How AI Fails: An Interactive Pedagogical Tool for Demonstrating Dialectal Bias in Automated Toxicity Models

Subhojit Ghimire

REVIEW 3 major objections 4 minor 23 references

How AI Fails: An Interactive Pedagogical Tool for Demonstrating Dialectal Bias in Automated Toxicity Models

T0 review · 3 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A widely used toxicity classifier systematically scores African-American English text as 1.8 times more toxic and 8.8 times higher for identity hate than Standard American English, and an interactive tool shows how a human-set threshold tur

desk verdict Honest, reproducible audit with a nice pedagogical tool; the headline numbers are weaker than they look because the corpora are unmatched, but the qualitative bias conclusion is consistent with prior work. read the letter →

arxiv 2511.06676 v3 pith:37MSHIB7 submitted 2025-11-10 cs.CL cs.CYcs.HC

Subhojit Ghimire This is my paper · ORCID

classification cs.CLcs.CYcs.HC

keywords dialectalbiastoxicitydetectioncontentmoderationAfrican-AmericanEnglishalgorithmicfalsepositiverateAIliteracyinteractivepedagogy

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish two linked claims: a widely used toxicity classifier systematically scores African-American English (AAE) text higher than Standard American English (SAE) text, and the practical harm of that bias comes from the human-set threshold that turns a score into a moderation decision. On a benchmark of 10,000 AAE and 10,000 SAE tweets, the model averages 1.8 times higher toxicity scores and 8.8 times higher identity-hate scores for AAE text. The paper's interactive tool adds a user-controllable sensitivity slider, so a person can watch the same threshold flag one dialect while sparing another for equivalent content. The point matters because automated moderation is now widespread, and the tool gives non-experts a direct way to see how a seemingly neutral policy can produce a discriminatory outcome.

What carries the argument

The load-bearing mechanism is the Toxicity Classification Threshold slider in the interactive tool. It does not recompute or recalibrate the model's scores; it simply draws the binary line between TOXIC and NOT TOXIC. Because the same line is applied to AAE and SAE text alike, any systematic difference in the score distributions becomes a systematic difference in flagging rates. The quantitative side of the paper uses a large public social-media corpus split by a demographic model into high-confidence AAE and SAE sets, then passes those through the classifier to establish the score distributions the slider exposes.

What would settle it

Have human annotators label a random sample of the same AAE and SAE tweets for true toxicity, remove tweets that are genuinely toxic, then recompute the false-positive-by-threshold curves. If the AAE curve no longer sits above the SAE curve, the paper's strongest no-fair-threshold claim does not survive; the same annotation could also verify whether the dialect labels themselves are accurate.

Watch

Extended reading notes

Core claim

The paper's central quantitative finding is that a widely used BERT-based toxicity model exhibits systematic dialectal bias: over the benchmark corpus, AAE text receives a mean toxicity score of 0.279 versus 0.148 for SAE text, a 1.8 ratio, and an identity-hate score of 0.0458 versus 0.0052, an 8.8 ratio. Score distributions show the gap is not driven by a few outliers; the AAE curve is flatter and spread across the range while the SAE curve is compressed near zero. The paper then argues the raw score gap is only half the story. Its Dialectal Bias Analyser lets users move a classification threshold and see that a single, uniformly applied policy level converts the score gap into a flag-rate

Load-bearing premise

In the false-positive analysis (Section 5.3), the load-bearing premise is that all 20,000 benchmark tweets are benign, so the threshold chart is a true false-positive comparison—a premise the paper itself concedes may not hold.

