Summary

In this chapter, we covered word embedding and why it is important in natural language processing. N-grams were used to show how the words are treated as a vector and how the count of words are stored to find the relevance. GloVe and word2vec are two common approaches to word embedding, where the word counts or probabilities are stored in vectors. Both of these approaches lead to high dimensionality, which is not feasible to process in the real world, especially on mobile devices or devices with less memory. We have seen two different approaches to reduce the dimensionality. In next chapter, Chapter 7, Information Retrieval we will see how information retrieval can be done from the unstructured format such as text.