraordinary to see the kinds of natural language queries that it gets right, and also to see how gracefully it displays the history of a series when available.

lthough it doesn’t enable all that you’re doing here

Right.

I don’t believe there is any way to just access the raw data they use, however.

Yes, and that’s a key point. WA’s robust curation and normalization is, so far, a black box. No way to verify how it’s done, or collaborate with it.

Fascinating, thanks!

different encodings because it was merged a while ago and never cleaned up (this happens way too often, unfortunately)

Yes. It’s time-consuming even for somebody like me to correct those issues, and for most people quite beyond the pale.

I’ll have lots more to say about this — and to ask you and David — but the tool certainly had me at hello. Terrific work!

http://www.wolframalpha.com/input/?i=chicken%20production%20united%20states%20vs%20china

I don’t believe there is any way to just access the raw data they use, however.

very glad you like Gridworks and I totally agree on the ‘spidey sense tingling’. We (the Gridworks dev team) feel precisely the same way and are very excited about it: can’t wait to get it out and start building a healthy and diverse open development community of users and contributors around it.

One note about lost encodings: one awesome trick is to create a ‘custom numeric facet’ out of the numeric values of the unicode characters of the cell value (can you can obtain using the expression “value.unicode()’), then filter on the outliers (normally the highest values) and then transform the cell values with the expression ‘value.reinterpret(“utf-8″)’ and test visually with different encoding values until problems get fixed.

I found this operational recipe to be extremely helpful not only when the encoding information is lost (or Gridworks guesses the encoding wrong), but also and most importantly when the dataset contains data from different encodings because it was merged a while ago and never cleaned up (this happens way too often, unfortunately).