What does downstream tasks terminology mean in NLP? I saw this terminology used in several articles but I can't understand the idea behind it.
2 Answers
"Downstream tasks is what the field calls those supervised-learning tasks that utilize a pre-trained model or component". Derived from this blogpost: http://jalammar.github.io/illustrated-bert/
What is the "downstream task" in NLP
In supervised learning, you can think of "downstream task" as the application of the language model.
Example
- article classification: To tell whether the news is fake news? or Patent classification
- sequence labeling: assigns a class or label to each token in a given input sequence. See wiki page of sequence labeling
Reference
Related