Machine Translation and Encoder-Decoder Models

Question 1

Machine Translation

Answer 1

The use of computers to translate from one language to another.

Question 2

3 Components of an encoder-decoder network

Answer 2

1. Encoder
2. Context vector
3. Decoder

Question 3

3 Components of an encoder-decoder network:

Encoder

Answer 3

An encoder accepts an input sequence, and generates a corresponding sequence of contextualized representations.

LSTMs, GRUs, convolutional networks, and Transformers can all be employed as encoders.

Question 4

3 Components of an encoder-decoder network:

Context vector

Answer 4

A context vector, c, is a function of the contextualized representations from the encoder.

It conveys the essence of the input to the decoder.