MétaCan
Menu
Back to cohort
Record W4401813233 · doi:10.1016/j.nlp.2024.100098

HarmonyNet: Navigating hate speech detection

2024· article· en· W4401813233 on OpenAlex

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.
fundA Canadian funder is recorded on the work.

Bibliographic record

VenueNatural Language Processing Journal · 2024
Typearticle
Languageen
FieldComputer Science
TopicHate Speech and Cyberbullying Detection
Canadian institutionsVector Institute
FundersVector InstituteGovernment of OntarioCanadian Institute for Advanced Research
KeywordsVoice activity detectionComputer scienceSpeech recognitionSpeech processing

Abstract

fetched live from OpenAlex

In the digital era, social media platforms have become central to communication across various domains. However, the vast spread of unregulated content often leads to the prevalence of hate speech and toxicity. Existing methods to detect this toxicity struggle with context sensitivity, accommodating diverse dialects, and adapting to varied communication styles. To tackle these challenges, we introduce an ensemble classifier that leverages the strengths of language models and traditional deep neural network architectures for more effective hate speech detection on social media. Our evaluations show that this hybrid approach outperforms individual models and exhibits robustness against adversarial attacks. Future efforts will aim to enhance the model’s architecture to further boost its efficiency and extend its capability to recognize hate speech across an even wider range of languages and dialects. • H armonyNet is a ensemble model that can detect hate speech. • H armonyNet is robust and can handle perturbations in texts. • H armonyNet shows better performance compared to individual models.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.001
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesScholarly communication
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Other design · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.930
Threshold uncertainty score0.999

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0010.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.001
Science and technology studies0.0010.000
Scholarly communication0.0020.002
Open science0.0000.000
Research integrity0.0000.001
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.006
GPT teacher head0.266
Teacher spread0.260 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it