Filters








79 Hits in 3.7 sec

Improving Robustness by Augmenting Training Sentences with Predicate-Argument Structures [article]

Nafise Sadat Moosavi, Marcel de Boer, Prasetya Ajie Utama, Iryna Gurevych
2020 arXiv   pre-print
In addition, we show that models can still be vulnerable to the lexical overlap bias, even when the training data does not contain this bias, and that the sentence augmentation also improves the robustness  ...  We will release our adversarial datasets to evaluate bias in such a scenario as well as our augmentation scripts at https://github.com/UKPLab/data-augmentation-for-robustness.  ...  This work is funded by the German Research Foundation through the research training group AIPHES (GRK 1994/1) and by the German Federal Ministry of Education and Research and the Hessian Ministry of Higher  ... 
arXiv:2010.12510v1 fatcat:hl5hq5qywzaytffz475uqscfui

Analyzing machine-learned representations: A natural language case study [article]

Ishita Dasgupta, Demi Guo, Samuel J. Gershman, Noah D. Goodman
2019 arXiv   pre-print
We then investigate the effect of the training distribution on learning these heuristic strategies, and study changes in these representations with various augmentations to the training set.  ...  Analyzing performance on these diagnostic tests indicates a lack of systematicity in the representations and decision rules, and reveals a set of heuristic strategies.  ...  NDG is supported by DARPA agreement number FA8750-14-2-0009, and a Sloan Foundation Research Fellowship.  ... 
arXiv:1909.05885v1 fatcat:wo7z5woybfdvhe732gmc452soy

Towards Reversal-Based Textual Data Augmentation for NLI Problems with Opposable Classes

Alexey Tarasov
2020 Proceedings of the First Workshop on Natural Language Interfaces   unpublished
Data augmentation methods are commonly used in computer vision and speech. However, in domains dealing with textual data, such techniques are not that common.  ...  We propose and evaluate two reversal-based methods on an NLI task of recognising a type of a simple logical expression from its description in plain-text form.  ...  Special thanks go to Awez Syed who sparkled author's interest in data augmentation.  ... 
doi:10.18653/v1/2020.nli-1.2 fatcat:fc3w2ckebbdr3m5ptf5cm5jtzm

Ten paradigms of ancillary antonymy: Evidence from Classical Arabic

Hamada S. A. Hassanein
2020 مجلة کلیة الآداب . القاهرة  
Results indicate that canonical, less canonical, and noncanonical pairs of opposition are interchangeably employed to signal, sharpen, and trigger each other on the syntagmatic axis.  ...  across a variety of languages: Swedish, Japanese, Dutch, Serbian, Qur'anic Arabic, Chinese, Modern Standard Arabic, and Classical Arabic.  ...  ones (Bpairs) and augmenting their contrastive power.  ... 
doi:10.21608/jarts.2020.135396 fatcat:f3hwd54zbrhcbhlm3tjf2bfv4e

An Empirical Study on Model-agnostic Debiasing Strategies for Robust Natural Language Inference [article]

Tianyu Liu, Xin Zheng, Xiaoan Ding, Baobao Chang, Zhifang Sui
2020 arXiv   pre-print
We also perform data augmentation including text swap, word substitution and paraphrase and prove its efficiency in combating various (though not all) adversarial attacks at the same time.  ...  Finally, we investigate several methods to merge heterogeneous training data (1.35M) and perform model ensembling, which are straightforward but effective to strengthen NLI models.  ...  Acknowledgments We would like to thank Sam Wiseman and Kevin Gimpel for very thoughtful discussions, and the anonymous reviewers for their helpful feedback.  ... 
arXiv:2010.03777v2 fatcat:m4tts7rvn5etzpa4sulmg45zza

TopX : efficient and versatile top-k query processing for text, structured, and semistructured data [article]

Martin Theobald, Universität Des Saarlandes, Universität Des Saarlandes
2006
text data.  ...  Chapter 2 defines our data model and the query language for text and semistructured data.  ...  We claim that (A) a packing for this instance of KNAPSACK has capacity ≤ C and utility ≥ U if and only if (B) the corresponding SAS instance has a scan of total depth C and score decrease of ≥ U.  ... 
doi:10.22028/d291-23776 fatcat:wsdfpjnpsjgbjn4x437dn53gdm

Recent Advances in Natural Language Inference: A Survey of Benchmarks, Resources, and Approaches [article]

Shane Storks, Qiaozi Gao, Joyce Y. Chai
2020 arXiv   pre-print
In the NLP community, recent years have seen a surge of research activities that address machines' ability to perform deep language understanding which goes beyond what is explicitly stated in text, rather  ...  relying on reasoning and knowledge of the world.  ...  Acknowledgements We would like to thank the anonymous reviewers for their greatly helpful comments and suggestions.  ... 
arXiv:1904.01172v3 fatcat:minzpxrrwfebdipu55udwe5dxq

VALSE: A Task-Independent Benchmark for Vision and Language Models Centered on Linguistic Phenomena [article]