Editorial extensions

If this is right

A fixed moderation threshold will always produce a higher flagging rate for AAE text, so moving the threshold cannot by itself remove the disparate impact.
In practice, AAE-speaking users are more likely to have benign posts removed or hidden by automated moderation systems that use such thresholds.
The interactive tool turns an abstract fairness statistic into a visible cause-and-effect demonstration, lowering the barrier for public understanding of AI bias.
Re-running the same benchmark on newer classifiers would reveal whether later models have mitigated the dialect bias or only shifted it.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A matched-pair experiment—feeding sentences that differ only in dialect—would test whether dialect features alone drive the score gap, rather than corpus-level differences in topic or sentiment.
The same threshold-slider mechanic could be extended to gender, religion, or other identity axes to see whether no fair threshold is a general property of score-based moderation.
Because the false-positive analysis assumes every tweet is benign, a human-annotated toxicity sample is the natural next step; until then the true false-positive gap between dialects remains unmeasured.
The tool's framing suggests a policy remedy that goes beyond the paper: platforms could audit per-dialect flagging rates and adjust systems or policies accordingly, since a single uniform threshold cannot be fair.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Dear colleague,

The thing to know: this is a candid, self-aware paper that builds a nice interactive tool and releases its code; the headline numbers (1.8x, 8.8x) are not as solid as the abstract suggests, because the AAE and SAE corpora are unmatched and the paper admits the FPR analysis assumes all tweets are benign. Still, the qualitative conclusion—toxic-bert scores AAE more toxic than SAE—is very likely correct, and the tool is a real contribution to AI literacy.

What's actually new: the threshold-slider mechanic, borrowed from Face Mis-ID, is applied to NLP dialect bias for the first time. The tool makes the point that a 'neutral' policy threshold turns a biased score into a disparate outcome. That is a good pedagogical insight. The paper also does the right thing by citing the existing literature and not pretending the bias is new. The limitations section is unusually honest: it flags the model-on-model dialect labels and the benign-corpus assumption.

Where the paper is soft: the 1.8x/8.8x ratios are computed on two 10k-tweet sets filtered only by dialect probability. No matching on topic, sentiment, length, or actual toxicity. So the ratio likely includes content differences, not just dialect. The FPR analysis in Figure 5 is explicitly based on assuming all 20k tweets are benign, which the paper admits is not true; therefore it is not a genuine FPR. There are also no confidence intervals or significance tests on the differences. The minimal-pair examples (e.g., 'She is at the library' vs 'She at the library') are much more convincing evidence of dialect bias, but they are anecdotal and not included as a systematic evaluation.

These are load-bearing soft spots only if you treat the paper as a rigorous measurement study. If you treat it as a pilot audit plus a pedagogical tool, it works. The paper's own framing is closer to the latter.

Who is this for? People working on AI fairness education, or anyone wanting a ready-made demonstration for a classroom. I would not cite the specific numbers in a research paper, but I would cite the tool and the honest methodology note.

It deserves a serious referee: it is reproducible (code on GitHub, tool on Hugging Face), and it takes a clear position with appropriate humility. I'd send it to review, with the expectation that the quantitative claims get reworded as score disparities rather than false-positive rates.

Best.

Referee Report

3 major / 4 minor

Summary. The paper presents a dual contribution: a quantitative audit of the unitary/toxic-bert toxicity classifier on 10,000 African-American English (AAE) and 10,000 Standard American English (SAE) tweets sampled from the TwitterAAE dataset, and an interactive web tool, the “Dialectal Bias Analyser,” whose central mechanism is a user-controlled toxicity threshold slider. The benchmark reports that AAE text receives average toxicity scores 1.8 times higher and identity-attack scores 8.8 times higher than SAE text, and describes a false-positive-rate analysis across threshold values. The pedagogical tool is intended to show that a seemingly neutral human-set threshold operationalizes discriminatory outcomes.

