In the last post we saw a very simple approach to represent words in the vector form. Now we'll take a look at how we can generate word vectors using a neural network in an unsupervised manner. We are hoping that we'll learn the vector representation similar to as we learned the weight parameters for the neural network, and at end of the process we'll have something which will capture the syntactic or semantic relationship between the words better than what we got by creating the co-occurance matrix.

References:

• Hierarchical distributed representations for statistical language modeling
• word2vec
• Efficient Estimation of Word Representations in Vector Space
• word2vec Parameter Learning Explained
• Word2Vec: an introduction
• Deep Learning, NLP, and Representations
• Distributed Representations of Words and Phrases and their Compositionality