Three Faces of Heroism: An Empirical Study of Indirect Literary Translation Between Chinese-English-Portuguese of Wuxia Fiction

2.1 Indirect Translation and Its Impact on Theme Transformation

Indirect translation, also known as relay translation (Ringmar 2012), involves translating a text into an intermediate language before translating it into the target language. The multi-step process of indirect translation can greatly affect how cultural concepts are handled in different language versions (Zhou 2018). It often results in the preservation, alteration, or transformation of these concepts as they move from the source language through intermediate translations to the final target language. Recent years, indirect translation has gained increasing attention in translation studies, highlighting both its potential benefits and challenges (Pięta 2019; Pięta, Ivaska, and Gambier 2022; Rosa, Pięta, and Bueno Maia 2017). As a mediated or second-handed translation method, indirect translation has often been criticized for its greater distance from the original text (Kadiu et al. 2016), manifesting the need for a thorough examination of its reliability and fidelity to the source material.

Research has shown that indirect translation can significantly impact the way themes and concepts are transformed across languages and cultures (Dollerup 2000), leading to the loss of cultural nuances (Hekkanen 2014), the alteration of stylistic elements (Pięta 2017), and the reinterpretation of themes (Washbourne 2013). While some elements in the translation process may remain consistent, others are subject to shifts and omissions, especially in complex translation chains. As Mossop (2017, 135) notes, “sameness is not an exotic feature in translation.” However, when translations go through multiple stages, as in indirect translation, there’s a higher risk of changes or omissions occurring (Hadley 2017; Ringmar 2012), sameness can reveal essential aspects about texts and their translators. Elements that remain unaltered likely reflect what translators and other agents (such as editors) consider the core meaning of a given text and worth transferring to the target text.

To further examine the fidelity of indirect translation, existing academic works often focus on qualitative, close readings and manual comparisons of different language versions (Rosa, Pięta, and Bueno Maia 2017). However, this manual textual comparison is time-consuming and labor-intensive, particularly with large datasets. Recently, researchers have begun to adopt computational approaches in indirect translation studies (Ivaska 2020; Ustaszewski 2021). According to Ustaszewski (2021):

… Given the multitude of linguistic, cognitive and sociocultural factors that constrain the production of indirect translations (and its subtypes), rigorous empirical research into indirect translation has the potential to broaden the understanding of the complexity of translation and its relation to other types of constrained communication …. (p. 327)

Despite the growing body of research on indirect translation, there is a lack of studies specifically focusing on how it shapes the preservation, alteration, or transformation of heroism concepts in Wuxia literature. This gap in the current understanding of heroism transformation through indirect translation necessitates further exploration, particularly in the context of Wuxia fiction.

Our study addresses this gap by being the first to explore indirect translation results across different languages using computational methods, specifically the zero-shot learning algorithm BERTopic modeling method. While Ivaska and Ustaszewski’s works primarily focus on identifying the source language in indirect translation, an important aspect in understanding the intermediate translation process, they do not extensively compare translation results across languages to explore fidelity. Our innovative approach allows us to identify which elements remain consistent and which undergo changes, providing insights into the core meanings that translators aim to preserve versus those elements more susceptible to cultural adaptation and transformation.

2.2 Recent Developments in BERTopic Modeling and Its Application

Topic modeling techniques, such as Latent Dirichlet allocation (LDA) and Non-negative Matrix Factorization (NMF), have shown promise in computational literary studies by providing a suite of algorithms to discover hidden thematic structures in large collections of texts. As Callaway et al. (2020) explain, topic modeling “discovers a set of ‘topics’ – recurring themes that are discussed in the collection – and the degree to which each document exhibits those topics.”

Traditional topic modeling techniques, such as LDA and NMF, have been widely used in computational literary studies to uncover hidden thematic structures in large collections of texts. These methods have proven effective in identifying and exploring themes and trends within single-language literary corpora, as demonstrated by the works of Jockers (2013), Goldstone and Underwood (2014), and Schöch (2021). However, these traditional topic modeling methods have limitations that reduce their effectiveness in multilingual research contexts. They face particular challenges when analyzing abstract concepts like heroism across different languages and translations.

