Filters








3,482 Hits in 10.9 sec

Are Pretrained Transformers Robust in Intent Classification? A Missing Ingredient in Evaluation of Out-of-Scope Intent Detection [article]

Jianguo Zhang, Kazuma Hashimoto, Yao Wan, Zhiwei Liu, Ye Liu, Caiming Xiong, Philip S. Yu
2022 arXiv   pre-print
Pre-trained Transformer-based models were reported to be robust in intent classification.  ...  In this work, we first point out the importance of in-domain out-of-scope detection in few-shot intent recognition tasks and then illustrate the vulnerability of pre-trained Transformer-based models against  ...  However, there is a lack of further study of pre-trained Transformers on few-shot fine-grained OOS detection where the OOS intents are more relevant to the in-scope intents.  ... 
arXiv:2106.04564v3 fatcat:5rjyxfm6tfaghcxijntdtu7jji

Are Pre-trained Transformers Robust in Intent Classification? A Missing Ingredient in Evaluation of Out-of-Scope Intent Detection

Jianguo Zhang, Kazuma Hashimoto, Yao Wan, Zhiwei Liu, Ye Liu, Caiming Xiong, Philip Yu
2022 Proceedings of the 4th Workshop on NLP for Conversational AI   unpublished
Pre-trained Transformer-based models were reported to be robust in intent classification.  ...  In this work, we first point out the importance of in-domain out-of-scope detection in few-shot intent recognition tasks and then illustrate the vulnerability of pre-trained Transformer-based models against  ...  However, there is a lack of further study of pre-trained Transformers on few-shot fine-grained OOS detection where the OOS intents are more relevant to the in-scope intents.  ... 
doi:10.18653/v1/2022.nlp4convai-1.2 fatcat:y5rjfo6mbvbffpypwesoxqwoju

Plex: Towards Reliability using Pretrained Large Model Extensions [article]

Dustin Tran, Jeremiah Liu, Michael W. Dusenberry, Du Phan, Mark Collier, Jie Ren, Kehang Han, Zi Wang, Zelda Mariet, Huiyi Hu, Neil Band, Tim G. J. Rudner (+14 others)
2022 arXiv   pre-print
tasks involving uncertainty (e.g., selective prediction, open set recognition), robust generalization (e.g., accuracy and proper scoring rules such as log-likelihood on in- and out-of-distribution datasets  ...  We devise 10 types of tasks over 40 datasets in order to evaluate different aspects of reliability on both vision and language domains.  ...  Acknowledgements We thank Ben Adlam, Dilip Krishnan, Ed Chi, Neil Houlsby, Rif A. Saurous, and Sharat Chikkerur for helpful discussions and feedback on earlier drafts of the paper.  ... 
arXiv:2207.07411v1 fatcat:txa5dne23fhxxcsqyurh2qfrly

Multi-word Entity Classification in a Highly Multilingual Environment

Sophie Chesney, Guillaume Jacquet, Ralf Steinberger, Jakub Piskorski
2017 Proceedings of the 13th Workshop on Multiword Expressions (MWE 2017)  
iii resources to predict the compositionality of MWEs. The program also included a panel discussion on the future directions of the MWE community and the SIGLEX Section.  ...  , and all involved participants for their interest in the workshop.  ...  Of course, any mistakes which remain are mine alone.  ... 
doi:10.18653/v1/w17-1702 dblp:conf/mwe/ChesneyJSP17 fatcat:bv7aavgth5eurmzuphuowtuuhq

A sequential distance-based approach for imputing missing data: Forward Imputation

Nadia Solaro, Alessandro Barbiero, Giancarlo Manzi, Pier Alda Ferrari
2016 Advances in Data Analysis and Classification  
In metrology, such experiments of international scope are carried out by the National Metrology Institutes in order to establish the degree of equivalence of national measurement standards known as the  ...  to fit the data and the application of inferential methods will be carried out in a more robust way.  ... 
doi:10.1007/s11634-016-0243-0 fatcat:yvrqlgllsbesbnvnzzci2egpl4

Gender as an 'Interplay of Rules': Detecting Epistemic Interplay of Medical and Legal Discourse with Sex and Gender Classification in Four Editions of the Dewey Decimal Classification

Melodie J. Fox
2015 Zenodo  
When groups of people are represented in classification systems, potential exists for them to be structurally or linguistically subordinated, erased or otherwise misrepresented (Olson & Schlegl, 2001).  ...  interplay with conceptualizations of similar concepts in four editions of the Dewey Decimal Classification.  ...  Epistemology in classification can be detected in a few ways.  ... 
doi:10.5281/zenodo.3472321 fatcat:rdtuyrm6ife3hnbqphx2ajmkr4

Frequency of injuries and health status of football players in Bosnia; classification by gender and age

Ratko Peric, Radojka Peric
2013 Zenodo  
Both males and females athletes were examined through systematic examination and medical records check out and diagnosis were classified using ICD-10 (10th revision of the International Statistical Classification  ...  A total number of 281 (30,6%) were younger than 15 years of age. Interesting was that there were no major differences in diagnoses between generations.  ...  It was evaluated at inclusion (PRE) after 4 weeks and after 10 weeks training (W4 and W10, respectively).  ... 
doi:10.5281/zenodo.4575221 fatcat:gps3ldi3crhmhjxoandbui77je

Framework for Deep Learning-Based Language Models using Multi-task Learning in Natural Language Understanding: A Systematic Literature Review and Future Directions

