Unravelling the stylistic nuances: a comparative multidimensional analysis of amateur and professional translations of Legends of the Condor Heroes

The paper investigates how amateur (crowdsourced/volunteer) translations differ stylistically from professional translations in the Chinese martial arts (wuxia) genre, focusing on Jin Yong’s Legends of the Condor Heroes (LCH). Motivated by the surge of online fan translations that have expanded global access to wuxia, the study examines linguistic variation across fictional speech and narration to reassess the role of amateur translators and their impact on target-culture reception. The research aims to: (1) identify the MDA dimensions where amateur and professional translations diverge in speech and narration; (2) pinpoint specific linguistic features driving these differences; and (3) explore factors (e.g., training, collaboration, editing) that may account for the stylistic patterns observed.

The study situates itself within research on wuxia translation and Jin Yong’s reception in English. Earlier work on professional translations addressed challenges in rendering martial arts techniques, domestication vs. foreignization strategies, translatability and identity discourse, and treatment of cultural terminology (e.g., Wong 1997; Lai 1997; Mok 2002; Shen 2007; Zhao 2009). Recent corpus-based studies (Wu & Li 2021, 2022) examined normalization, lexical richness, and acceptance, while Chen & Dai (2021) documented purposeful omissions in LCH’s English translation. Despite the proliferation of amateur translations via platforms (e.g., spcnet, wuxiasociety), direct comparisons between amateur and professional renditions remain rare. The literature highlights the need to examine stylistic and linguistic differences systematically, given the large readership of fan translations and their influence on publishing and reception.

Design: A corpus-based comparative study of amateur translation (AT) and published professional translation (PT) of LCH. The corpus includes 40 corresponding chapters per version. Dialog was programmatically extracted via quotation marks, yielding four subcorpora: AT-speech, AT-narration, PT-speech, PT-narration. Reporting clauses (e.g., “he said”) were removed from narration to keep dialog integrity. This produced 160 files. Measures: Biber’s Multidimensional Analysis (MDA; 67 features across six dimensions) was applied to each subcorpus. Annotation/tagging used the Multidimensional Analysis Tagger (MAT; Nini 2019) built on the Stanford tagger. Because dimension and feature scores were non-normally distributed, Mann–Whitney U tests assessed differences between AT and PT. Descriptives: Speech—AT types 8,606; tokens 229,243; TTR 3.75; STTR 38.94. PT types 8,432; tokens 178,050; TTR 4.74; STTR 40.56. Narration—AT types 12,550; tokens 482,841; TTR 2.60; STTR 39.89. PT types 15,232; tokens 465,280; TTR 3.28; STTR 44.06.

Fictional speech (per chapter n=40 per version): Significant differences in Dimensions 1, 2, 5, and 6 (Mann–Whitney U): D1 U=593, p<0.05; D2 U=402, p<0.01; D5 U=531, p<0.01; D6 U=67, p<0.01. Means (AT vs PT): D1 3.12 vs 3.98 (PT more involved/affective); D2 −1.20 vs −0.40 (AT more non-narrative); D5 −0.90 vs −1.37 (AT more abstract); D6 0.56 vs −1.01 (AT more on-line informational elaboration). Feature-level differences highlighted greater PT use of 1st/2nd person pronouns, that-deletion, private verbs, etc., while AT showed higher rates of amplifiers, adverbs, conditional subordination, and features contributing to abstractness and on-line elaboration (e.g., THAT clauses as verb complements). Fictional narration (per chapter n=40 per version): Significant differences in Dimensions 1, 2, 5, and 6: D1 U=347.5, p<0.00; D2 U=548, p<0.05; D5 U=88, p<0.01; D6 U=6, p<0.01. Means (AT vs PT): D1 −9.98 vs −12.36 (PT more informationally dense); D2 6.61 vs 7.21 (PT more narrative); D5 1.60 vs −0.43 (AT more abstract/formal; PT more concrete); D6 −0.81 vs −1.99 (AT more on-line elaboration; PT more concise/planned). Feature-level differences include higher AT ranks for THAT verb complements, demonstratives, and adverbial subordinators (supporting more on-line elaboration and abstractness), while PT showed higher ranks for present participial clauses and other devices that enhance immediacy and narrative vividness. Overall: AT exhibits less separation between speech and narration, lower involvement in dialog, reduced informational density in narration, and greater abstractness across both registers. PT better preserves interpersonal engagement in dialog and informational density and concreteness in narration.

Findings address the research questions by showing systematic, statistically significant stylistic divergences between AT and PT across key MDA dimensions for both speech and narration. PT more closely aligns with expected register characteristics of fictional dialog (higher involvement, concreteness, less on-line elaboration) and narration (greater informational density, stronger narrativity, concision). AT’s tendencies toward abstractness and on-line elaboration, plus weaker dialog involvement and less dense narration, suggest potential impacts on reader engagement, character voice differentiation, and access to narrative information. Proposed explanatory factors include differences in translator training and native proficiency, team size and coordination, goals (niche community vs broader readership), and professional editing/quality control, all of which can shape register realization and stylistic cohesion.

The study demonstrates that amateur and professional translations of LCH differ systematically in style across speech and narration when analyzed via Biber’s MDA. PT maintains higher involvement in dialog and greater informational density and concreteness in narration, while AT shows less register differentiation, more abstractness, and more on-line informational elaboration. These differences may affect readers’ comprehension and engagement. The work contributes a corpus-based, multidimensional comparison of amateur vs professional literary translation in a major Chinese genre and underscores the importance of considering the stylistic effects of fan translation in cross-cultural dissemination. Future research should broaden text coverage beyond a single novel, apply complementary quantitative metrics (e.g., activity/descriptivity, dependency distance, entropy-based approaches), and integrate qualitative methods (e.g., translator interviews) to probe decision processes behind stylistic choices.

The analysis is limited to a single source text (LCH) and its two English versions, which may restrict generalizability. It focuses on MDA register features, leaving other linguistic/stylistic dimensions underexplored. The study does not incorporate translator process data (e.g., interviews), which could clarify motivations behind stylistic choices.