AMS_CREATIVITY AND TRANSLATION IN THE AGE OF ARTIFICIAL INTELLIGENCE

International Conference: Creativity and Translation in the Age of Artificial Intelligence January 11–13, 2024

Machine Translation in Restaurant Menus: A Comparative Study with ChatGPT, DeepL and Google Translate

ALBA MONTES SÁNCHEZ Universidad de Córdoba (Spain) l32mosaa@uco.es

INDEX

1. Introduction

2. A brief history of machine translation

3. Methodology3.1. Baselines 3.2. Data

4. Results and discussion

5. Conclusions

6. References

" The art of creating machines that perform functions that require intelligence when performed by people.''

Kurzweil (1990, p. 17)

"Contrary to viewing AI as a threat to human professional development, it is understood as an organizer of new tasks within the profession"

Kirov (2022)

AI & Translations Studies

Human linguistic complexity cannot be fully assimilated by technology a threat to the skills and requirements traditionally expected of translators. Wilks, 1972, 1978; Parton, 2006; Tomasello, 2019, Vargas-Sierra, 2020, Hasyim et al., 2021,, or Canavilhas, 2022.

AI & Translations Studies

"Co-dependency" (Luo, 2018) : introducing quantitative thinking of machines and a more comprehensive incorporation of data, --> novel scenarios for translation. Le Scao, a specialist in language technologies motivated by the desire to order a kebab in any language, applied mathematics to NLP. It is precisely in the advancement of research combining AI and translation that scientific studies in the field are currently focusing.

Objectives

How the process of machine translation from one source language to another may modify the meaning in restaurant menus: "cuisine and restaurants are powerful tools for cultural, social and tourist image-building, and projection for tourist promotion" (Fuentes, 2016 P. 177).

Questions related to machine translation and AI are made, as well as the analytical data in Spanish gastronomy texts (menus) retrieved from the Andalusian cuisine.

We provide a preliminary study of ChatGPT on machine translation to gain a better understanding of this specialized translation driven by technologies.

Timeline of Machine Translation

2016 Google revamped its system: neural machine translation (NMT)

1949Warren Weaver translation memorandum

1964Automatic Language Processing Advisory Committee (ALPAC)

Golden 1950sIBM and Georgetown University

2006Google Translate 2007 MOSES, an open-source statistical system

Baselines

The methodology of the study focuses on a descriptive and qualitative analysis aimed at contrasting the accuracy of machine translation in systems like Google Translate or DeepL with the output generated by ChatGPT.

Restaurant Menu 1

Restaurant Menu 3

Restaurant Menu 4

Info

Restaurant Menu 2

Baselines

How we provide information in the source language to guide machine translation systems towards the target language, as the quality of the target text depends on the instructions or prompts given. Following Jiao et al. (2023), we directly engage the system to design instructions that will enhance ChatGPT's machine translation capabilities as accurately as possible.

PROMPTS

(P1) 'Translate this sentence into English'(P2) 'Please translate this sentence into English' (P3) 'Translate this restaurant menu into English'

Data of the study

Cohesive sample

Qualitative study

Representative sample

The corpus is composed by gastronomic menus from Andalusian restaurants with typical dishes from the southern region of Spain.

The selected restaurants and their geographical origins provides a balanced and cohesive sample of study

The analysis is divided into three categories: starters [ST], main courses [MC], and desserts [DD].

