Word Embeddings (Word2Vec)
AI · Intermediate
Definition
Word embeddings represent words as dense vectors so similar words are close together.
Gives models a numeric sense of meaning.
Diagram
Flowchart
Example: Vector arithmetic
king, man, woman = [5, 3], [2, 1], [2, 4]
print([k - m + w for k, m, w in zip(king, man, woman)])Output: [5, 6]
Key points
- Contextual embeddings (BERT) vary by sentence
- Typical size 100–1000 dims
Interview questions
Static vs contextual embeddings?
Static give one vector per word; contextual change with surrounding text.