Rahul Manohar Samant, Mrinal Bachute, Shilpa Gite, Ketan Kotecha
2022 IEEE Access  
Combining these approaches may result in a more efficient and robust multi-task NLU.  ...  This SLR proposes building steps for a conceptual framework to achieve goals of enhancing the performance of language models in the field of NLU.  ...  out-of-domain task detection and task interference by using the active learning technique.  ... 
doi:10.1109/access.2022.3149798 fatcat:k3kdt4eryzdfpk5k6w62jtlskm

Recipes for Safety in Open-domain Chatbots [article]

Jing Xu, Da Ju, Margaret Li, Y-Lan Boureau, Jason Weston, Emily Dinan
2021 arXiv   pre-print
We introduce a new human-and-model-in-the-loop framework for both training safer models and for evaluating them, as well as a novel method to distill safety considerations inside generative models without  ...  We investigate a variety of methods to mitigate these issues in the context of open-domain generative dialogue models.  ...  While this is a larger scope than conversational models, much of the work discussed such as training classifiers to detect abusive content, and scoping out what qualifies as "abusive," is largely relevant  ... 
arXiv:2010.07079v3 fatcat:qvbchivryrcdrj2evt6awl37fm

Data Collection and Quality Challenges in Deep Learning: A Data-Centric AI Perspective [article]

Steven Euijong Whang, Yuji Roh, Hwanjun Song, Jae-Gil Lee
2022 arXiv   pre-print
As a result, data-centric AI practices are now becoming mainstream. Unfortunately, many datasets in the real world are small, dirty, biased, and even poisoned.  ...  Even if the data cannot be fully cleaned, we can still cope with imperfect data during model training using robust model training techniques.  ...  There have been some approaches to successfully denoising the natural noise using Sparse Coding [145] and Feature Attention [22] , but they are out of the scope of this survey.  ... 
arXiv:2112.06409v2 fatcat:ztyeo7kalbgmpohynjudy5x5ei

Goal-Directed Resilience in Training (GRIT): A Biopsychosocial Model of Self-Regulation, Executive Functions, and Personal Growth (Eudaimonia) in Evocative Contexts of PTSD, Obesity, and Chronic Pain

Martha Kent, Crystal Rivers, Glenda Wrenn
2015 Behavioral Sciences  
Implications of the training for the prevention of maladaptation, including psychological distress and health declines, and for promoting healthy development are addressed.  ...  This paper presents a biopsychosocial model of self-regulation, executive functions, and personal growth that we have applied to Goal-Directed Resilience in Training (GRIT) interventions for posttraumatic  ...  Comparable to the New Wave approaches, we set out to train missing skills.  ... 
doi:10.3390/bs5020264 pmid:26039013 pmcid:PMC4493448 fatcat:aeoh2ocmg5cvzbiycub2p5x5su

Cognitive Relevance Transform for Population Re-Targeting

Gregor Koporec, Andrej Košir, Aleš Leonardis, Janez Perš
2020 Sensors  
The paper presents a complete methodology for either adapting the output of a pre-trained, state-of-the-art object classification algorithm to the target population or inferring a proper, user-friendly  ...  The machine is useful as much as the output classification labels are correct and match the dataset-provided labels.  ...  In the context of a CNN as a human assistant, the question is which pre-trained object classification algorithm better imitate target-users?  ... 
doi:10.3390/s20174668 pmid:32825013 pmcid:PMC7506759 fatcat:b5b6ozon6nf7hiigijyaf3unmq

A Survey on Trustworthy Recommender Systems [article]

Yingqiang Ge, Shuchang Liu, Zuohui Fu, Juntao Tan, Zelong Li, Shuyuan Xu, Yunqi Li, Yikun Xian, Yongfeng Zhang
2022 arXiv   pre-print
Recommender systems (RS), serving at the forefront of Human-centered AI, are widely deployed in almost every corner of the web and facilitate the human decision-making process.  ...  , robustness in recommendation, user controllable recommendation, as well as the relationship between these different perspectives in terms of trustworthy and responsible recommendation.  ...  being detected by the model [13, 29] . • Attack Intent.  ... 
arXiv:2207.12515v1 fatcat:lsnuwdtl5rboznmhhux2n5y5om

A Review on Scene Prediction for Automated Driving

Anne Stockem Stockem Novo, Martin Krüger, Marco Stolpe, Torsten Bertram
2022 Physics  
The subtile indications of other drivers' intentions, which are often intuitively clear to the human driver, require data-driven models such as deep learning techniques.  ...  A quantitative comparison of the model results reveals the dominance of deep learning methods in current state-of-the-art research in this area, leading to a competition on the cm scale.  ...  The purpose of the training data in the context of learning-based models is obvious. After training (and validation) a separate test dataset is needed to evaluate the trained models.  ... 
doi:10.3390/physics4010011 fatcat:4rbh64feprdnrflcf4ydyyre4a

Detection of Irradiated Food and Evaluation of the Given Dose by Electron Spin Resonance, Thermoluminescence, and Gas Chromatographic/Mass Spectrometric Analysis [chapter]

Maria C. D'Oca, Antonio Bartolotta
2018 Food Control and Biosecurity  
Acknowledgments This work was done in the framework of state project no. 0035-2014-0008 (Institute of Applied Physics, Russian Academy of Sciences). Acknowledgments  ...  Acknowledgments This work was supported by national funds through the Pluriannual Program (PEst-OE/AGR/U10276/2014), financed by FCT-Fundação para a Ciência e a Tecnologia (FCT).  ...  out or added the ingredient.  ... 
doi:10.1016/b978-0-12-811445-2.00010-6 fatcat:flc2mqpiwrhn3cmrmnbkgwgfxe
« Previous Showing results 1 — 15 out of 3,482 results