Performance Impact Caused by Hidden Bias of Training Data for Recognizing Textual Entailment

By Masatoshi Tsuchiya. 2018

168 citations recorded by the source corpus.

The quality of training data is one of the crucial problems when a learning-centered approach is employed. This paper proposes a new method to investigate the quality of a large corpus designed for the recognizing textual entailment (RTE) task. The proposed method, which is inspired by a statistical hypothesis test, consists of two phases: the first phase is to introduce the predictability of textual entailment labels as a null hypothesis which is extremely unacceptable if a target corpus has no hidden bias, and the second phase is to test the null hypothesis using a Naive Bayes model. The experimental result of the Stanford Natural Language Inference (SNLI) corpus does not reject the null hypothesis. Therefore, it indicates that the SNLI corpus has a hidden bias which allows prediction of textual entailment labels from hypothesis sentences even if no context information is given by a premise sentence. This paper also presents the performance impact of NN models for RTE caused by this hidden bias.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding (2018, background): “In light of recent work showing that crowdsourced data often contains artifacts which can be exploited to perform well without solving the intended task (Schwartz et al., 2017; Gururangan et al., 2018; Poliak et al., 2018; Tsuchiya, 2018), we audit the data for such artifacts.”
Adversarial NLI: A New Benchmark for Natural Language Understanding (2019, background): “models learn to exploit spurious statistical patterns in datasets (Gururangan et al., 2018; Poliak et al., 2018; Tsuchiya, 2018; Glockner et al., 2018; Geva et al., 2019; McCoy et al., 2019), instead of learning meaning in the flexible and generalizable way”
XNLI: Evaluating Cross-lingual Sentence Representations (2018, methodology): “As in the original MultiNLI corpus, we expect that cues like these (‘artifacts’, in Gururangan’s terms, also observed by Poliak et al., 2018; Tsuchiya, 2018) allow a baseline system to achieve better-than-random accuracy with access only to the premise sentences.”
WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation (2022, neutral): “Despite being one of the most resource-available tasks in NLP, analysis and challenge sets repeatedly demonstrate the limitations of existing datasets and the brittleness of NLI models trained on them (Gururangan et al., 2018; Poliak et al., 2018; Tsuchiya, 2018).”
Symbolic Knowledge Distillation: from General Language Models to Commonsense Models (2021, methodology): “Data Generation Although manual creation of datasets is expensive and introduces annotation artifacts (Schwartz et al., 2017; Gururangan et al., 2018; Agrawal et al., 2018; Tsuchiya, 2018; Geva et al., 2019; Bras et al., 2020), crowdsourcing remains the most popular method for curating goaloriented ”
Hypothesis Only Baselines in Natural Language Inference (2018, neutral): “3 Con-currently, Tsuchiya (2018) and Gururangan et al. (2018) similarly trained an NLI classiﬁer with access limited to hypotheses and discovered similar results on three of the ten datasets that we study.”
Dynabench: Rethinking Benchmarking in NLP (2021, background): “Natural language inference models can be fooled by simple syntactic heuristics or hypothesis-only biases (Gururangan et al., 2018; Poliak et al., 2018; Tsuchiya, 2018; Belinkov et al., 2019; McCoy et al., 2019).”
Are We Modeling the Task or the Annotator? An Investigation of Annotator Bias in Natural Language Understanding Datasets (2019, neutral): “In this paper, we continue recent efforts to understand biases that are introduced during the process of data creation (Levy et al., 2015; Schwartz et al., 2017; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Tsuchiya, 2018; Aharoni and Goldberg, 2018; Paun et al., 2018).”
Adversarial Filters of Dataset Biases (2020, background): “Such biases or artifacts 1 are often introduced during data collection (Fouhey et al., 2018) or during human annotation (Gururangan et al., 2018; Poliak et al., 2018; Tsuchiya, 2018; Geva et al., 2019).”
Abductive Commonsense Reasoning (2019, methodology): “Baselines We include baselines that rely on simple features to verify that ART is not trivially solv-able due to noticeable annotation artifacts, observed on several crowdsourced datasets Gururangan et al. (2018); Poliak et al. (2018); Tsuchiya (2018).”
What Will it Take to Fix Benchmarking in Natural Language Understanding? (2021, background): “Existing datasets focus heavily on repetitive, easy cases and often fail to isolate key behaviors (Jia and Liang, 2017; Tsuchiya, 2018; McCoy et al., 2019).”
Explaining Question Answering Models through Text Generation (2020, background): “In fact, it has been repeatedly shown that often models obtain high performance through “shortcuts” rather than language understanding (Tsuchiya, 2018; Poliak et al., 2018; Gu-rurangan et al., 2018; Geva et al., 2019).”
Analyzing Dynamic Adversarial Training Data in the Limit (2021, background): “only information (Poliak et al., 2018; Gururangan et al., 2018; Tsuchiya, 2018) and high-overlap entailment examples (McCoy et al.”
Get Your Vitamin C! Robust Fact Verification with Contrastive Evidence (2021, methodology): “Sentence-pair inference tasks such as fact veriﬁcation (Paul Panenghat et al., 2020; Schuster et al., 2019) and NLI (Gururangan et al., 2018; McCoy et al., 2019; Poliak et al., 2018; Tsuchiya, 2018) are no exception.”
Diversify Your Datasets: Analyzing Generalization via Controlled Variance in Adversarial Datasets (2019, background): “test sets just by exploiting artifacts in the generation of the hypotheses, rather than learning to model the complex entailment relationship (Gururangan et al., 2018; Tsuchiya, 2018; Poliak et al., 2018).”
DynaSent: A Dynamic Benchmark for Sentiment Analysis (2020, background): “This narrowness seems to be an underlying cause of many of the artifacts that have been identified in prominent NLU benchmarks (Poliak et al., 2018; Gururangan et al., 2018; Tsuchiya, 2018; Belinkov et al., 2019).”
The Curse of Performance Instability in Analysis Datasets: Consequences, Source, and Suggestions (2020, background): “In NLI, hypothesis bias was also found in SNLI and MultiNLI (Tsuchiya, 2018; Gururangan et al., 2018).”
DocNLI: A Large-scale Dataset for Document-level Natural Language Inference (2021, background): “In consideration of the common artifacts in some popular sentence-level NLI benchmarks (Gururangan et al., 2018; Poliak et al., 2018; Tsuchiya, 2018), we tried a “hypothesis-only” base-line based on RoBERTa on this raw-D OC NLI.”
Teach Me to Explain: A Review of Datasets for Explainable Natural Language Processing (2021, background): “Lack of variety can result in annotation artifacts, which are prevalent in the popular SNLI [16] and MNLI [129] datasets [97, 45, 120], among others [40].”