Personas with Attitudes: Controlling LLMs for Diverse Data Annotation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Fröhling, Leon, Demartini, Gianluca, Assenmacher, Dennis
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917804045762560
author Fröhling, Leon
Demartini, Gianluca
Assenmacher, Dennis
author_facet Fröhling, Leon
Demartini, Gianluca
Assenmacher, Dennis
contents We present a novel approach for enhancing diversity and control in data annotation tasks by personalizing large language models (LLMs). We investigate the impact of injecting diverse persona descriptions into LLM prompts across two studies, exploring whether personas increase annotation diversity and whether the impacts of individual personas on the resulting annotations are consistent and controllable. Our results show that persona-prompted LLMs produce more diverse annotations than LLMs prompted without personas and that these effects are both controllable and repeatable, making our approach a suitable tool for improving data annotation in subjective NLP tasks like toxicity detection.
format Preprint
id arxiv_https___arxiv_org_abs_2410_11745
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Personas with Attitudes: Controlling LLMs for Diverse Data Annotation
Fröhling, Leon
Demartini, Gianluca
Assenmacher, Dennis
Computation and Language
Human-Computer Interaction
We present a novel approach for enhancing diversity and control in data annotation tasks by personalizing large language models (LLMs). We investigate the impact of injecting diverse persona descriptions into LLM prompts across two studies, exploring whether personas increase annotation diversity and whether the impacts of individual personas on the resulting annotations are consistent and controllable. Our results show that persona-prompted LLMs produce more diverse annotations than LLMs prompted without personas and that these effects are both controllable and repeatable, making our approach a suitable tool for improving data annotation in subjective NLP tasks like toxicity detection.
title Personas with Attitudes: Controlling LLMs for Diverse Data Annotation
topic Computation and Language
Human-Computer Interaction
url https://arxiv.org/abs/2410.11745