Question1 Consider the word embeddings of restaurant, bottle, and galaxy. Note that restaurant is neither a synonym of bottle nor of galaxy. The cosine similarity between the embedding of restaurant and bottle is likely to be _________ the similarity between the embeddings of restaurant and galaxy. Group of answer choices

Similar to

Greater than

Smaller than

Question2

The document-term matrix obtained by a CountVectorizer on a large corpus is likely to have _______ columns than/as the document-features matrix obtained using pre-trained embeddings.

Group of answer choices

The same

More

Fewer

Question 3

When implementing a semantic search engine, training embeddings on our own data set generally leads to a better performance than using pre-trained embeddings.

Group of answer choices

True

False

Question 4

Consider four words w1,w2,w3,w4. To check whether w1 is to w2 as w3 is to w4 (e.g., king is to man as queen is to woman), the most appropriate equation whose validity needs to be checked is: w1w2=w3w4w1+w2=w3w4w2w1=w4w3