Communicating artificial neural networks develop efficient color-naming systems

Mostra el registre complet Registre parcial de l'ítem

• dc.contributor.author Chaabouni, Rahma
• dc.contributor.author Kharitonov, Eugene
• dc.contributor.author Dupoux, Emmanuel
• dc.contributor.author Baroni, Marco
• dc.date.accessioned 2021-04-08T08:19:29Z
• dc.date.available 2021-04-08T08:19:29Z
• dc.date.issued 2021
• dc.description.abstract Words categorize the semantic fields they refer to in ways that maximize communication accuracy while minimizing complexity. Focusing on the well-studied color domain, we show that artificial neural networks trained with deep-learning techniques to play a discrimination game develop communication systems whose distribution on the accuracy/complexity plane closely matches that of human languages. The observed variation among emergent color-naming systems is explained by different degrees of discriminative need, of the sort that might also characterize different human communities. Like human languages, emergent systems show a preference for relatively low-complexity solutions, even at the cost of imperfect communication. We demonstrate next that the nature of the emergent systems crucially depends on communication being discrete (as is human word usage). When continuous message passing is allowed, emergent systems become more complex and eventually less efficient. Our study suggests that efficient semantic categorization is a general property of discrete communication systems, not limited to human language. It suggests moreover that it is exactly the discrete nature of such systems that, acting as a bottleneck, pushes them toward low complexity and optimal efficiency.
• dc.description.sponsorship We thank Emmanuel Chemla, Thomas Brochhagen, Roger Levy, Louise McNally, Rachid Riad, Noga Zaslavsky, the PNAS reviewers, and, especially, Gemma Boleda and Diane Bouchacourt for feedback. Ted Gibson and Bevil Conway generously shared their data with us. This research was funding by Agence Nationale pour la Recherche Grants ANR17-EURE-0017 Frontcog, ANR-10-IDEX-0001-02 PSL∗, and ANR-19-P3IA-0001 PRAIRIE 3IA
• dc.format.mimetype application/pdf
• dc.identifier.citation Chaabouni R, Kharitonov E, Dupoux E, Baroni M. Communicating artificial neural networks develop efficient color-naming systems. Proc Natl Acad Sci. 2021;118(12):e2016569118 DOI: 10.1073/pnas.2016569118
• dc.identifier.doi http://dx.doi.org/10.1073/pnas.2016569118
• dc.identifier.issn 0027-8424
• dc.identifier.uri http://hdl.handle.net/10230/47044
• dc.language.iso eng
• dc.publisher National Academy of Sciences
• dc.relation.ispartof Proceedings of the National Academy of Science of the United States of America. 2021;118(12):e2016569118
• dc.rights This open access article is distributed under Creative Commons Attribution License 4.0 (CC BY) https://creativecommons.org/licenses/by/4.0/
• dc.rights.accessRights info:eu-repo/semantics/openAccess
• dc.rights.uri https://creativecommons.org/licenses/by/4.0/
• dc.subject.keyword Efficiency of human language
• dc.subject.keyword Language emergence in artificial neural networks
• dc.subject.keyword Color-naming systems
• dc.title Communicating artificial neural networks develop efficient color-naming systems
• dc.type info:eu-repo/semantics/article
• dc.type.version info:eu-repo/semantics/publishedVersion