1.
What output shape would you expect from model.encode() using the model 'all-MiniLM-L6-v2' for a single sentence?
Select the correct answer.
The 'all-MiniLM-L6-v2' model produces embeddings with 384 dimensions. So for a single sentence, model.encode() returns a vector of shape (384,) — a list of 384 float values.
2.
What does the following code return: model.encode(['Hello world'])?
Select the correct answer.
SentenceTransformer's encode() method returns a numpy array representing the embedding vector for the input text. For a single sentence, you get one vector (a 1D array of floats with as many elements as the model's output dimensions).
3.
What is the main limitation of Word2Vec compared to transformer-based embedding models?
Select the correct answer.
Word2Vec produces static (fixed) embeddings. The word 'bank' in 'river bank' gets the same vector as 'bank' in 'financial bank'. Transformer-based models (BERT, etc.) are contextual — they read the whole sentence and produce different vectors for the same word in different contexts.
