Vector space models are a family of models that represent data as vectors within space. They are well-suited to NLP tasks that involve measuring similarity between words or documents. Similarity is usually measured by either the Euclidean Distance or Cosine Similarity, the latter of which is usually preferred to avoid bias incurred when documents are of different lengths. For a given word in the vocabulary, the word that has a vector representation with the highest similarity to the word of interest would be considered the word closest in meaning.