Explanation-Based Human Debugging of NLP Models: A Survey
AbstractDebugging a machine learning model is hard since the bug usually involves the training data and the learning process. This becomes even harder for an opaque deep learning model if we have no clue about how the model actually works. In this survey, we review papers that exploi...
Main Authors: | , |
---|---|
Format: | Article |
Language: | English |
Published: |
The MIT Press
2021-01-01
|
Series: | Transactions of the Association for Computational Linguistics |
Online Access: | https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00440/108932/Explanation-Based-Human-Debugging-of-NLP-Models-A |
_version_ | 1811290779978563584 |
---|---|
author | Piyawat Lertvittayakumjorn Francesca Toni |
author_facet | Piyawat Lertvittayakumjorn Francesca Toni |
author_sort | Piyawat Lertvittayakumjorn |
collection | DOAJ |
description |
AbstractDebugging a machine learning model is hard since the bug usually involves the training data and the learning process. This becomes even harder for an opaque deep learning model if we have no clue about how the model actually works. In this survey, we review papers that exploit explanations to enable humans to give feedback and debug NLP models. We call this problem explanation-based human debugging (EBHD). In particular, we categorize and discuss existing work along three dimensions of EBHD (the bug context, the workflow, and the experimental setting), compile findings on how EBHD components affect the feedback providers, and highlight open problems that could be future research directions. |
first_indexed | 2024-04-13T04:18:35Z |
format | Article |
id | doaj.art-514cb33eaea84bc3be5e36199fb70be6 |
institution | Directory Open Access Journal |
issn | 2307-387X |
language | English |
last_indexed | 2024-04-13T04:18:35Z |
publishDate | 2021-01-01 |
publisher | The MIT Press |
record_format | Article |
series | Transactions of the Association for Computational Linguistics |
spelling | doaj.art-514cb33eaea84bc3be5e36199fb70be62022-12-22T03:02:54ZengThe MIT PressTransactions of the Association for Computational Linguistics2307-387X2021-01-0191508152810.1162/tacl_a_00440Explanation-Based Human Debugging of NLP Models: A SurveyPiyawat Lertvittayakumjorn0Francesca Toni1Department of Computing, Imperial College London, UK. pl1515@imperial.ac.ukDepartment of Computing, Imperial College London, UK. ftft@imperial.ac.uk AbstractDebugging a machine learning model is hard since the bug usually involves the training data and the learning process. This becomes even harder for an opaque deep learning model if we have no clue about how the model actually works. In this survey, we review papers that exploit explanations to enable humans to give feedback and debug NLP models. We call this problem explanation-based human debugging (EBHD). In particular, we categorize and discuss existing work along three dimensions of EBHD (the bug context, the workflow, and the experimental setting), compile findings on how EBHD components affect the feedback providers, and highlight open problems that could be future research directions.https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00440/108932/Explanation-Based-Human-Debugging-of-NLP-Models-A |
spellingShingle | Piyawat Lertvittayakumjorn Francesca Toni Explanation-Based Human Debugging of NLP Models: A Survey Transactions of the Association for Computational Linguistics |
title | Explanation-Based Human Debugging of NLP Models: A Survey |
title_full | Explanation-Based Human Debugging of NLP Models: A Survey |
title_fullStr | Explanation-Based Human Debugging of NLP Models: A Survey |
title_full_unstemmed | Explanation-Based Human Debugging of NLP Models: A Survey |
title_short | Explanation-Based Human Debugging of NLP Models: A Survey |
title_sort | explanation based human debugging of nlp models a survey |
url | https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00440/108932/Explanation-Based-Human-Debugging-of-NLP-Models-A |
work_keys_str_mv | AT piyawatlertvittayakumjorn explanationbasedhumandebuggingofnlpmodelsasurvey AT francescatoni explanationbasedhumandebuggingofnlpmodelsasurvey |