26 November 2024 in Digital Scholarship in the Humanities
This study utilized machine learning algorithms and entropy-based features to identify translators of two English translations of Hongloumeng, a great classical Chinese novel written in the mid-18th century. The translations under examination were completed, respectively, by David Hawkes and the Yangs (Yang Hsien-yi and Gladys Yang). Two feature sets were extracted as input for the identification of translator styles: wordform features (wordform unigrams, bigrams, and trigrams) and part-of-speech (POS) features (POS unigrams, bigrams, and trigrams). Additionally, four machine learning classifiers were tested: linear support vector machines (SVMs), linear discriminant analysis (LDA), random forest (RF), and multilayer perceptron (MLP). Analysis of feature importance and SHAP value identified the most influential features within each classifier. Results showed that LDA achieved the best performance, with 81 per cent accuracy in distinguishing between translations, showing promise for translator identification. In contrast, MLP struggled to reliably differentiate between translations, achieving only 50 per cent accuracy. Furthermore, POS features had the greatest influence in SVM and LDA, while wordform features dominated in RF. SHAP analysis revealed that Hawkes’ translation tended to exhibit higher POS unigram and lower POS trigram entropy compared to the Yangs’. This increased contribution of POS unigrams and trigrams suggests a link to explicitation differences in translation. In summary, the combination of machine learning and entropy-based stylometric features shows potential for automatic translator identification and analysis.
22 June 2026 in Across Languages and Cultures
Translated language often carries subtle linguistic markers that set it apart from text originally written in the target language. This study investigates the potential of powerful AI language models to automatically identify these differences, specifically focusing on distinguishing translated texts from original texts. More specifically, this study utilized transformer-based large language models to differentiate translational English and original English. FLOB (Freiburg–LOB Corpus of British English) was selected as the original English corpus, and COCE (Corpus of Chinese-English) was selected as the translational English corpus. Two models were tested: Bidirectional Encoder Representations from Transformers (BERT) and Robustly Optimized BERT Pretraining Approach (RoBERTa). The factors affecting classification results are investigated through SHAP analysis and analysis of text types that have significantly different error rates. Results show that both models achieved excellent performance, with F1-scores of .864 for BERT and .998 for RoBERTa. The text types miscellaneous, general fiction, skills trade, and hobbies, and humor exhibit significantly higher error rates. Reportage, review, and science exhibit significantly lower error rates. Through SHAP analysis, we find that higher error rates may be attributed to simpler sentences in these three text types and the shared characteristics of translational texts, such as the tendencies for simplification and explicitation. Conversely, lower error rates were associated with text types that did not share these characteristics. In summary, transformer-based large language models show great potential for the automatic classification and analysis of translational sentences.