Human and Machine: Language Processing in Translation Tasks
Identifikátory výsledku
Kód výsledku v IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F24%3A10492914" target="_blank" >RIV/00216208:11320/24:10492914 - isvavai.cz</a>
Výsledek na webu
<a href="https://aclanthology.org/2024.icnlsp-1.27/" target="_blank" >https://aclanthology.org/2024.icnlsp-1.27/</a>
DOI - Digital Object Identifier
—
Alternativní jazyky
Jazyk výsledku
angličtina
Název v původním jazyce
Human and Machine: Language Processing in Translation Tasks
Popis výsledku v původním jazyce
The present study analyzes the influence of linguistic factors (sentence ambiguities) and non-linguistic factors (visual cues) on online language processing in translation tasks. Moreover, it also offers an attempt at relating machine and human translation in a multimodal setting, an aspect that has received less attention before. We qualitatively evaluated translation outputs between subjects across different experimental conditions, as well as between human and machine translation processes. We observed a positive correlation between humans' reading time and models' next token prediction, with a higher similarity score for the translation of unambiguous sentences compared to translations of ambiguous sentences. We also found that a context-relevant image has a significant influence on translation updates
Název v anglickém jazyce
Human and Machine: Language Processing in Translation Tasks
Popis výsledku anglicky
The present study analyzes the influence of linguistic factors (sentence ambiguities) and non-linguistic factors (visual cues) on online language processing in translation tasks. Moreover, it also offers an attempt at relating machine and human translation in a multimodal setting, an aspect that has received less attention before. We qualitatively evaluated translation outputs between subjects across different experimental conditions, as well as between human and machine translation processes. We observed a positive correlation between humans' reading time and models' next token prediction, with a higher similarity score for the translation of unambiguous sentences compared to translations of ambiguous sentences. We also found that a context-relevant image has a significant influence on translation updates
Klasifikace
Druh
O - Ostatní výsledky
CEP obor
—
OECD FORD obor
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Návaznosti výsledku
Projekt
Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.
Návaznosti
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)
Ostatní
Rok uplatnění
2024
Kód důvěrnosti údajů
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů