Count based Word Embeddings

#deep-learning #interview

Count based word embeddings depends on the frequency of the word in the sentence / paragraph / documents. There are several methods for count based word embeddings:

  1. TF-IDF
  2. Bag of Words

Limitations:

  1. Lack of context and word order
  2. Lack of semantic understanding
  3. Doesn't account for Homonym or Polysemy
  4. Limitation of Vocabulary
  5. Poor performance with short text

References

Related Notes