Show HN: Sense2vc – Semantic Analysis of the Reddit Hivemind

it would be cool to have this with a more neutral data set like wikipedia or google news.

I think the biggest problem with word2vec is that it only captures the context of the word and not the meaning. eg. the top results for "cold" includes "warm" because they're often used in the same sentence. Not sure if anyone's made progress on that front?

I fail to understand how word2vec is useful? It outputs a bunch of related words but there no link between the words so no way to make sens of the "relation" between input and output.