First, traditional topic modeling techniques rely on bag-of-words representations, which disregard word order and context. This simplification can result in the loss of important semantic details, making it difficult to capture the precise meaning of words in their specific contexts. In literary works, the precise use of language is crucial for conveying themes and ideas. Therefore, preserving these semantic details is essential for accurate analysis.

Second, when dealing with multilingual corpora, traditional topic modeling approaches face challenges in effectively comparing and aligning topics across languages. Researchers have proposed various strategies to address this issue, such as calculating separate language-specific models, adapting topic models to handle multilingual corpora, or translating documents into a common language (Lind et al. 2022). However, these methods have their own drawbacks. They often require large sets of pre-translated texts or depend on the accuracy of machine translation. These requirements can introduce new problems and potential errors in the analysis process.

Third, traditional topic modeling techniques often require extensive preprocessing and parameter tuning, which can be time-consuming and may introduce subjective biases into the analysis (Maier et al. 2022). This is particularly challenging when dealing with abstract concepts like heroism, as it may be difficult to determine the optimal parameters and preprocessing steps that effectively capture this theme across different languages and cultures.

In contrast, the application of advanced techniques like zero-shot BERTopic offers several advantages that make them particularly well-suited for our research on translingual thematic analysis. BERTopic leverages the power of BERT embeddings, which capture contextual information and semantic relationships between words (Devlin et al. 2018). By using these word representations in the topic modeling process, BERTopic can identify more accurate and meaningful topics. These topics better reflect the actual content of the texts (Grootendorst 2022). This enhanced semantic understanding is crucial when analyzing literary works, where the subtle use of language and context plays a significant role in conveying themes and ideas. The multilingual nature of BERT-based models also makes them inherently well-suited for translingual research. These models are trained on large-scale multilingual corpora, enabling them to capture cross-lingual semantic similarities and align topics across languages more effectively than traditional methods (Vulić et al. 2015). This ability is important when comparing original literary works with their translations. It helps researchers identify thematic patterns and similarities across different languages.

The zero-shot learning capability of BERT-based models is particularly useful in our current research, as it enables the classification of documents into predefined categories without the need for labeled training data (Alcoforado et al. 2022; Savelka and Ashley 2023). This feature allows us to predefine topics related to the abstract concept of heroism and explore which words are more closely associated with this theme, without relying on language-specific labeled corpora. Zero-shot topic modeling sets itself apart from traditional topic modeling approaches by enabling the discovery of predefined topics in large amounts of documents. In our study, we focus on the concept of heroism, aiming to investigate which words are more strongly connected to this relatively abstract topic, without the need for language-specific training data. This ability to work with predefined topics and the absence of a requirement for labeled training data are key aspects that distinguish zero-shot BERTopic from conventional topic modeling methods, making it particularly suitable for our research on the abstract concept of heroism across languages and cultures.

While traditional topic modeling techniques have been valuable in computational literary studies, their limitations in handling semantic nuances, multilingual corpora, and the need for labeled data make them less suitable for our research on translingual thematic analysis. The advanced capabilities of zero-shot BERTopic, including their ability to capture contextual information, handle multilingual data, and perform predefined topic classification without labeled training data, make them a better fit for our research goals.

3.1 Data Collection and Pre-Processing

Data collection: The data for this study is constructed from the first nine chapters of She Diao Ying Xiong Zhuan 2nd edition (射雕英雄传 第二版) and its English translation, A Hero Born: Legends of the Condor Heroes Vol. 1 (trans. Anna Holmwood), as well as its Portuguese translation, Nasce um Herói Lendas dos Heróis do Condor – Livro 1 (trans. Elsa T.S. Vieira).^[1] Since the target (Portuguese) version only translates the first volume (chapters one to nine) from English, this study will analyze only the first nine chapters in the Chinese, English, and Portuguese versions. The Portuguese version, being an indirect translation derived from the English text, positions the data as a sequential translation chain, ideal for analyzing thematic shifts across the translation process.

Standardization of Named Entities: To ensure analytical accuracy, especially concerning character names, martial arts techniques, and place names, we standardized these terms across the different language versions to ensure consistency. For example, Ke Zhen’e is standardized to Ke_Zhen_E, and Lotus Huang (Huang Rong’s name in the English translation) is standardized to Lotus_Huang. This step is vital for addressing variations in how these elements are presented in different contexts and ensuring they are correctly linked to avoid segmentation issues. Two student helpers from translation major participated in this process to ensure consistency and accuracy. The standardized terms were then verified and cross-referenced to maintain consistency.

