Word2Vec is famous for demonstrating local linear properties on analogy tasks.
The task and dataset were introduced in Tomas Mikolov, Wen-tau Yih, and Geoffrey Zweig , Linguistic Regularities in Continuous Space Word Representations (NAATL 2013).
That paper gives a link to where the Syntactic task set can be downloaded from. Unfortunately, it is broken. Googling for it has not yielded me anything.
The semantic set is derived from SemEval-2012 Task 2, so I am pretty sure I can track that down (haven't tried yet).
I know these both exist, and are generally available. People are still using them to benchmark variations on word-embeddings.