Significance. If the quantitative claims are properly qualified, the paper makes a useful contribution to AI-literacy and fairness auditing. The interactive threshold-slider mechanic is a clear and transferable pedagogical idea, building directly on prior work such as Face Mis-ID. The paper also ships reproducible code and data-processing notebooks, which is a genuine strength. However, the quantitative evidence as currently presented overstates what can be concluded from the experimental design: the mean-score comparison uses unmatched, model-labeled corpora, and the false-positive-rate analysis relies on an unverified “all benign” assumption. The paper is therefore promising but needs a substantial revision to align its claims with its evidence.

major comments (3)

[§5.1, Table 4] The headline ratios (1.8x toxicity, 8.8x identity hate) are computed on two corpora filtered only by TwitterAAE dialect probability. No matching or adjustment is made for topic, length, sentiment, or human-annotated toxicity, and no confidence intervals or significance tests are reported. The AAE and SAE samples could therefore differ in content that legitimately affects the model's outputs. As the paper itself notes in §7b, the dialect labels come from a model, so this is a model-on-model comparison. I recommend either (a) adding matched or minimal-pair analyses, (b) reporting uncertainty, or (c) reframing the claims as “the model scores these AAE and SAE samples differently” rather than “clear, systematic bias.” The current abstract and conclusion make the stronger claim without the needed support.
[§5.3, §7c] The False Positive Rate analysis assumes all 20,000 sampled tweets are benign. Section 7c explicitly concedes that “some tweets could genuinely be toxic.” Therefore Figure 5 cannot be described as a comparison of false positive rates or of “benign text incorrectly flagged as toxic.” The plotted quantity is a flag rate, not an error rate, since no ground-truth toxicity labels are available. This is a load-bearing issue because the FPR/threshold framing is central to the paper’s policy argument. Please rename the quantity, qualify the interpretation, or obtain a human-annotated toxicity sample for at least a subset of the data.
[§7b] The paper acknowledges that the TwitterAAE “ground truth” is itself the output of a model, making the analysis model-on-model. This is not fatal for a descriptive audit, but it limits the causal claim that the toxic-bert model has ‘dialectal bias’ independent of the labeling procedure. The text should consistently state that the benchmark measures differences across model-derived dialect groups, and that human-annotated dialect and toxicity labels are needed to separate dialectal bias from content differences. The current wording in the abstract and §6 overstates the strength of the evidence.

minor comments (4)

[§3.2] The model pipeline is described as “the transformers pipeline,” but the exact model revision, PyTorch/transformers version, and scoring configuration are not specified. Please include these details for reproducibility.
[§5.1, §8] The paper uses “identity hate” while the model label is “identity_attack.” Define this mapping, and avoid phrasing such as “8.8 times more likely to contain identity hate” in the conclusion, as the metric is an average score, not a prevalence or likelihood.
[Figures 3–5] The figures are described but not displayed in the submitted text. Ensure captions define the box-plot elements and the axes of the FPR/flag-rate plot, and label the red/blue lines consistently with the text.
[Throughout] There are several typographical and phrasing issues, e.g., “which I stumbled later on during this research,” “Dialectical Bias” vs. “Dialectal Bias,” and inconsistent capitalization of “Biaslyze.” A careful proofreading pass is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the benchmark is an observed comparison of two independently trained models, not a fitted parameter disguised as a prediction.

full rationale

The central quantitative claim (§5.1, Table 4) is a direct mean-score comparison of unitary/toxic-bert on two fixed corpora. No parameter is fitted to the outcome, and the 1.8x/8.8x ratios are arithmetic consequences of the stored scores, not quantities that were optimized or defined in terms of the conclusion. The dialect grouping comes from the TwitterAAE demographic model (Blodgett et al., 2016), and the toxicity scores come from a separately trained Jigsaw-fine-tuned BERT; neither model's output is used to construct the other, so the disparity is not equivalent to the inputs by construction. The paper's own §7b acknowledges this is a model-on-model analysis, which is a validity/confounding concern rather than circularity. Similarly, §5.3's FPR analysis assumes all 20,000 samples are benign; §7c concedes some tweets could genuinely be toxic. That assumption makes the label false positive unsupported, but the plotted curves are computed directly from the model scores and a threshold, so no result is being derived from itself. The pedagogical minimal pairs (§4.2) are illustrative examples, not part of the benchmark derivation. There are no load-bearing self-citations, no imported uniqueness theorems, and no ansatz smuggled in via citation. Accordingly, the derivation chain is self-contained in the circularity sense.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new mechanistic entities. Its central measurement depends on two hand-chosen data-processing parameters (confidence threshold, corpus size) and three domain assumptions about the validity of the TwitterAAE labels, the benign status of the sampled tweets, and the representativeness of toxic-bert. The FPR claim is the most sensitive to the benign-status assumption.