Data Cleaning Process: The texts undergo a cleaning process to ensure data quality. First, paratextual content, including prefaces, footnotes, and appendices that are not directly related to the main text, is manually removed to maintain focus on the core narrative. In the following parts, Python (v. 3.11.4) was used for other preprocessing tasks, including removal of stop words and text segmentation. A standard set of stop words is then applied using well-established dictionaries: the Harbin Institute of Technology Stop Word List for the Chinese text and the Natural Language Toolkit (NLTK) (v. 3.8.1) Stop Word List for the English and Portuguese texts. Text segmentation is performed using Jieba (v. 0.42.1) for the Chinese text and NLTK for the English and Portuguese texts. To fit the requirements of the BERTopic model, the novel text is segmented into slides of around 50 words each.

3.2 Topic Modelling Procedure

Predefined topic names: Zero-shot Topic Modelling is suitable for scenarios where topics are predefined or potentially contained in the context. In the case of this study, it is assumed that some topics reflect the nature of heroism. Therefore, by applying zero-shot topic modelling, the level of heroism and the differences in expression between different languages can be determined. To effectively utilize this approach, topic names that capture the essence of heroism in the context of the studied work need to be identified.

The identification of these topic names requires a deep understanding of the concept of Chinese heroism, or Xia Yi. Scholars have attempted to define the essential qualities of the Xia, the heroic figures in Chinese literature, despite the abstract nature of the concept. James Liu, in his work The Chinese Knight-errant (2022) presents a list of heroic qualities, including altruism, justice, individual freedom, personal loyalty, courage, truthfulness and mutual faith, honor and fame, and generosity and contempt for wealth. In contrast, Xu (1995) distills the defining traits of the xia down to two core aspects: “disdain for wealth” and “disdain for [one’s own] life” (p. 5). While these analyses provide valuable insights, it is crucial to recognize that such essentializing definitions are abstractions and may not fully capture the specific manifestations of heroism in particular literary works.

Considering the general understanding of Chinese heroism and the specific narrative of The Legend of the Condor Heroes, four topic names were identified to represent the core aspects of heroism in the novel: patriotic, brotherhood, courage, and loyalty. These topic names serve as the basis for the zero-shot BERTopic model to identify and classify related thematic content within the corpus. The rationale for the choice of these four topic names is as follows:

Patriotic: The narrative of The Legend of the Condor Heros is set in historical background of the Song dynasty’s (960–1279) struggles against northern invaders. Although the fate of the Chinese empire in the face of foreign aggression was one of the main concerns in Jinyong’s Wuxia novel, The Legend of the Condor “represents the consummation of that heroic nationalism” (Hamm 2005, p. 79) The characters’ actions are often driven by a deep love for their homeland, manifesting in the defense of their country and the upholding of its honor. The word ‘patriotic’ captures this narrative arc, reflecting the heroes’ dedication to a cause greater than themselves. Patriotism resonates deeply through the narrative, as protagonists like Guo Jing find themselves embroiled in conflicts that transcend personal ambitions, bound by a higher calling to uphold justice and defend the nation against the insidious forces of the Mongol invaders. The topic name “patriotic” is chosen to describe this devotion to one’s country and the willingness to make immense sacrifices for the greater good.

Brotherhood: In The Legend of the Condor Heroes, brotherhood is a fundamental aspect of heroism. The story begins with two sworn brothers Guo Xiaotian and Yang Tiexin exemplify this by forging an unbreakable bond through shared martial arts heritage and traditional oaths of brotherhood. The naming of their unborn children, Guo Jing and Yang Kang, by Qiu Chuji further reinforces the significance of brotherhood, linking their shared destiny and responsibility to redeem the nation’s honor. Throughout the novel, the concept of brotherhood extends to various characters united in their struggles, loyalty, and support for one another, serving as the foundation for their heroic journeys and triumphs. Choosing “brotherhood” as a topic name acknowledges the vital role of these profound, non-familial relationships in shaping the heroic identities and actions of the characters.

