Morphology: a field of study in which people study how a word is formed and how it connects with other words!
Coarse-grained vs fine-grained: Coarse-grained tags (three labels in sentiment analysis without having subtler tags / NER like location and no city, country, state, etc.) are more discrete and general whereas fine-grained are more specific.
Common sense reasoning 💁♀️: It is an intuitive understanding of the world and how it works (e.g., the president didn’t want to show the riot → this is true because it is dangerous for his position.)
Transduction Model A model that transduces the input and turns it into the output. (converts it - e.g., the energy of body is transduced into nervous signals in the brain.)
Transductive models ≠ Inductive models
<aside> 💡 Difference between transductive and inductive?
Transductive models see the whole dataset.
Inductive models only see the training dataset
</aside>
Label Smoothing - link Spreading noise in the output, because teaching a certain sequence of words with specific probability of happening is dangerous.
Deterministic Not conditional - based on determined facts.