Lauscher et al propose five pronoun rules for NLP models
TL;DR
- A 2022 COLING paper by Lauscher, Crowley and Hovy argues NLP's handling of 3rd person pronouns excludes non-binary users and neopronouns.
- The authors define five desiderata for modelling pronouns in language technology and validate them through a survey.
- They evaluate existing and novel modelling approaches against those desiderata and quantify the impact on an established benchmark dataset.
"The world of pronouns is changing," three NLP researchers write in a 2022 COLING paper that two of the specialists in our directory have shared: "from a closed word class with few members to an open set of terms to reflect identities." Current language technology, they argue, has not followed.
The paper, published in the ACL Anthology by Anne Lauscher, Archie Crowley and Dirk Hovy, focuses on 3rd person pronouns and what the authors call "various phenomena like neopronouns, i.e., novel pronoun sets that are not (yet) widely established." The abstract argues current modelling of these pronouns "contributes to the discrimination of marginalized and underrepresented groups, e.g., non-binary individuals" and runs against "one of the UN's sustainable development goals (goal 5)."
The contribution is a set of five desiderata for modelling pronouns, validated through a survey, then applied to existing and novel approaches. The authors say they "quantify the impact of a more discrimination-free approach on an established benchmark dataset."
The abstract does not list the five desiderata, name the benchmark, or publish the quantified impact.
Shared on Bluesky by 2 AI experts
-
#TBT #NLProc 'Welcome to the Modern World of Pronouns: Identity-Inclusive Natural Language Processing beyond Gender' by Anne Lauscher, Archie Crowley, Dirk Hovy (2022) discusses non-binary representation in tech and NLP…
View on Bluesky →
Originally reported by aclanthology.org
Read the original article →Original headline: Welcome to the Modern World of Pronouns: Identity-Inclusive Natural Language Processing beyond Gender