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

Question 1

What is the encoder-decoder architecture in NLP?

Accepted Answer

A model where the encoder compresses an input sequence into a context representation and the decoder generates the output sequence from it

Answer

A model that only encodes text without generating output

Answer

A two-layer feed-forward network for text classification

Answer

A model that encodes text as images for multimodal tasks

Question 2

What is the BLEU score used to evaluate in NLP?

Accepted Answer

Machine translation quality by measuring n-gram overlap between the generated output and human reference translations

Answer

Sentiment analysis accuracy on benchmark datasets

Answer

The perplexity of a language model on a held-out corpus

Answer

Named entity recognition F1 on CoNLL datasets

Question 3

What is the attention mechanism's key benefit in neural machine translation?

Accepted Answer

It allows the decoder to focus on relevant parts of the source sentence at each decoding step, overcoming the fixed-length bottleneck

Answer

It reduces the vocabulary size of the target language model

Answer

It speeds up training by reducing the number of parameters

Answer

It eliminates the need for parallel bilingual training data

Question 4

What is beam search used for in sequence generation?

Accepted Answer

Exploring multiple candidate output sequences simultaneously and returning the most probable one

Answer

Finding the shortest path in a dependency parse tree

Answer

Selecting the best training batch based on difficulty

Answer

Pruning the vocabulary to only the top k most frequent words

Question 5

What is the main difference between statistical machine translation (SMT) and neural machine translation (NMT)?

Accepted Answer

SMT uses probabilistic phrase-based models with separately trained components, while NMT uses end-to-end neural networks trained jointly

Answer

SMT requires more data than NMT

Answer

NMT cannot handle morphologically rich languages

Answer

SMT is faster at inference time than NMT

Question 6

What is back-translation in machine translation?

Accepted Answer

Translating monolingual target-language text into the source language to create synthetic parallel training data

Answer

Translating a document and then translating it back to verify accuracy

Answer

A technique to reverse the order of tokens in training data

Answer

Augmenting training data by replacing words with their antonyms