Using Hackernews search API as a compressor, computing indexation distance between terms:
Cow + Pig 0.4843358722880948
Hen + Sheep 0.5328799348807779
Principles from Kolmogorov Complexity (Compression Distance) can give many approximate answers, even in the context of an IQ test.
Multiple choice is particularly easy, if you are allowed to note the result counts on queries like "Answer A + Keywords from Question" and "Answer B + Keywords from Question"
Speaking of cultural differences mentioned in the post -- You could even search locally on these terms and notice a difference in the semantic distances. If cows were to go extinct, in a 100 years, the search index would probably reflect on them being the least fit candidate for a set of three farm animals.
Multiple choice is particularly easy, if you are allowed to note the result counts on queries like "Answer A + Keywords from Question" and "Answer B + Keywords from Question"
Speaking of cultural differences mentioned in the post -- You could even search locally on these terms and notice a difference in the semantic distances. If cows were to go extinct, in a 100 years, the search index would probably reflect on them being the least fit candidate for a set of three farm animals.
[1] http://homepages.cwi.nl/~paulv/papers/amdug.pdf