Differences

This shows you the differences between two versions of the page.

--- ml:confidence [2021/02/25 09:59] – jmflanig
+++ ml:confidence [2023/06/15 07:36] (current) – external edit 127.0.0.1
@@ Line 5: / Line 5: @@
 ===== In NLP =====
-(search [[https://www.aclweb.org/anthology/|acl anthology]] for "confidence scores")
+(search [[https://www.aclweb.org/anthology/|ACL Anthology]] for "confidence scores")
   * [[https://www.aclweb.org/anthology/N04-4028.pdf|Culotta & McCallum 2003 - Confidence Estimation for Information Extraction]] Uses three evaluation metrics of confidence scores:
      * "Pearson’s r, a correlation coefficient ranging from -1 to 1 that measures the correlation between a confidence score and whether or not the field (or record) is correctly labeled."
      * "average precision, used in the Information Retrieval community... the precision at each point in the ranked list where a relevant document is found and then averages these values. Instead of ranking documents by their relevance score, here we rank fields (and records) by their confidence score, where a correctly labeled field is analogous to a relevant document"
-     * "accuracy-coverage graph. Better confidence estimates push the curve to the upper-right"
+     * "accuracy-coverage graph. Better confidence estimates push the curve to the upper-right"  Precision-recall curve.  See fig 1.
   * [[https://www.aclweb.org/anthology/C04-1046.pdf|2004 - Confidence Estimation for Machine Translation]]
   * [[https://www.aclweb.org/anthology/P08-2055.pdf|2008 - Computing Confidence Scores for All Sub Parse Trees]]
   * [[https://www.aclweb.org/anthology/P11-1022.pdf|Nguyen Bach 2011  - Goodness: A Method for Measuring Machine Translation Confidence]] Has a good explanation of MT confidence
+  * [[https://www.aclweb.org/anthology/N12-1068.pdf|2012 - Are You Sure? Confidence in Prediction of Dependency Tree Edges]]
   * [[https://www.aclweb.org/anthology/P18-1069.pdf|2018 - Confidence Modeling for Neural Semantic Parsing]]  Measures "the relationship between confidence scores and F1 using Spearman’s ρ correlation coefficient which varies between −1 and 1 (0 implies there is no correlation)."
   * [[https://www.aclweb.org/anthology/W19-8671.pdf|2019 - Modeling Confidence in Sequence-to-Sequence Models]]
   * [[https://www.aclweb.org/anthology/2020.acl-main.188.pdf|2020 - Calibrating Structured Output Predictors for Natural Language Processing]]
+  * [[https://arxiv.org/pdf/2006.09462.pdf|2020 - Selective Question Answering under Domain Shift]]
+  * [[https://arxiv.org/pdf/2102.08501.pdf|Lahlou et al 2021 - DEUP: Direct Epistemic Uncertainty Prediction]]
+  * [[https://arxiv.org/pdf/2204.06546.pdf|Zerva et al 2022 - Better Uncertainty Quantification for Machine Translation Evaluation]]  See the related work
+===== Other Areas =====
+  * [[https://arxiv.org/pdf/1706.02690.pdf|Liang et al 2017 - Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networks]]