free parameters (2)

AAE/SAE confidence threshold = >0.80
Used to filter TwitterAAE into AAE and SAE corpora in §3.1; chosen by hand and not varied in a sensitivity analysis.
Corpus size = 10,000 per group
Arbitrary sample size; results may depend on this choice and no power analysis is given.

assumptions (3)

domain assumption All 20,000 sampled tweets are benign
Needed to interpret Figure 5 as a false positive rate; acknowledged as a limitation in §7c.
domain assumption TwitterAAE demographic probabilities are valid dialect ground truth
Used to label AAE vs SAE in §3.1; the author acknowledges this is a model-on-model analysis in §7b.
domain assumption toxic-bert is representative of deployed moderation systems
Used as the only model under test in §3.2; the author notes in §7a that newer models may behave differently.

how reviews work

0 comments

Cite this review

Pith. "Pith review of How AI Fails: An Interactive Pedagogical Tool for Demonstrating Dialectal Bias in Automated Toxicity Models." pith.science (2026). https://pith.science/paper/37MSHIB7

@misc{pith2026251106676,
  author       = {Pith},
  title        = {Pith review of: How AI Fails: An Interactive Pedagogical Tool for Demonstrating Dialectal Bias in Automated Toxicity Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/37MSHIB7}},
  note         = {Machine review of arXiv:2511.06676}
}

read the original abstract

Now that AI-driven moderation has become pervasive in everyday life, we often hear claims that "the AI is biased". While this is often said jokingly, the light-hearted remark reflects a deeper concern. How can we be certain that an online post flagged as "inappropriate" was not simply the victim of a biased algorithm? This paper investigates this problem using a dual approach. First, I conduct a quantitative benchmark of a widely used toxicity model (unitary/toxic-bert) to measure performance disparity between text in African-American English (AAE) and Standard American English (SAE). The benchmark reveals a clear, systematic bias: on average, the model scores AAE text as 1.8 times more toxic and 8.8 times higher for "identity hate". Second, I introduce an interactive pedagogical tool that makes these abstract biases tangible. The tool's core mechanic, a user-controlled "sensitivity threshold," demonstrates that the biased score itself is not the only harm; instead, the more-concerning harm is the human-set, seemingly neutral policy that ultimately operationalises discrimination. This work provides both statistical evidence of disparate impact and a public-facing tool designed to foster critical AI literacy.

Figures

Figures reproduced from arXiv: 2511.06676 by the authors.

**Figure 1.** Web UI for Dialectal Bias Analyser Upon clicking “Analyse Text,” the model’s scores (aforementioned six output labels) are shown, based on the evaluation of the sentence in the text area. 4.2 Threshold Slider: The “Aha!” moment The central pedagogical feature of the tool is the “Toxicity Classification Threshold” slider. This slider does not re-calculate the model’s scores; rather, it acts as the human-set policy l… view at source ↗

**Figure 2.** An example text displaying discriminatory result upon adjusting slider. This mechanic is the "Aha!" moment. A user can analyse a sentence like "That Christian guy is pointing a gun at that lady." and see it receives a low “toxicity” score (0.6806). They can then analyse "That Muslim guy is pointing a gun at that lady.” and seea higher score (0.8995). By setting the threshold slider to 0.70 (more lenient), they di… view at source ↗

