I wonder if researchers working on textual entailment using explicit modelling tools (mathematical semantics / logic /formal semantics) have an explanation of recent success of neural networks in textual entailment (GLUE Benchmark)? Any links would be greatly appreciated.
-
2I am not researching formal semantics (syntax, speech understanding and production are my areas), and I do apply and ML methods and have a first person knowledge of them. If nobody writes an answer in a day more, I will. It's not going to be specific, naturally, but I can explain why neural nets are good at X for almost any X, for a specific notion "being good at X." Please tag me if it slips off my radar. – kkm Oct 8 '18 at 6:26
Add a comment
|