Courage: At the heart of the novel’s action-packed episodes is courage – the willingness to confront fear, uncertainty, and significant risk. As Wang (1975) stated in his paper on “Towards Defining a Chinese Heroism”: “It is the heroism that exhibits the courage of approaching the cause with valor and appreciating the effect with wisdom” (p. 26). The heroes of the story exhibit this trait in abundance, whether in battle or in moral dilemmas. This label aligns with the numerous instances where characters are called upon to act bravely, thus embodying the traditional hero’s journey. Throughout the narrative, heroes like Guo Jing and Huang Rong must confront formidable adversaries and daunting trials with unwavering bravery and determination, exemplifying the virtue of courage. The topic name “courage” connects with this spirit of fearlessness and resilience that defines the archetypal wuxia protagonist.

Loyalty: In Wuxia novels, “A hero pursues loyalty to their family, nation, and world, especially the loyalty to the “justice” in their minds” (Bai 2022, pp. 11–12). Loyalty emerges as a paramount ideal, reflected in characters’ steadfast commitment to their principles, causes, and relationships. Integral to the novel’s portrayal of its protagonists is their unwavering loyalty – to each other, to nation, to their masters, and to the ideals they cherish. The choice of topic name “loyalty” dictates the narrative flow, influencing decisions and alliances, and is a reflection of the Confucian values of fidelity and righteousness that are prevalent in Chinese culture.

Through the lens of these topic names – patriotic, brotherhood, courage, and loyalty – our computational analysis using zero-shot BERTopic modeling aims to illuminate how the foundational themes of The Legend of the Condor Heroes are translated and potentially transformed across cultural boundaries. These four topics, chosen for their central role and frequent appearance in the text, represent the main values and virtues highlighted in the narrative. By incorporating these terms, we not only focus our analysis on the cultural specificity of the original text but also create a basis for examining the thematic consistencies and variations within the translated works. Other potential topic names were considered, but these four were chosen for their relatively comprehensive representation of the heroic themes in the novel. Moreover, the choice of four predefined topic names aligns with the natural generation of 3–4 topics by BERTopic under unsupervised modeling, ensuring a rational range and avoiding potential topic overlapping that might occur with a higher number of topics.

Topic Modeling Process: Zero-shot BERTopic, a highly modular topic modeling workflow, is employed to analyze the prepared corpus using predefined topic names related to heroism in The Legend of the Condor Heroes. Using the Python package bertopic (v. 0.16.0), this approach utilizes various components, including embeddings, dimensionality reduction, clustering, tokenization, and weighting schemes, to generate topics and their most related words, along with probability scores indicating the likelihood of documents belonging to each theme.

In this study, the settings for BERTopic are as follows (Table 1):

After processing the corpus through these components, the model generates topics and their corresponding top 10 most related words. These top 10 words are considered the most representative of each theme due to their high relevance and contribution (Grootendorst 2022). Words beyond the 10th rank have relatively low TF-LDF scores and are considered less valuable for the analysis, as they may not significantly contribute to the overall understanding of the themes.

3.3 Data Analysis

Following the application of Zero-shot BERTopic modeling to each language version, the study conducts a data analysis to identify similarities and differences in heroism representation across the translations. The analysis consists of two main components: comparative analysis and cosine similarity analysis.

4.1 Topic Cases: Word Inventories by Topic Names

This section examines the representation of heroism across language versions by analyzing specific words used in four thematic topics. The analysis focuses on bar charts that illustrate the distribution of keywords and scores for themes of brotherhood, courage, and loyalty. All keywords will be placed in three tables, one for each language used in the English, Chinese, and Portuguese versions, so that a side-by-side comparison of the vocabulary at hand can be made within each topic. Such a tabular format will better present the data clearly and concisely for cross-referencing purposes regarding similarities and contrasts in lexical choice the translators chose to use.

In the bar carts, for consistency and easy comparison across languages, all the keywords across the three versions will be translated to English for presentation purposes. The green bars represent the appearance of the keywords in the original Chinese text, the blue bars represent that in the English translation, and the orange ones represent that in the Portuguese. On the other hand, in each table, they will appear in ascending order of the probability of relevance to the topic. This ordering will highlight the words that are of especially high prominence and frequency within any given theme, which will be instructive in showing the cultural and linguistic emphases found in each translation regarding what it means to be a hero.