**Figure 3.** Distribution of Toxicity Scores (AAE versus SAE) (Box Plot) [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 5.** Figure 5: False Positive Rate v ersus Classification Threshold 6. Discussion The results from the quantitative benchmark are unambiguous. They are not a series of random errors but a reproducible pattern. The model exhibits a clear, systematic bias against one group. Why does t…

**Figure 4.** Figure 4: Distribution of Toxicity Scores (AAE versus SAE) ( Histogram with Bell Curve) 5.3 False Positive Rate (FPR) Analysis The most critical finding is displayed in [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 1 canonical work pages

[1]

and Munot, Mousami V

Gongane, Vaishali U. and Munot, Mousami V. and Anuse, Alwin D. , title=. Social Network Analysis and Mining , year=. doi:10.1007/s13278-022-00951-3 , url=

work page doi:10.1007/s13278-022-00951-3
[2]

Big Data & Society , volume =

Jenna Burrell , title =. Big Data & Society , volume =. 2016 , doi =. https://doi.org/10.1177/2053951715622512 , abstract =

doi:10.1177/2053951715622512 2016
[3]

Algorithms of Oppression , series =

Safiya Umoja Noble , publisher =. Algorithms of Oppression , series =. 2018 , lastchecked =. doi:doi:10.18574/nyu/9781479833641.001.0001 , isbn =

arXiv 2018
[4]

Demographic Dialectal Variation in Social Media: A Case Study of A frican- A merican E nglish

Blodgett, Su Lin and Green, Lisa and O ' Connor, Brendan. Demographic Dialectal Variation in Social Media: A Case Study of A frican- A merican E nglish. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. 2016. doi:10.18653/v1/D16-1120

doi:10.18653/v1/d16-1120 2016
[5]

The Risk of Racial Bias in Hate Speech Detection

Sap, Maarten and Card, Dallas and Gabriel, Saadia and Choi, Yejin and Smith, Noah A. The Risk of Racial Bias in Hate Speech Detection. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019. doi:10.18653/v1/P19-1163

doi:10.18653/v1/p19-1163 2019
[6]

Investigating Label Bias and Representational Sources of Age-Related Disparities in Medical Segmentation , year=

Parikh, Aditya and Das, Sneha and Feragen, Aasa , booktitle=. Investigating Label Bias and Representational Sources of Age-Related Disparities in Medical Segmentation , year=
[7]

Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection

Vidgen, Bertie and Thrush, Tristan and Waseem, Zeerak and Kiela, Douwe. Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021. d...

doi:10.18653/v1/2021.acl-long.132 2021
[8]

The Language Interpretability Tool: Extensible, Interactive Visualizations and Analysis for NLP Models

Tenney, Ian and Wexler, James and Bastings, Jasmijn and Bolukbasi, Tolga and Coenen, Andy and Gehrmann, Sebastian and Jiang, Ellen and Pushkarna, Mahima and Radebaugh, Carey and Reif, Emily and Yuan, Ann. The Language Interpretability Tool: Extensible, Interactive Visualizations and Analysis for NLP Models. Proceedings of the 2020 Conference on Empirical ...

doi:10.18653/v1/2020.emnlp-demos.15 2020

Show all 23 references

[9]

Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency , pages =

Gomez, Juan Felipe and Machado, Caio and Paes, Lucas Monteiro and Calmon, Flavio , title =. Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency , pages =. 2024 , isbn =. doi:10.1145/3630106.3659036 , abstract =

2024
[10]

2025 , eprint=

FairPy: A Toolkit for Evaluation of Prediction Biases and their Mitigation in Large Language Models , author=. 2025 , eprint=

2025
[11]

Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems , articleno =

Kabir, Samia and Li, Lixiang and Zhang, Tianyi , title =. Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems , articleno =. 2024 , isbn =. doi:10.1145/3613904.3642111 , abstract =

2024
[12]

Raz, Daniella and Bintz, Corinne and Guetler, Vivian and Tam, Aaron and Katell, Michael and Dailey, Dharma and Herman, Bernease and Krafft, P. M. and Young, Meg , title =. Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society , pages =. 2021 , isbn =. doi:10.1...

