tm4roon / survey

Survey on machine learning.
14 stars 1 forks source link

Confidence Modeling for Neural Semantic Parsing #122

Open tm4roon opened 4 years ago

tm4roon commented 4 years ago

Confidence Modeling for Neural Semantic Parsing

本論文ではSemantic Parsingにおける出力の確信度/不確かさを測る方法として、次の3つの観点に着目している.

また、不確かさが入力文のどの部分から発生しているのかを出力から逆算する、Uncertainty Backpropagationを提案している. 結果として、アテンション機構を利用するよりもうまく不確かさを検出できる.

文献情報