Tolmačenje z velikimi jezikovnimi modeli

posnemanje tolmaških strategij z iztočnico

Matija Šprogar Perko (Author), Vojko Gorjanc (Mentor)

Abstract

Raziskave o vplivu iztočnic na jezikovni prenos z velikimi jezikovnimi modeli se ukvarjajo izključno s prevajanjem oziroma zanemarjajo, da se prevajanje in tolmačenje razlikujeta. Tolmači zaradi narave svojega dela pogosteje in bolj izrazito kot prevajalci posegajo po parafraziranju, povzemanju in jedrnatem izražanju – postopkih, v katerih so veliki jezikovni modeli zelo dobri. V magistrskem delu se zato posvečam vprašanju, ali lahko s prilagojeno iztočnico za velike jezikovne modele, ki posnema tolmaške strategije, generiramo prevode, ki bi se po kakovosti kosali s človeškim tolmačenjem. Prilagojena iztočnica jezikovni model do rešitve usmerja prek treh zaporednih korakov (angl. chain-of-thought prompting): 1) parafraze besedila v jedrnatem, jasnem jeziku, 2) jezikovnega prenosa ter 3) preoblikovanja besedila z rabo pogostih kolokacij in poudarkom na koherentnosti. Iztočnica s tem posnema osnovno tolmaško tehniko deverbalizacije in ponovne ubeseditve. Preizkusil sem jo z jezikovnima modeloma Claude 3.5 Sonnet in GPT-4o na govorih iz institucij EU, ocenjeval pa sem prenos vsebine, terminološko ustreznost in besedišče, skladnjo in slog ter slovnično pravilnost. Analiza prevodov pokaže, da je imela prilagojena iztočnica na pomensko točnost le majhen vpliv, močno pa je izboljšala zveznost, skladnjo in naravno jezikovno rabo – do te mere, da so končni prevodi primerljivi s človeškim tolmačenjem. Gre za prvo delo, ki analizira vpliv posnemanja tolmaškega miselnega procesa na kakovost prevoda.

Keywords

veliki jezikovni modeli;iztočnice;tolmaške tehnike;miselni proces;razmišljanje po korakih;tolmačenje;magistrska dela;

Data

Language:	Slovenian
Year of publishing:	2025
Typology:	2.09 - Master's Thesis
Organization:	UL FF - Faculty of Arts
Publisher:	[M. Šprogar Perko]
UDC:	81'253:81'322.4(043.2)
COBISS:	234136579
Views:	72
Downloads:	10
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Prompt Engineering for Interpreting: Exploring Human-Like Interpreting Strategies with Large Language Models
Secondary abstract:	Research on prompt engineering for translation with large language models (LLMs) completely overlooks interpreting or fails to distinguish it from translation. Unlike translators, interpreters rely heavily on paraphrasing, summarizing, and concise expression—skills at which LLMs excel. This master's thesis explores whether LLMs can generate translations that match the quality of human interpretation when guided by an augmented prompt designed to imitate interpreting strategies. The method employs chain-of-thought prompting, guiding the model through a three-step process: 1) paraphrasing the text in clear and concise language, 2) transferring it into the target language, and 3) refining it using natural collocations while ensuring coherence. This approach mirrors the core interpreting technique of deverbalization and reformulation. The prompt was tested with Claude 3.5 Sonnet and GPT-4o on speeches from EU institutions. The translations were evaluated in terms of content accuracy, terminology and vocabulary usage, syntax, style and grammar. The results suggest that while the customized prompt had only a minor effect on semantic accuracy, it significantly improved cohesion, syntax, and idiomatic quality of the translations—to the extent that the final outputs were comparable to human interpretation. This study is the first to explore how imitating interpreters’ cognitive process influences translation quality.
Secondary keywords:	large language models;prompt engineering;interpreting techniques;human thought process;chain-of-thought prompting;interpreting;master's theses;
Type (COBISS):	Master's thesis/paper
Study programme:	0
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Filozofska fak., Oddelek za prevajalstvo
Pages:	105 str.
ID:	26259728