Why might embedding models place eggplant and aubergine close in vector space?

Why might embedding models place eggplant and aubergine close in vector space?

• A. Because there is a universal tag for synonyms.
• B. Because synonyms always share exact spelling.
• C. Because words with similar contexts tend to have high similarity in embedding space.
• D. Because the model memorizes every sentence verbatim.

Answer: (C)

Word embeddings rely on the idea that words used in similar contexts have similar meanings. When two words share many similar contexts, their vectors in the embedding space become close to each other. Eggplant and aubergine are just two names for the same thing, so they tend to appear in the same kinds of sentences—recipes, grocery lists, vegetable sections, cooking terms—which makes their contextual patterns align. That alignment pulls their vectors together, reflecting their semantic similarity. The other options don’t capture how embeddings work: there isn’t a universal tagging system for synonyms in these models, synonyms don’t require identical spelling, and embeddings don’t memorize every sentence verbatim; they generalize from patterns in the data.

Word embeddings rely on the idea that words used in similar contexts have similar meanings. When two words share many similar contexts, their vectors in the embedding space become close to each other. Eggplant and aubergine are just two names for the same thing, so they tend to appear in the same kinds of sentences—recipes, grocery lists, vegetable sections, cooking terms—which makes their contextual patterns align. That alignment pulls their vectors together, reflecting their semantic similarity. The other options don’t capture how embeddings work: there isn’t a universal tagging system for synonyms in these models, synonyms don’t require identical spelling, and embeddings don’t memorize every sentence verbatim; they generalize from patterns in the data.