word embeddings

My thoughts on word2vec

If language is a medium of communication then words are symbols of communication. Words represent some concepts which when agreed upon by both the parties involved, make the information conveyed unambiguously. How do these words behave? Can we somehow model the behaviour of words? Let us consider a corpus as some closed system. We care only about the particles inside this closed system and not about any outside influence. In thermodynamics, the particles of interest are molecules and their kinetic energies.

Unsupervised most frequent sense detection using word embeddings