22,728 Hits in 6.5 sec

Improving Neural Model Performance through Natural Language Feedback on Their Explanations [article]

Aman Madaan, Niket Tandon, Dheeraj Rajagopal, Yiming Yang, Peter Clark, Keisuke Sakaguchi, Ed Hovy
2021 arXiv   pre-print
Our goal is to allow users to interactively correct explanation structures through natural language feedback.  ...  We introduce MERCURIE - an interactive system that refines its explanations for a given reasoning task by getting human feedback in natural language.  ...  Acknowledgments This material is partly based on research sponsored in part by the Air Force Research Laboratory under agreement number FA8750-19-2-0200. The U.S.  ... 
arXiv:2104.08765v1 fatcat:xoid6dkq2jaxdji5auywspxzma

A Review on Explainability in Multimodal Deep Neural Nets

Gargi Joshi, Rahee Walambe, Ketan Kotecha
2021 IEEE Access  
through language.  ...  natural language explanation for VQA using scene graph and visual attention mechanism.  ... 
doi:10.1109/access.2021.3070212 fatcat:5wtxr4nf7rbshk5zx7lzbtcram

ALICE: Active Learning with Contrastive Natural Language Explanations [article]

Weixin Liang, James Zou, Zhou Yu
2020 arXiv   pre-print
We propose Active Learning with Contrastive Explanations (ALICE), an expert-in-the-loop training framework that utilizes contrastive natural language explanations to improve data efficiency in learning  ...  ALICE learns to first use active learning to select the most informative pairs of label classes to elicit contrastive natural language explanations from experts.  ...  Acknowledgments We would like to sincerely thank EMNLP 2020 Chairs and Reviewers for their review efforts and helpful feedback.  ... 
arXiv:2009.10259v1 fatcat:u4d44vcv75enjgb46b3nipbphe

Speak to your Parser: Interactive Text-to-SQL with Natural Language Feedback [article]

Ahmed Elgohary, Saghar Hosseini, Ahmed Hassan Awadallah
2020 arXiv   pre-print
We focus on natural language to SQL systems and construct, SPLASH, a dataset of utterances, incorrect SQL interpretations and the corresponding natural language feedback.  ...  We study the task of semantic parse correction with natural language feedback.  ...  Acknowledgments We thank our ACL reviewers for their feedback and suggestions.  ... 
arXiv:2005.02539v2 fatcat:dhlmkldgorc63grptwdepolr3u

Automatic Generation of Natural Language Explanations [article]

Felipe Costa, Sixun Ouyang, Peter Dolog, Aonghus Lawlor
2017 arXiv   pre-print
In this paper, we propose a method for the automatic generation of natural language explanations, for predicting how a user would write about an item, based on user ratings from different items' features  ...  We design a character-level recurrent neural network (RNN) model, which generates an item's review explanations using long-short term memories (LSTM).  ...  ACKNOWLEDGMENTS This work is supported by Science Foundation Ireland through through the Insight Centre for Data Analytics under grant number SFI/12/RC/2289, and Conselho Nacional de Desenvolvimento Científico  ... 
arXiv:1707.01561v1 fatcat:6xrvjlzabvg3nh5r3asvxoozsa

Practical Benefits of Feature Feedback Under Distribution Shift [article]

Anurag Katakkar, Weiqin Wang, Clay H. Yoo, Zachary C. Lipton, Divyansh Kaushik
2021 arXiv   pre-print
By contrast, on natural language inference tasks, performance remains comparable. Finally, we compare those tasks where feature feedback does (and does not) help.  ...  In experiments addressing sentiment analysis, we show that feature feedback methods perform significantly better on various natural out-of-domain datasets even absent differences on in-domain evaluation  ...  Improv- ing the adversarial robustness and interpretability of deep neural networks by regularizing their input gradients. In AAAI Conference on Artificial Intelli- gence.  ... 
arXiv:2110.07566v1 fatcat:lrqekvfruzdgzc75ehgtsdqmde

Improving Explainable Recommendations with Synthetic Reviews [article]

Sixun Ouyang and Aonghus Lawlor and Felipe Costa and Peter Dolog
2018 arXiv   pre-print
Besides language model evaluation methods, we employ DeepCoNN, a novel review-oriented recommender system using a deep neural network, to evaluate the recommendation performance of generated reviews by  ...  To our knowledge, this presents the first machine-generated natural language explanations for rating prediction.  ...  We build 6 explanation generation models in this framework, and analyse their performance on both natural language generating and recommendation interpretation tasks.  ... 
arXiv:1807.06978v1 fatcat:vudlqa5zqbfztd3ulcko4ojx4a

Explanation-Based Human Debugging of NLP Models: A Survey [article]

Piyawat Lertvittayakumjorn, Francesca Toni
2021 arXiv   pre-print
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  ...  Also, we thank Brian Roark and Cindy Robinson for their technical support concerning the submission system.  ... 
arXiv:2104.15135v3 fatcat:u6equfv2yrbhxeiexyrpzadip4

Teaching the Machine to Explain Itself using Domain Knowledge [article]

