Quiz: Word Embedding

1. What is the purpose of word embeddings in NLP?

To convert words into numerical vectors for machine learning

To display font styles in documents

To translate audio into text

To generate HTML markup

2. How are similar words typically represented in embedding space?

In reverse alphabetical order

With very different vector lengths

In close proximity to each other

Using the same vector

3. What does each dimension in a word embedding vector typically represent?

A frequency of punctuation

A specific sentence

A document type

A latent linguistic attribute

4. Which model provides context-sensitive embeddings?

Bag-of-Words

TF-IDF

Word2Vec

BERT

5. What is a key difference between traditional and contextual word embeddings?

Traditional embeddings change with each sentence

Contextual embeddings use image data

Traditional embeddings require no training

Contextual embeddings vary depending on word usage in context

6. Which type of embedding would most likely capture the difference between 'king' and 'queen'?

Distributed word embedding

TF-IDF vector

Character-level count

One-hot encoding

7. In embedding visualizations, what does clustering typically indicate?

Words have the same length

Words share similar semantic or grammatical features

Words were generated synthetically

Words are rarely used

8. Why are high-dimensional embeddings often used in practice?

They simplify network architectures

They capture more semantic detail and word relationships

They reduce training time

They generate more documents

9. What common analogy demonstrates the power of word embeddings?

'King' - 'Man' + 'Woman' = 'Queen'

'Red' + 'Yellow' = 'Orange'

'Dog' + 'Tree' = 'Forest'

'Word' * 'Vector' = 'Matrix'

10. Which of the following is an application that relies on word embeddings?

Image classification

Audio transcription

Video rendering

Sentiment analysis

Quiz 27: Word Embedding