40

8 provinces

Restaurant menus

Main categories

Data

Main courses [MC]

Desserts [DD]

Starters [ST]

• [MC1] Huevas fritas o plancha • [MC2] Centros de alcauciles con habitas • [MC3] Rosada en adobo • [MC4] Flamenquín • [MC5] Puchero con habichuelas blancas y pringá

• [ST1] Tomate aliñado al estilo sevillano• [ST2] Cogollos de lechuga al ajillo• [ST3] Ensaladilla rusa• [ST4] Mazamorra con sardina ahumada, uvas y pasas• [ST5] Ajoblanco

• [DD1] Tocino de cielo casero • [DD2] Mousse de queso fresco de cabra con dulce de membrillo • [DD3] Gachas dulces • [DD4] Leche frita • [DD5] Torrijas

Translation Techniques by Molina and Hurtado (2002)

• Generalization vs. particularization
• Linguistic amplification vs. linguistic compression
• Literal translation
• Modulation
• Reduction
• Substitution
• Transposition
• Variation

• Adaptation
• Amplification
• Borrowing
• Calque
• Compensation
• Description
• Discursive creation
• Established equivalent

First category: Starters [ST]

Second category: Main Course [MC]

Third category: Desserts [DD]

Conclusions

Paratextual elements: The textual and visual components found in menus, and being able to be identified by IA, have the ability to evoke mental images, conjure up tastes and aromas, bring to mind familiar or novel textures, and elicit memories.

Technologies can perform 'unconscious' tasks, actions that humans perform almost involuntarily, and consequently, this becomes a translation element.

The comparative study with three machine translation systems reveal the ongoing difficulty of translating culturally-bound terms: semantic, grammatical, or lexical errors impact the intelligibility of the translated text.

The fear that human capacity in translation tasks might be replaced should make way for collaboration and the pursuit of symbiosis between the realms of technology and artificial intelligence, particularly with human translators.

References

• Canavilhas, J. (2022). Inteligencia artificial aplicada al periodismo: traducción automática y recomendación de contenidos en el proyecto “A European Perspective” (UER). Revista latina de comunicación social, (80), 24.
• Fuentes-Luque, A. (2017). An approach to analysing the quality of menu translations in southern Spain restaurants. Journal of multilingual and multicultural development, 38(2), 177-188.
• Jiao, W., Wang, W., Huang, J. T., Wang, X., & Tu, Z. (2023). Is ChatGPT a good translator? A preliminary study. arXiv preprint arXiv:2301.08745.
• Luo, X. (2018). Artificial intelligence and the crisis of translation. Asia Pacific Translation and Intercultural Studies, 5(1), 1-2.
• Molina, L., & Hurtado Albir, A. (2002). Translation techniques revisited: A dynamic and functionalist approach. Meta, 47(4), 498-512.
• Taylor, J. E. T., & Taylor, G. W. (2021). Artificial cognition: How experimental psychology can help generate explainable artificial intelligence. Psychonomic Bulletin & Review, 28(2), 454-475.
• Tomasello, L. (2019). Neural machine translation and artificial intelligence: what is left for the human translator? Università degli Studi di Padova.

• Weaver, W. (1952). Translation. In Proceedings of the Conference on Mechanical Translation.
• Wilks, Y. (1978). Machine translation and artificial intelligence Implementing machine aids to translation. In Translating and the Computer.

International Conference: Creativity and Translation in the Age of Artificial Intelligence January 11–13, 2024

Machine Translation in Restaurant Menus: A Comparative Study with ChatGPT, DeepL and Google Translate

ALBA MONTES SÁNCHEZ Universidad de Córdoba (Spain) l32mosaa@uco.es

¿Tienes una idea?

¡Que fluya la comunicación!

Con las plantillas de Genially podrás incluir recursos visuales para dejar a tu audiencia con la boca abierta. También destacar alguna frase o dato concreto que se quede grabado a fuego en la memoria de tu público e incluso embeber contenido externo que sorprenda: vídeos, fotos, audios... ¡Lo que tú quieras! ¿Necesitas más motivos para crear contenidos dinámicos? Bien: el 90% de la información que asimilamos nos llega a través de la vista y, además, retenemos un 42% más de información cuando el contenido se mueve.

• Genera experiencias con tu contenido.
• Tiene efecto WOW. Muy WOW.
• Logra que tu público recuerde el mensaje.
• Activa y sorprende a tu audiencia.