As we have seen, word2vec is a very good technique for generating distributional similarity. There are other advantages of it as well, which I've listed here:
- Word2vec concepts are really easy to understand. They are not so complex that you really don't know what is happening behind the scenes.
- Using word2vec is simple and it has very powerful architecture. It is fast to train compared to other techniques.
- Human effort for training is really minimal because, here, human tagged data is not needed.
- This technique works for both a small amount of datasets and a large amount of datasets. So it is an easy-to-scale model.
- Once you understand the concept and algorithms, then you can replicate the whole concept and algorithms on your dataset as well.
- It does exceptionally well on capturing semantic similarity.
- As this is a kind of unsupervised approach...