Deep Learning for NLP Quiz

What is the definition of 'meaning' in the context of NLP?

• the idea represented by a word, phrase, etc.

• the color of a word

• the size of a word

• the shape of a word

What is the problem with using a taxonomy like WordNet for representing 'meaning' in a machine?

• Too few synonyms

• Great as resource but missing nuances

• Too many synonyms

• Not enough hypernyms

What is the problem with using a 'one-hot' representation for words in NLP?

• Not enough storage

• Too few dimensions

• Too many dimensions

• Very high dimensional and require a lot of storage

What is the purpose of using a co-occurrence matrix in NLP?

• To capture both syntactic and semantic information

• To capture only syntactic information

• To capture only semantic information

• To capture only hypernyms

What is the main advantage of using low-dimensional word vectors in NLP?

• Faster computation

• Require less storage

• Ability to propagate any information into them via neural networks

• Capture more dimensions

What is the objective function of Word2Vec?

• Maximize log probability of any context word given the current center word

• Minimize log probability of any context word given the current center word

• Maximize log probability of any center word given the current context word

• Minimize log probability of any center word given the current context word

What is the main advantage of using GloVe for word representation?

• Capture more dimensions

• Require less storage

• Faster computation

• Fast training

What is the purpose of using a pretrained embedding matrix in future models?

• Faster computation

• Ability to propagate any information into them via neural networks

• Capture more dimensions

• Require less storage

What is the main benefit of deep learned word vectors in NLP?

• Capture more dimensions

• Require less storage

• Faster computation

• Ability to propagate any information into them via neural networks

What is the purpose of using linear substructures in GloVe?

• To capture only hypernyms

• To capture only semantic information

• To capture only syntactic information

• To capture as much as possible the meaning specified by the juxtaposition of two words