1. Patriotic (爱国/Patriota)

Based on the comparative analysis, it is evident that the keywords associated with patriotism in the three languages are relatively dispersed (see Figure 2 and Table 2). The only keyword that appears in all three languages is “soldier,” which is highly relevant to the theme of patriotism in both the English and Portuguese versions. This indicates that the concept of patriotism is linked to soldiers and warfare in both Chinese and Western contexts. However, aside from this term, the other keywords vary across the three languages.

Figure 2:

Keyword probability scores for patriotic topic.

In the original Chinese text, the protagonist “Guo Jing” and the national hero “Yue Fei” are prominent under the theme of patriotism. Guo Jing, the main character in The Legend of the Condor Heroes, symbolizes traditional Chinese patriotism, characterized by his loyalty to the country and determination to resist the Mongol invaders. Yue Fei, a historical figure known for his patriotic spirit, further reinforces this theme (Hamm 2005). However, these figures are absent in the English and Portuguese versions, suggesting a shift in how patriotism is portrayed in these translations. The English version includes more references to Mongol characters such as “Temujin” (Genghis Khan), highlighting the historical and military context of the narrative. This emphasis aligns with Western readers’ understanding of patriotism through the lens of historical conflicts and military endeavors. Terms like “fight” and “patriot” reinforce the association with warfare and national defense, making the theme of patriotism accessible and relatable to a Western audience. In the Portuguese version, patriotism is conveyed through a combination of martial terms like “soldados” (soldiers) and emotional expressions such as “coração” (heart). This suggests that in the Portuguese cultural context, patriotism is both a public duty and a personal conviction, characterized by emotional depth and individual valor. This dual representation reflects the blend of Chinese cultural elements and Portuguese cultural interpretations in the translation.

The concept of patriotism in the Chinese context is often intertwined with historical and collective identity, deeply rooted in the mission to protect the nation and cultural heritage. Figures like Yue Fei represent the highest ideals of sacrifice for the country, emphasizing collective loyalty over individual desires (Hamm 2005). In contrast, Western interpretations of patriotism focus more on individual bravery and personal commitment to national ideals, often embodied by soldiers and acts of heroism on the battlefield. This can be seen in the prominence of terms like “soldier” and “fight” in the English version, which aligns with Western heroic narratives. In the process of indirect translation, the transformation of the theme of patriotism is evident. The Portuguese version, as an indirect translation through English, retains some elements of the source text and the intermediary text, but also incorporates unique aspects of the Portuguese cultural context. The absence of key characters like Guo Jing and Yue Fei in the English and Portuguese versions suggests that the translators did not see these figures as central to the theme of patriotism in the same way as the original Chinese text. This indicates that indirect translation can lead to shifts in the portrayal of key themes, as the intermediary language adds its own layer of interpretation before the final translation (Pięta 2019). The thematic shift observed in the portrayal of patriotism across the different language versions reflects how each culture interprets and emphasizes this concept. In The Legend of the Condor Heroes, patriotism is a key theme that shapes characters’ motivations and actions. The different ways this theme appears in various translations show how challenging it can be to convey culturally specific ideas through direct and indirect translation processes (Ivaska 2020).

1. Brotherhood (兄弟情/Irmandade):

The analysis of the topic “brotherhood” in the three languages reveals a significant degree of consensus in the understanding of this theme (see Figure 3 and Table 3). Unlike the more varied interpretations of patriotism, the theme of brotherhood exhibits more commonalities across Chinese, English, and Portuguese. The graph indicates that five keywords overlap among the three languages: the protagonist “郭靖” (Guo Jing), his martial arts mentor “柯镇恶” (Ke Zhen’e), Guo Jing’s father’s sworn brother “杨铁心” (Yang Tiexin), and Guo Jing’s father’s close friend “丘处机” (Qiu Chuji). This suggests a shared recognition of friendship and brotherhood in all three languages.

Figure 3:

Keyword probability scores for brotherhood topic.