2021
[13]

A Survey of Race, Racism, and Anti-Racism in NLP

Field, Anjalie and Blodgett, Su Lin and Waseem, Zeerak and Tsvetkov, Yulia. A Survey of Race, Racism, and Anti-Racism in NLP. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language...

2021 doi
[14]

2023 , eprint=

ToxBuster: In-game Chat Toxicity Buster with BERT , author=. 2023 , eprint=

2023
[15]

TwitterAAE Dataset , howpublished =
[16]

unitary/toxic-bert , howpublished =
[17]

bert-base-uncased , howpublished =
[18]

Jigsaw Toxic Comment Classification Challenge , howpublished =
[19]

Dialectal Bias Analyser Repository , howpublished =
[20]

Dialectal Bias Analyser: Hugging Face Space , howpublished =
[21]

biaslyze , howpublished =
[22]

How Normal Am I , howpublished =
[23]

evaluate-measurement/toxicity , howpublished =

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [1]

and Munot, Mousami V

Gongane, Vaishali U. and Munot, Mousami V. and Anuse, Alwin D. , title=. Social Network Analysis and Mining , year=. doi:10.1007/s13278-022-00951-3 , url=

work page doi:10.1007/s13278-022-00951-3

[2] [2]

Big Data & Society , volume =

Jenna Burrell , title =. Big Data & Society , volume =. 2016 , doi =. https://doi.org/10.1177/2053951715622512 , abstract =

doi:10.1177/2053951715622512 2016

[3] [3]

Algorithms of Oppression , series =

Safiya Umoja Noble , publisher =. Algorithms of Oppression , series =. 2018 , lastchecked =. doi:doi:10.18574/nyu/9781479833641.001.0001 , isbn =

arXiv 2018

[4] [4]

Demographic Dialectal Variation in Social Media: A Case Study of A frican- A merican E nglish

Blodgett, Su Lin and Green, Lisa and O ' Connor, Brendan. Demographic Dialectal Variation in Social Media: A Case Study of A frican- A merican E nglish. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. 2016. doi:10.18653/v1/D16-1120

doi:10.18653/v1/d16-1120 2016

[5] [5]

The Risk of Racial Bias in Hate Speech Detection

Sap, Maarten and Card, Dallas and Gabriel, Saadia and Choi, Yejin and Smith, Noah A. The Risk of Racial Bias in Hate Speech Detection. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019. doi:10.18653/v1/P19-1163

doi:10.18653/v1/p19-1163 2019

[6] [6]

Investigating Label Bias and Representational Sources of Age-Related Disparities in Medical Segmentation , year=

Parikh, Aditya and Das, Sneha and Feragen, Aasa , booktitle=. Investigating Label Bias and Representational Sources of Age-Related Disparities in Medical Segmentation , year=

[7] [7]

Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection

Vidgen, Bertie and Thrush, Tristan and Waseem, Zeerak and Kiela, Douwe. Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021. d...

doi:10.18653/v1/2021.acl-long.132 2021

[8] [8]

The Language Interpretability Tool: Extensible, Interactive Visualizations and Analysis for NLP Models

Tenney, Ian and Wexler, James and Bastings, Jasmijn and Bolukbasi, Tolga and Coenen, Andy and Gehrmann, Sebastian and Jiang, Ellen and Pushkarna, Mahima and Radebaugh, Carey and Reif, Emily and Yuan, Ann. The Language Interpretability Tool: Extensible, Interactive Visualizations and Analysis for NLP Models. Proceedings of the 2020 Conference on Empirical ...

doi:10.18653/v1/2020.emnlp-demos.15 2020