41 |
What Should/Do/Can LSTMs Learn When Parsing Auxiliary Verb Constructions?
|
|
|
|
In: Computational Linguistics, Vol 46, Iss 4, Pp 763-784 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
42 |
Efficient Computation of Expectations under Spanning Tree Distributions
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 675-690 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
43 |
Revisiting Multi-Domain Machine Translation
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 17-35 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
44 |
Interpretability Analysis for Named Entity Recognition to Understand System Predictions and How They Can Improve
|
|
|
|
In: Computational Linguistics, Vol 47, Iss 1, Pp 117-140 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
45 |
Semantic Data Set Construction from Human Clustering and Spatial Arrangement
|
|
|
|
In: Computational Linguistics, Vol 47, Iss 1, Pp 69-116 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
46 |
WikiAsp: A Dataset for Multi-domain Aspect-based Summarization
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 211-225 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
47 |
Latent Compositional Representations Improve Systematic Generalization in Grounded Question Answering
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 195-210 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
48 |
Aligning Faithful Interpretations with their Social Attribution
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 294-310 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
49 |
Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 1408-1424 (2021) (2021)
|
|
Abstract:
Abstract⚠ This paper contains prompts and model outputs that are offensive in nature.When trained on large, unfiltered crawls from the Internet, language models pick up and reproduce all kinds of undesirable biases that can be found in the data: They often generate racist, sexist, violent, or otherwise toxic language. As large models require millions of training examples to achieve good performance, it is difficult to completely prevent them from being exposed to such content. In this paper, we first demonstrate a surprising finding: Pretrained language models recognize, to a considerable degree, their undesirable biases and the toxicity of the content they produce. We refer to this capability as self-diagnosis. Based on this finding, we then propose a decoding algorithm that, given only a textual description of the undesired behavior, reduces the probability of a language model producing problematic text. We refer to this approach as self-debiasing. Self-debiasing does not rely on manually curated word lists, nor does it require any training data or changes to the model’s parameters. While we by no means eliminate the issue of language models generating biased text, we believe our approach to be an important step in this direction.1
|
|
Keyword:
Computational linguistics. Natural language processing; P98-98.5
|
|
URL: https://doaj.org/article/7865d581bc554481bb1d3d28fe5f98e4 https://doi.org/10.1162/tacl_a_00434
|
|
BASE
|
|
Hide details
|
|
50 |
Evaluating Document Coherence Modeling
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 621-640 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
51 |
Comparing Knowledge-Intensive and Data-Intensive Models for English Resource Semantic Parsing
|
|
|
|
In: Computational Linguistics, Vol 47, Iss 1, Pp 43-68 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
52 |
Model Compression for Domain Adaptation through Causal Effect Estimation
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 1355-1373 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
53 |
Planning with Learned Entity Prompts for Abstractive Summarization
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 1475-1492 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
54 |
Supertagging the Long Tail with Tree-Structured Decoding of Complex Categories
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 243-260 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
55 |
Lexically Aware Semi-Supervised Learning for OCR Post-Correction
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 1285-1302 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
56 |
Supervised and Unsupervised Neural Approaches to Text Readability
|
|
|
|
In: Computational Linguistics, Vol 47, Iss 1, Pp 141-179 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
57 |
A Graph-Based Framework for Structured Prediction Tasks in Sanskrit
|
|
|
|
In: Computational Linguistics, Vol 46, Iss 4, Pp 785-845 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
58 |
What Helps Transformers Recognize Conversational Structure? Importance of Context, Punctuation, and Labels in Dialog Act Recognition
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 1163-1179 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
59 |
Efficient Outside Computation
|
|
|
|
In: Computational Linguistics, Vol 46, Iss 4, Pp 745-762 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
60 |
There Once Was a Really Bad Poet, It Was Automated but You Didn’t Know It
|
|
|
|
In: Transactions of the Association for Computational Linguistics, Vol 9, Pp 605-620 (2021) (2021)
|
|
BASE
|
|
Show details
|
|
|
|