4. Neural Methods

Word Embeddings — Quiz

Test your understanding of word embeddings with 5 practice questions.

Practice Questions

Question 1

In the context of Word2Vec's skip-gram model, if the vocabulary size is $V$ and the embedding dimension is $D$, what are the dimensions of the input-to-hidden weight matrix and the hidden-to-output weight matrix, respectively?

Question 2

Consider the training objective of GloVe. If $P_{ij}$ is the co-occurrence probability of word $i$ and word $j$, and $\vec{v}_i$ and $\vec{v}_j$ are their respective word vectors, which of the following expressions best represents the core of the GloVe objective function?

Question 3

When performing vector algebra on word embeddings, if we have the vectors for 'man' ($\vec{M}$), 'woman' ($\vec{W}$), and 'doctor' ($\vec{D}$), which operation would most likely yield a vector representing 'female doctor'?

Question 4

A researcher is evaluating a new set of word embeddings using a word analogy task. They find that the analogy 'king - man + woman = queen' holds with high accuracy, but the analogy 'doctor - surgeon + nurse = ?' yields inconsistent results. What does this suggest about the embeddings?

Question 5

In the context of evaluating word embeddings, an 'extrinsic evaluation' method would typically involve which of the following?