What describes dense vector representations of words that capture meaning and relationships?

What describes dense vector representations of words that capture meaning and relationships?

• A. One-Hot Encoding
• B. Word Embeddings
• C. Sparse Vectors
• D. Bag-of-Words

Answer: (B)

Dense vector representations of words that capture meaning and relationships are Word Embeddings. They map each word to a dense, low-dimensional vector learned from large text data, so words with similar meanings sit close together in the vector space. This setup also allows relational patterns to emerge, such that certain vector operations reflect analogies (like how directions between words can align to show relationships such as king minus man plus woman approximating queen). Embeddings are produced by models like Word2Vec, GloVe, or fastText and enable measuring similarity with metrics such as cosine similarity and better generalization across contexts. This differs from one-hot encoding, which gives each word its own position in a very high-dimensional, sparse vector with no information about similarity between words. Bag-of-Words similarly yields sparse, count-based representations that ignore word order and semantics. Dense embeddings address both sparsity and semantic relationships, making them the best description here.

Dense vector representations of words that capture meaning and relationships are Word Embeddings. They map each word to a dense, low-dimensional vector learned from large text data, so words with similar meanings sit close together in the vector space. This setup also allows relational patterns to emerge, such that certain vector operations reflect analogies (like how directions between words can align to show relationships such as king minus man plus woman approximating queen). Embeddings are produced by models like Word2Vec, GloVe, or fastText and enable measuring similarity with metrics such as cosine similarity and better generalization across contexts.

This differs from one-hot encoding, which gives each word its own position in a very high-dimensional, sparse vector with no information about similarity between words. Bag-of-Words similarly yields sparse, count-based representations that ignore word order and semantics. Dense embeddings address both sparsity and semantic relationships, making them the best description here.