Vladimir Balayan, Pedro Saleiro, Catarina Belém, Ludwig Krippahl, Pedro Bizarro
2020 arXiv   pre-print
Moreover, we collect the domain feedback from a pool of certified experts and use it to ameliorate the model (human teaching), hence promoting seamless and better suited explanations.  ...  Furthermore, obtained results indicate that human teaching can further improve the explanations prediction quality by approximately 13.57%.  ...  Acknowledgements The project CAMELOT (reference POCI-01-0247-FEDER-045915) leading to this work is cofinanced by the ERDF -European Regional Development Fund through the Operational Program for Competitiveness  ... 
arXiv:2012.01932v1 fatcat:igzh3ll45bbfzcc5smfgldywiu

NPRF: A Neural Pseudo Relevance Feedback Framework for Ad-hoc Information Retrieval

Canjia Li, Yingfei Sun, Ben He, Le Wang, Kai Hui, Andrew Yates, Le Sun, Jungang Xu
2018 Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing  
Extensive experiments on two standard test collections confirm the effectiveness of the proposed NPRF framework in improving the performance of two state-of-theart neural IR models.  ...  Pseudo relevance feedback (PRF) is commonly used to boost the performance of traditional information retrieval (IR) models by using top-ranked documents to identify and weight new query terms, thereby  ...  Acknowledgments This work is supported in part by the National Natural Science Foundation of China (61433015/61472391), and the Beijing Natural Science Foundation under Grant No. (4162067/4142050).  ... 
doi:10.18653/v1/d18-1478 dblp:conf/emnlp/LiSHWHYSX18 fatcat:jxizmm2nc5glldd4e5afq2pzbm

LEAN-LIFE: A Label-Efficient Annotation Framework Towards Learning from Explanation [article]

Dong-Ho Lee, Rahul Khanna, Bill Yuchen Lin, Jamin Chen, Seyeon Lee, Qinyuan Ye, Elizabeth Boschee, Leonardo Neves, Xiang Ren
2020 arXiv   pre-print
On three popular NLP tasks (named entity recognition, relation extraction, sentiment analysis), we find that using this enhanced supervision allows our models to surpass competitive baseline F1 scores  ...  Successfully training a deep neural network demands a huge corpus of labeled data.  ...  We would like to thank all the collaborators in USC INK research lab for their constructive feedback on the work.  ... 
arXiv:2004.07499v1 fatcat:yd7exkzxq5bg7pz7ndihfgplqu

Learning to Learn Semantic Parsers from Natural Language Supervision

Igor Labutov, Bishan Yang, Tom Mitchell
2018 Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing  
As humans, we often rely on language to learn language. For example, when corrected in a conversation, we may learn from that correction, over time improving our language fluency.  ...  We construct a novel dataset of natural language feedback in a conversational setting, and show that our method is effective at learning a semantic parser from such natural language supervision.  ...  Our work is similar in that natural language is used as additional supervision during learning, however, our natural language annotations consist of user feedback on system predictions instead of explanations  ... 
doi:10.18653/v1/d18-1195 dblp:conf/emnlp/LabutovYM18 fatcat:7avxs2tjozdurcsdsqoyjuvas4

PyTorch, Explain! A Python library for Logic Explained Networks [article]

Pietro Barbiero, Gabriele Ciravegna, Dobrik Georgiev, Franscesco Giannini
2021 arXiv   pre-print
is a Python module integrating a variety of state-of-the-art approaches to provide logic explanations from neural networks. This package focuses on bringing these methods to non-specialists.  ...  Acknowledgments and Disclosure of Funding We thank Stefano Melacci, Pietro Lió, and Marco Gori for useful feedback and suggestions.  ...  Thanks to their explainable nature, LENs can be effectively used to understand the behavior of an existing algorithm, to reverse engineer products, to find vulnerabilities, or to improve system design.  ... 
arXiv:2105.11697v2 fatcat:blfz6k527vbqhovafhk3cmiuuy

A Study on Multimodal and Interactive Explanations for Visual Question Answering [article]

Kamran Alipour, Jurgen P. Schulze, Yi Yao, Avi Ziskind, Giedrius Burachas
2020 arXiv   pre-print
Explainability and interpretability of AI models is an essential factor affecting the safety of AI.  ...  The results indicate that the explanations help improve human prediction accuracy, especially in trials when the VQA system's answer is inaccurate.  ...  The core model extracts features from natural language questions as well as images, combines them, and generates a natural language answer.  ... 
arXiv:2003.00431v1 fatcat:ycjgoi65mbdwpgfkthozrhjt6e

Explainable Automated Essay Scoring: Deep Learning Really Has Pedagogical Value

Vivekanandan Kumar, David Boulanger
2020 Frontiers in Education  
In doing so, it evaluates the impact of deep learning (multilayer perceptron neural networks) on the performance of AES.  ...  It leverages the state-of-the-art in natural language processing, applying feature selection on a pool of 1592 linguistic indices that measure aspects of text cohesion, lexical diversity, lexical sophistication  ...  Finally, this paper highlights the means to debug and compare the performance of predictive models through their explanations.  ... 
doi:10.3389/feduc.2020.572367 fatcat:eumqfzpovfbtjbq3cx6z2mp6pm
« Previous Showing results 1 — 15 out of 22,728 results