Learn Before
Explaining Divergent Translations
A machine translation model is tasked with translating the English sentence: 'She saw a bat.' On two separate occasions, it produces two different French translations:
- 'Elle a vu une chauve-souris.' (She saw a bat [the animal].)
- 'Elle a vu une batte.' (She saw a bat [for baseball].)
Assuming the model is functioning correctly, explain how the step-by-step generation process can lead to two different, plausible translations from the exact same input. Focus your explanation on what happens at the point of translating the word 'bat'.
0
1
Tags
Ch.1 Pre-training - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Application in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
A machine translation system generates an output sentence one word at a time, where each new word is chosen based on the original input sentence and the words already generated. When translating a long, complex sentence, the system produces an output that starts accurately but then devolves into a repetitive, nonsensical phrase (e.g., '...the boy is going to the is going to the is going to the...'). What is the most likely flaw in the generation process that would cause this specific type of error?
A machine translation model is translating the English sentence 'The cat sat' into French. It has already generated the tokens 'Le' and 'chat'. Arrange the following actions in the correct order to generate the very next token in the sequence.
Explaining Divergent Translations