Advanced Deep Learning with Python

fastText

fastText (https://fasttext.cc/) is a library for learning word embeddings and text classification created by the Facebook AI Research (FAIR) group. Word2Vec treats each word in the corpus as an atomic entity and generates a vector for each word, but this approach ignores the internal structure of the words. In contrast, fastText decomposes each word, w, to a bag of character n-grams. For example, if n = 3, we can decompose the word there to the character 3-grams and the special sequence <there> for the whole word:

<th, the, her, ere, re>

Note the use of the special characters < and > to indicate the start and the end of the word. This is necessary to avoid mismatching between n-grams from different words. For example, the word