Which metric is used to quantify how close two word vectors are in meaning?

Get ready for the GARP Risk and AI Exam with flashcards and multiple choice questions. Each question comes with hints and explanations. Prepare for success!

Multiple Choice

Which metric is used to quantify how close two word vectors are in meaning?

Measuring semantic similarity between word vectors is about comparing their directions in the embedding space. Cosine similarity does exactly that: it computes the cosine of the angle between two vectors, yielding a value near 1 when they point in the same direction, near -1 when they point in opposite directions, and near 0 when they are orthogonal. Because it ignores magnitude, it focuses on orientation, which better reflects meaning regardless of how long the vectors are due to word frequency or other factors. This makes cosine similarity the natural choice for assessing how close two word vectors are in meaning. Euclidean distance, by contrast, cares about both direction and magnitude, so two words with similar meanings but different vector lengths can appear farther apart than expected. Jaccard similarity is designed for comparing sets and doesn’t readily apply to continuous, real-valued word embeddings. Pearson correlation looks at linear relationships across dimensions and isn’t as well aligned with the intuitive notion of directional similarity between word vectors.

Measuring semantic similarity between word vectors is about comparing their directions in the embedding space. Cosine similarity does exactly that: it computes the cosine of the angle between two vectors, yielding a value near 1 when they point in the same direction, near -1 when they point in opposite directions, and near 0 when they are orthogonal. Because it ignores magnitude, it focuses on orientation, which better reflects meaning regardless of how long the vectors are due to word frequency or other factors. This makes cosine similarity the natural choice for assessing how close two word vectors are in meaning.

Euclidean distance, by contrast, cares about both direction and magnitude, so two words with similar meanings but different vector lengths can appear farther apart than expected. Jaccard similarity is designed for comparing sets and doesn’t readily apply to continuous, real-valued word embeddings. Pearson correlation looks at linear relationships across dimensions and isn’t as well aligned with the intuitive notion of directional similarity between word vectors.

More practice questions