We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Towards a Deep Multi-layered Dialectal Language Analysis: A Case Study of African-American English

Authors: Jamell Dacon
Abstract: Currently, natural language processing (NLP) models proliferate language discrimination leading to potentially harmful societal impacts as a result of biased outcomes. For example, part-of-speech taggers trained on Mainstream American English (MAE) produce non-interpretable results when applied to African American English (AAE) as a result of language features not seen during training. In this work, we incorporate a human-in-the-loop paradigm to gain a better understanding of AAE speakers' behavior and their language use, and highlight the need for dialectal language inclusivity so that native AAE speakers can extensively interact with NLP systems while reducing feelings of disenfranchisement.
Comments: Accepted to the NAACL 2022 (HCI+NLP) Workshop
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2206.08978 [cs.CL]
  (or arXiv:2206.08978v1 [cs.CL] for this version)

Submission history

From: Jamell Dacon [view email]
[v1] Fri, 3 Jun 2022 01:05:58 GMT (68kb,D)

Link back to: arXiv, form interface, contact.