580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How do linguistic and pragmatic properties differentiate types of offensive words and their communicative functions?
This research theme focuses on dissecting the nuanced linguistic categories of offensive words—including insults, slurs, epithets, and expletives—and explores how they differentially encode speaker meaning, psychological states, and social attitudes. Understanding these distinct mechanisms is crucial for refining theoretical models of language aggression and pragmatics, as well as for improving computational detection systems that rely on linguistic cues.
Key finding: This paper advances a relevance-theoretic framework differentiating insults, epithets, slurs, and expressive expletives based on their encoding of conceptual and procedural meaning. It finds that some offensive words... Read more
Key finding: The COLD dataset introduces an annotation scheme capturing complex offensive language phenomena including implicitly offensive language and reclaimed slurs, demonstrating that automatic detection systems struggle particularly... Read more
Key finding: This study reveals that offensive discourse can be identified through stance analysis even in the absence of explicit offensive words, by modeling writer’s pros and cons and role framings using polarity lexicons and... Read more
Key finding: Examining the global spread and semantic adaptation of the English swear word 'fuck', this paper finds that the word has diffused widely across languages and regions with varying degrees of integration and taboo strength. It... Read more
Key finding: By comparing native (L1) and foreign (LX) English speakers, the study finds that non-native users both underestimate the offensiveness and use the highly offensive term 'cunt' less frequently. It establishes that pragmatic... Read more
2. What sociocultural and legal dynamics shape the variation, prohibition, and impact of offensive slurs and taboo words?
This theme examines how social contexts, group identity, historical background, and legal frameworks influence the offensive power, regulation, and acceptance of slurs and taboo words. Research here addresses cultural variability in perceived offensiveness, reclamation of slurs by in-group members, and differing legal approaches to hate speech and offensive language, emphasizing that offensive words are embedded in socio-legal systems and contested identity politics rather than fixed linguistic entities.
Key finding: The authors propose the Prohibitionism theory, arguing that slurs are offensive not primarily due to inherent content but because of social and group-based prohibitions determined by targeted communities. They document... Read more
Key finding: Focusing on the n-word’s historical and cultural trajectory, this paper argues for its continued prohibition outside the African American community due to its enduring capacity to inflict racial harm. It differentiates... Read more
Key finding: Through experimental evidence, this research shows that slurs are generally perceived as more hurtful than neutral group labels, but that some neutral nouns themselves can evoke significant hurt compared to adjectival forms.... Read more
Key finding: This comprehensive work analyzes the linguistic and legal challenges of hate speech, highlighting the absence of a universal definition due to differing jurisprudential traditions (common law vs civil law) and the complex... Read more
Key finding: This comparative study finds statistically significant differences in how Iranian EFL learners versus native English speakers perceive the severity of English taboo words. It concludes that offensiveness is culturally... Read more
3. What computational and methodological strategies improve the automatic detection and classification of offensive language across diverse linguistic resources?
This theme addresses the challenges and innovations in building effective automated offensive language detection systems. It encompasses the creation and refinement of annotated lexicons and corpora, the comparison between supervised and unsupervised learning methods, and algorithmic techniques to detect masked or implicit offensive content, especially in under-resourced languages or code-mixed social media contexts. The focus is on expanding detection beyond explicit keywords to nuanced, context-dependent offensive language.
Key finding: This study compares supervised and unsupervised machine learning approaches to offensive language detection on a novel YouTube comment corpus in Tamil, a low-resource Dravidian language. Results show that unsupervised... Read more
Key finding: The paper proposes an automated detection method capable of identifying masked offensive words on social media that evade traditional filters by letter substitutions or symbol insertions. Utilizing phonetic algorithms like... Read more
Key finding: This research develops a multilingual language identification system integrated with offensive term detection using machine learning and NLP techniques such as logistic regression, TF-IDF, and ensemble soft voting... Read more
Key finding: The authors present a curated and enriched lexicon for Modern Greek offensive words (HURTLEX-EL), employing a multidimensional cross-classification by context, reference, and thematic domain. This systematic approach reveals... Read more
Key finding: This study proposes an enriched offensive language taxonomy integrating explicit and implicit forms using corpus reviews and state-of-the-art word embeddings and transformer models. It provides a conceptual and computational... Read more