Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385571566> ?p ?o ?g. }
Showing items 1 to 69 of
69
with 100 items per page.
- W4385571566 abstract "Transformer architectures are complex and their use in NLP, while it has engendered many successes, makes their interpretability or explainability challenging. Recent debates have shown that attention maps and attribution methods are unreliable (Pruthi et al., 2019; Brunner et al., 2019). In this paper, we present some of their limitations and introduce COCKATIEL, which successfully addresses some of them. COCKATIEL is a novel, post-hoc, concept-based, model-agnostic XAI technique that generates meaningful explanations from the last layer of a neural net model trained on an NLP classification task by using Non-Negative Matrix Factorization (NMF) to discover the concepts the model leverages to make predictions and by exploiting a Sensitivity Analysis to estimate accurately the importance of each of these concepts for the model. It does so without compromising the accuracy of the underlying model or requiring a new one to be trained. We conduct experiments in single and multi-aspect sentiment analysis tasks and we show COCKATIEL’s superior ability to discover concepts that align with humans’ on Transformer models without any supervision, we objectively verify the faithfulness of its explanations through fidelity metrics, and we showcase its ability to provide meaningful explanations in two different datasets.Our code is freely available: https://github.com/fanny-jourdan/cockatiel" @default.
- W4385571566 created "2023-08-05" @default.
- W4385571566 creator A5019863253 @default.
- W4385571566 creator A5025032659 @default.
- W4385571566 creator A5038564554 @default.
- W4385571566 creator A5056924085 @default.
- W4385571566 creator A5059933120 @default.
- W4385571566 creator A5082554847 @default.
- W4385571566 date "2023-01-01" @default.
- W4385571566 modified "2023-10-13" @default.
- W4385571566 title "COCKATIEL: COntinuous Concept ranKed ATtribution with Interpretable ELements for explaining neural net classifiers on NLP" @default.
- W4385571566 doi "https://doi.org/10.18653/v1/2023.findings-acl.317" @default.
- W4385571566 hasPublicationYear "2023" @default.
- W4385571566 type Work @default.
- W4385571566 citedByCount "0" @default.
- W4385571566 crossrefType "proceedings-article" @default.
- W4385571566 hasAuthorship W4385571566A5019863253 @default.
- W4385571566 hasAuthorship W4385571566A5025032659 @default.
- W4385571566 hasAuthorship W4385571566A5038564554 @default.
- W4385571566 hasAuthorship W4385571566A5056924085 @default.
- W4385571566 hasAuthorship W4385571566A5059933120 @default.
- W4385571566 hasAuthorship W4385571566A5082554847 @default.
- W4385571566 hasBestOaLocation W43855715661 @default.
- W4385571566 hasConcept C119857082 @default.
- W4385571566 hasConcept C121332964 @default.
- W4385571566 hasConcept C154945302 @default.
- W4385571566 hasConcept C162324750 @default.
- W4385571566 hasConcept C165801399 @default.
- W4385571566 hasConcept C187736073 @default.
- W4385571566 hasConcept C204321447 @default.
- W4385571566 hasConcept C2776459999 @default.
- W4385571566 hasConcept C2780451532 @default.
- W4385571566 hasConcept C2781067378 @default.
- W4385571566 hasConcept C41008148 @default.
- W4385571566 hasConcept C50644808 @default.
- W4385571566 hasConcept C62520636 @default.
- W4385571566 hasConcept C66322947 @default.
- W4385571566 hasConcept C76155785 @default.
- W4385571566 hasConceptScore W4385571566C119857082 @default.
- W4385571566 hasConceptScore W4385571566C121332964 @default.
- W4385571566 hasConceptScore W4385571566C154945302 @default.
- W4385571566 hasConceptScore W4385571566C162324750 @default.
- W4385571566 hasConceptScore W4385571566C165801399 @default.
- W4385571566 hasConceptScore W4385571566C187736073 @default.
- W4385571566 hasConceptScore W4385571566C204321447 @default.
- W4385571566 hasConceptScore W4385571566C2776459999 @default.
- W4385571566 hasConceptScore W4385571566C2780451532 @default.
- W4385571566 hasConceptScore W4385571566C2781067378 @default.
- W4385571566 hasConceptScore W4385571566C41008148 @default.
- W4385571566 hasConceptScore W4385571566C50644808 @default.
- W4385571566 hasConceptScore W4385571566C62520636 @default.
- W4385571566 hasConceptScore W4385571566C66322947 @default.
- W4385571566 hasConceptScore W4385571566C76155785 @default.
- W4385571566 hasLocation W43855715661 @default.
- W4385571566 hasOpenAccess W4385571566 @default.
- W4385571566 hasPrimaryLocation W43855715661 @default.
- W4385571566 hasRelatedWork W1986582023 @default.
- W4385571566 hasRelatedWork W2883749686 @default.
- W4385571566 hasRelatedWork W2968260065 @default.
- W4385571566 hasRelatedWork W3006943036 @default.
- W4385571566 hasRelatedWork W3170815031 @default.
- W4385571566 hasRelatedWork W4200511449 @default.
- W4385571566 hasRelatedWork W4206534706 @default.
- W4385571566 hasRelatedWork W4229079080 @default.
- W4385571566 hasRelatedWork W4298141768 @default.
- W4385571566 hasRelatedWork W4299487748 @default.
- W4385571566 isParatext "false" @default.
- W4385571566 isRetracted "false" @default.
- W4385571566 workType "article" @default.