Information-Theoretic Method for Assessing the Quality of Translations

Area of Science:

• Computational Linguistics
• Literary Studies
• Information Theory

Background:

• Machine translation and literary translation are vital practical applications attracting significant research interest.
• Assessing translation quality requires methods sensitive to the nuances of authorial style, which can be influenced by the translator.
• Existing methods may not adequately capture the preservation of the original author's unique style in translated works.

Purpose of the Study:

• To propose and validate an information-theoretic method for evaluating the quality of translations.
• To specifically assess how accurately an author's unique style is conveyed across different translators.
• To apply this method to cross-lingual literary translations and text attribution tasks.

Main Methods:

• Utilizing an information-theoretic approach based on A.N. Kolmogorov's classification of text variability (information content, form, author's style).
• Developing a novel method to quantify the preservation of unconscious author's style in translated texts.
• Applying the method to analyze translations between English and Russian classical literature.

Main Results:

• The proposed method effectively estimates translation quality by focusing on stylistic elements.
• The study successfully differentiated translations based on the translator's influence on authorial style.
• The method proved capable of determining the attribution of literary texts, indicating its robustness.

Conclusions:

• An information-theoretic framework offers a powerful tool for assessing translation quality, particularly concerning authorial style.
• The developed method provides a quantitative means to evaluate translator performance and stylistic fidelity.
• This approach has significant implications for literary analysis, translation studies, and text authentication.