Distinguish static word embeddings from contextual text embeddings in workplace retrieval tasks.
The move: add enough context for the vector to mean the right thing. A word by itself is often under-specified. In static word embeddings, one word can have one general location. That helps with broad relationships, but it struggles when the same word does different jobs in different business contexts. Contextual embeddings improve the situation by using surrounding text. The sentence "the customer opened a chargeback" and the sentence "we charge back the campaign cost to EMEA" should not behave as the same query. The surrounding words tell the model which neighborhood of meaning matters. Sentence and passage embeddings make…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in