FastText Subword Embeddings Quiz

Answers and explanations appear as you take the quiz.

1. How does FastText represent a single word?
   • As a bag of character n-grams summed together
   • As one indivisible token only
   • As its definition from a dictionary
   • As a random vector
2. What major limitation of Word2Vec and GloVe does FastText overcome?
   • They run too fast
   • They cannot create vectors for out-of-vocabulary words
   • They ignore punctuation
   • They only work in English
3. For the word "where" with 3-character n-grams, which is a valid n-gram (with boundary markers)?
   • <where>only
   • xyz
   • whe
   • the
4. Which underlying training objective does FastText's embedding model build on?
   • GloVe matrix factorization
   • Sentence-level translation
   • Image classification loss
   • Skip-Gram with negative sampling
5. Why is FastText especially useful for languages like Finnish or Turkish?
   • They are morphologically rich with many inflected word forms that share subwords
   • They have very few words
   • They use no spaces
   • They have no rare words
6. Which separate method did Sennrich and colleagues describe for representing rare words in neural machine translation?
   • Byte-pair encoding as subword units
   • FastText’s hashed character n-gram bucket table
   • A full-word-only vocabulary that discards rare tokens
   • GloVe’s global co-occurrence matrix
7. How does FastText store learnable features for its character n-grams?
   • It stores every complete sentence as a separate vector
   • It assigns each n-gram a pixel grid in a convolutional decoder
   • It hashes n-grams into a fixed-size bucket table with learned vectors
   • It replaces all subword features with a hand-written dictionary
8. What can fastText's supervised mode do efficiently with its bag-of-features model?
   • Estimate the 3D shape of an object from a single image
   • Generate new audio waveforms from a spectrogram
   • Compute a global word-word co-occurrence matrix only
   • Classify text for tasks such as language identification or tagging