In the original Chinese text, the theme of brotherhood is deeply rooted in traditional values and interpersonal relationships. The prominent keywords, such as “郭靖” (Guo Jing), “柯镇恶” (Ke Zhen’e), “丘处机” (Qiu Chuji), and “杨铁心” (Yang Tiexin), highlight the strong bonds of mentorship, loyalty, and camaraderie that are central to the Wuxia genre. These relationships reflect the Confucian ethic of loyalty and the importance of camaraderie and alliances in the face of adversity (Hamm 2005). Guo Jing’s relationships with his mentors and peers are pivotal in illustrating the theme of brotherhood, emphasizing the traditional Confucian values of loyalty, respect for teachers, and strong familial bonds. In the English version, terms like “brother,” “the_seven_freaks,” and “zhu_cong” reflect a more active and combative expression of brotherhood, aligning it with physical camaraderie and shared battles. The inclusion of “ironheart_yang” and “Temujin” suggests a focus on historical and legendary figures who exemplify brotherhood through their actions and leadership. This portrayal aligns with Western heroic archetypes that emphasize individual valor and collective action. The Portuguese version mirrors these themes with terms such as “irmão” (brother), “coração_de_ferro_yang” (Ironheart Yang), and “sete_excêntricos_do_sul” (the Seven Freaks of the South). The use of “olhos” (eyes) and “mão” (hand) in the Portuguese context could symbolize perception and action within these relationships, suggesting a tactile and observant aspect to brotherhood. This indicates that in the Portuguese cultural context, brotherhood is portrayed with a blend of emotional depth and physical camaraderie.

The overlap of key figures in all three languages highlights how the relationships central to Guo Jing’s narrative are universally recognized as embodying the concept of brotherhood. Guo Jing’s relationships with his mentors, Ke Zhen’e and Qiu Chuji, are key to his growth as a hero. These connections highlight Chinese cultural values of loyalty and respect for teachers, while also showing the importance of mentorship and friendship in Wuxia stories (Hamm 2005). Furthermore, the indirect translation process also plays a role in the thematic representation of brotherhood. The Portuguese version, as an indirect translation through English, retains elements from both the source text and the intermediary text while incorporating unique aspects of the Portuguese cultural context.

1. Loyalty (忠诚/Lealdade):

The topic of “loyalty” demonstrates both convergence and divergence among the three languages (see Figure 4 and Table 4). Shared keywords such as “郭靖” (Guo Jing), “完颜洪烈” (Wanyan Honglie), “柯镇恶” (Ke Zhen’e), and “铁木真” (Temujin) indicate a common recognition of loyalty across Chinese, English, and Portuguese versions. These overlaps suggest that these characters universally embody loyalty within the narrative framework. However, the Chinese original text uniquely emphasizes familial and marital loyalty, evident through keywords like “王妃” (princess), “娘子” (wife), “丈夫” (husband), and “女子” (woman). This highlights the cultural importance of loyalty in Chinese marriage and family relationships, where personal and familial duties are inseparable from the concept of loyalty. In contrast, the English and Portuguese translations present a different perspective. The English text includes terms like “prince,” “son,” “khan,” and “young,” which align loyalty with personal honor, leadership, and hierarchical relationships rather than familial ties. The Portuguese version highlights terms such as “irmão” (brother), “mestre” (master), “kung_fu,” and “artes_marciais” (martial arts), emphasizing a sense of camaraderie and martial honor, thereby framing loyalty within the context of martial brotherhood and fellowship.

Figure 4:

Keyword probability scores for loyalty topic.

These differences show how each translation adjusts the idea of loyalty to fit its cultural setting. The Chinese text focuses on loyalty in marriage and family, while the English and Portuguese versions link it more to warrior virtues and personal honor. This demonstrates that translation, especially indirect translation, involves significant challenges and changes as it moves cultural and thematic nuances from one language to another.

1. Courage (勇气/Coragem)

The analysis of the “courage” theme across the three language versions reveals both shared elements and unique cultural interpretations, much like the themes of patriotism, brotherhood, and loyalty previously examined (see Figure 5 and Table 5). Common Terms like uch as “郭靖” (Guo Jing), “铁木真” (Temujin), “丘处机” (Qiu Chuji), and “战斗” (fight), suggest their common understanding as symbols of courage in the novel. In all translations, Guo Jing is portrayed as the typical fearless hero. Throughout the novel, he performs many brave acts against the strong Jin invaders. He also faces moral challenges that test his loyalty to both his Mongolian upbringing and his Han Chinese heritage. For example, Guo Jing’s defiance of Genghis Khan’s mandate that civilians be slaughtered in such a way as to safeguard the innocent demonstrates a more fundamental moral type of courage that extends well beyond physical bravery.

