sem6_ai_week_8

Which preprocessing step converts “Running” to “run”?

• Lowercasing for word simplification

• Stop‑word removal operation applied

• Stemming using Porter method

• Lemmatization with vocabulary lookup

In a bag‑of‑words model the feature vector length equals the _____?

• Number of processed documents there

• Count of sentences per corpus

• Size of chosen stop‑word list

• Total unique tokens in vocabulary

The TF component in TF‑IDF is commonly computed as _____?

• Binary presence for each term

• Log of inverse term length

• Raw position index in text

• Term frequency within document

Given N documents, IDF = log (N / df) where df is _____?

• Document format numeric code

• Desired feature vector size

• Distinct feature hash bucket

• Number of docs containing term

One‑hot vectors are extremely _____ compared with dense embeddings.

• Computationally fast for GPUs

• Contextually aware of semantics

• Compact due to low dimension

• Sparse and memory inefficient

Word2Vec’s Skip‑gram objective maximizes probability of _____?

• Document class given features

• Global co‑occurrence statistics

• Subword n‑gram predictions

• Context words given target word

In negative sampling, “k” most commonly refers to _____?

• Kernel width inside network

• Layers stacked in the model

• Context window size chosen

• Randomly drawn negative samples

FastText improves rare‑word handling by using _____?

• Sentence‑level recurrent memory

• Attention across document parts

• Byte‑pair token compression rule

• Subword character n‑gram vectors

GloVe embeddings are trained by factorizing _____ matrices.

• Grammar rule transition count

• Neural attention weight maps

• Topic probability distributions

• Global word co‑occurrence counts

In Keras, Embedding(input_dim, output_dim) input_dim equals _____?

• Hidden state recurrent length

• Maximum review character size

• Batch dimension from dataset

• Vocabulary size including OOV