Natural Language Processing Machine Translation and Sequence-to-Sequence Models 2

Question 1

What is text summarization in NLP?

Accepted Answer

The task of automatically producing a shorter version of a document that retains the most important information

Answer

Summarization can be extractive (selecting key sentences) or abstractive (generating new sentences), both aiming to condense source content while preserving meaning.

Question 2

What is the difference between extractive and abstractive summarization?

Accepted Answer

Extractive summarization selects existing sentences from the source, while abstractive summarization generates novel sentences that may not appear verbatim in the source

Answer

Extractive systems copy salient spans directly; abstractive systems paraphrase and synthesize information, producing more fluent but harder-to-evaluate summaries.

Question 3

What is the ROUGE metric used for in NLP evaluation?

Accepted Answer

Evaluating text summarization quality by measuring n-gram recall overlap between generated and reference summaries

Answer

ROUGE (Recall-Oriented Understudy for Gisting Evaluation) measures the overlap of n-grams and longest common subsequences between system and reference summaries.

Question 4

What is the copy mechanism in sequence-to-sequence models?

Accepted Answer

A mechanism that allows the decoder to directly copy tokens from the source input sequence rather than always generating from the vocabulary

Answer

The copy mechanism (pointer networks) lets the model point to and copy source tokens, which is critical for tasks like summarization where proper nouns or rare words should be reproduced exactly.

Question 5

What is the teacher forcing technique in training sequence-to-sequence models?

Accepted Answer

Feeding the ground-truth previous token as input to the decoder at each step during training, rather than the model's own previous prediction

Answer

Teacher forcing speeds up convergence by providing correct previous tokens during training, though it can cause exposure bias since the model sees its own errors only at inference time.

Natural Language Processing Practice Test

Natural Language Processing Practice Test

Natural Language Processing Machine Translation and Sequence-to-Sequence Models 2