Letitia Parcalabescu, Michele Cafagna, Lilitta Muradjan, Anette Frank, Iacer Calixto, Albert Gatt
2021 arXiv   pre-print
We propose VALSE (Vision And Language Structured Evaluation), a novel benchmark designed for testing general-purpose pretrained vision and language (V&L) models for their visio-linguistic grounding capabilities  ...  We build VALSE using methods that support the construction of valid foils, and report results from evaluating five widely-used V&L models.  ...  interpret affirmation and negation correctly.  ... 
arXiv:2112.07566v1 fatcat:wl7gwm3hpjeq7bbhg7q5ats4o4

The Growth of Cognitive Structure in Monkeys and Men [chapter]

Brendan McGonigle, Margaret Chalmers
2002 Animal Cognition and Sequential Behavior  
As in the size conditions described above, Chalmers and McGonigle (1984) and McGonigle and Chalmers (1977, 1992) followed binary tests of transitivity with triadic tests derived permutatively from  ...  However, given that the subjects were human and were capable of covert if not explicit linguistically based augmentation of the test conditions, it was left to studies such as McGonigle and Jones (1978  ... 
doi:10.1007/978-1-4615-0821-2_12 fatcat:ykntruqd2bf25j3mhxzhdcfzp4

Knowledge-Based Textual Inference via Parse-Tree Transformations

Roy Bar-Haim, Ido Dagan, Jonathan Berant
2015 The Journal of Artificial Intelligence Research  
We also present a novel packed data-structure and a corresponding inference algorithm that allows efficient implementation of this formalism.  ...  We proved the correctness of the new algorithm and established its efficiency analytically and empirically.  ...  -2 Network of Excellence of the European Community FP7-ICT-2007-1-216886, the Israel Internet Association (ISOC-IL), grant 9022, and the FBK-irst/Bar-Ilan University collaboration.  ... 
doi:10.1613/jair.4584 fatcat:daozqescn5an3ofcm5ax6foqwm

Program Committee

2006 2006 Sixth IEEE International Workshop on Source Code Analysis and Manipulation  
The conference and workshops were partially supported by the NTU Centre for Liberal Arts and Social Sciences (CLASS) and the Singapore MOE TRF Grant Syntactic Well-Formedness Diagnosis and Error-Based  ...  The program combined the main conference with a special day on wordnets and word-embeddings and finished with a half day workshop on technology enhanced learning (TEL).  ...  and Kieraś, 2016) 6 .  ... 
doi:10.1109/scam.2006.23 dblp:conf/scam/X06c fatcat:2dhsf7loj5hlffu2jxpmlo2qcq

Linked open data to represent multilingual poetry collections. A proposal to solve interoperability issues between poetic repertoires

Elena González-Blanco, Gimena del Rio, Clara Martínez Cantón
2016 Zenodo  
Its final objective is to interconnect, reuse and locate data disseminated through poetic repertoi res, in order to boost int eroperability among them.  ...  The conceptual semantic model, written in OWL, includes classes and metadata from standard ontological models related to humanities fields (such as CIDOC or Dublin Core), and adds specific elements and  ...  and Irina Nevskaya and Anna Dybo for providing us with Starling data and guidance with respect to this.  ... 
doi:10.5281/zenodo.2551595 fatcat:4nbzl534ebgnbort742fhfoqam

An Empirical Study on Model-agnostic Debiasing Strategies for Robust Natural Language Inference

Tianyu Liu, Zheng Xin, Xiaoan Ding, Baobao Chang, Zhifang Sui
2020 Proceedings of the 24th Conference on Computational Natural Language Learning   unpublished
We also perform data augmentation including text swap, word substitution and paraphrase and prove its efficiency in combating various (though not all) adversarial attacks at the same time.  ...  Finally, we investigate several methods to merge heterogeneous training data (1.35M) and perform model ensembling, which are straightforward but effective to strengthen NLI models.  ...  Acknowledgments We would like to thank Sam Wiseman and Kevin Gimpel for very thoughtful discussions, and the anonymous reviewers for their helpful feedback.  ... 
doi:10.18653/v1/2020.conll-1.48 fatcat:z22nnmbdmfgrrhyphc6qxitmsy

Rhetorical figures as argument schemes – The proleptic suite

Ashley Rose Mehlenbacher
2017 Argument & Computation  
One such figure is prolepsis (πρóληψις ), which this essay charts out in some detail, addressing the challenges and the benefits of rendering such figures computationally tractable through the use of argument  ...  (like antithesis, oxymoron, and litotes, indexed by antonyms and forms of negation) are likely to offer a high return on our efforts.  ...  What is particularly interesting for the case of prolepsis is a subtler rhetorical act: negotiating incomplete data "as even biodiversity scientists admit, complete data is lacking" [58] .  ... 
doi:10.3233/aac-170028 fatcat:bc6r7wihdzdrvcthzda7u6ne6m

Free Choice in Deontic Inquisitive Semantics (DIS) [chapter]

Martin Aher
2012 Lecture Notes in Computer Science  
We will propose a novel solution to the free choice puzzle that is driven by empirical data from legal discourse and does not suffer from the same problems as implicature-based accounts.  ...  Possibilities and counter-possibilities.  ...  Network of Excellence (FP7-ICT-216886) and by the Fundació ICREA.  ... 
doi:10.1007/978-3-642-31482-7_3 fatcat:vqnczv27vrbh5bekucsnhlocxm
« Previous Showing results 1 — 15 out of 79 results