I want to use pre-train word vectors (e.g., fasttest on Wikipedia) to find clusters of a set of words. However, in the list of words I have words like 'kindness', 'kind', 'kindly' and they fall in different clusters. That is sometimes words with similar part of speech are clusters together. I want to know how can I have word vectors that only captures meaning?
You can lemmatize or stem the words before using word2vec.
stemming library has several such algorithm implemented.