Figure 5:

Keyword probability scores for courage topic.

The original Chinese text puts particular stress on martial valor and combative contexts through its usage of such keywords as “敌人” (enemy), “武功” (martial arts), and “柯镇恶” (Ke Zhen’e). This reliance on physical bravery and military might is far from the original Wuxia genre olm where the heroes were not only praised for fighting skills, but also for bravery in facing formidable opponents. Personalities like Qiu Chuji and Temujin are seen to appreciate not only their fighting capabilities but also the strategies employed, and leadership exhibited as a broader perspective of courage. This is contrasted with the title in English, “young,” “prince,” and “kung_fu,” which expands the definition of courage to include youthful bravery and noble characteristics. Among these elements, the Western heroic archetypes of individual valor and moral fortitude may well have been followed. This incorporation of “kung_fu” maintains a cultural particularity that caters to the Western reader’s expectations of an exotic mysticism in martial arts, but “prince” pairs courage with nobility and leadership, not as directly associated perhaps, in the original text. The particular aspects of the Portuguese translation of “mestre” (master) and “caridade_bao” (Charity Bao/Bao Xiruo), offer support for a crossbreed story of emotional drama and combat honor. This bravery, in that respect, is componential in operation since it spreads from individual courage to the influence of mentors and comrades. It is particularly the use of the term “charity_bao” that confirms that between the original book in Chinese and the Portuguese text, the latter will always tend towards having characters that are bodily and morally courageous as humanitarian becoming prominent tends to dissimilate culturally with value placed on meanings.

A comparison with the theme of patriotism shows that, although there were commonalities in terms of “soldiers” but were otherwise quite different, there is something here that shows up consistently: heroism is represented everywhere, just as it was in all cultures, but the particular characteristics and stories are different. In the theme of patriotism, the different languages represented their cultural contexts: Chinese emphasized historical figures, English military vocabulary, Portuguese affective expressions with militaristic vocabulary. In the same way, brotherhood as a theme shared common keywords: “郭靖” (Guo Jing) and “柯镇恶” (Ke Zhen’e) but varied in cultural emphasis. The Chinese version presented it through family and mentor relationships while the English one did so through camaraderie and shared battles. The aspect came out through emotional bonds and mutual support in the Portuguese version. In the theme of loyalty, differences were more pronounced, with the Chinese text emphasizing familial and marital loyalty, which was less apparent in the English and Portuguese texts. The English translation aligned loyalty with personal honor and leadership, while the Portuguese version presented a multifaceted view combining emotional depth with martial camaraderie.

These are the shifts in how courage, and really all of these other themes, are presented during indirect translation. All the base elements of bravery, patriotism, brotherhood, and loyalty remain in place, but each version of the different language shifts those themes again into those that will hold weight for them culturally. More importantly, the Chinese focus on martial valor, the English emphasis on the individual and noble boldness, and the Portuguese combination of martial and moral courage --all this, together, enriches the narrative by giving the reader a diverse perspective as to what it is like to be a hero. Such differences point to the challenges of indirect translation with thematic fidelity. The core heroic traits of main characters like Guo Jing and Temujin remain consistent across translations. However, the way these traits are presented and celebrated varies, reflecting different cultural interpretations. This leads towards an argument about cultural mediation in translation, where each language carries with it its own spectacular requisite for a shared theme, essentially reconfiguring the thematic essence of the original narrative.

4.2 Cross-Linguistic Similarity Analysis: English-Portuguese, English-Chinese, Portuguese-Chinese

The cosine similarity scores provide valuable insights into the thematic concordance between the three translations. The English to Portuguese (EN_PT) translation exhibits the highest level of thematic alignment, with a score of 0.572, likely due to their shared Western cultural context and direct translation relationship. The English to Chinese (EN_ZH) comparison shows a moderate score of 0.5059. This suggests that while some themes are consistent, there are also noticeable changes due to cultural differences in translation. The lowest similarity score of 0.4362 between Portuguese and Chinese (PT_ZH) reveals the challenges of indirect translation. Each step in the translation process adds new interpretations that can significantly change the original themes (Figure 6).

Figure 6:

Similarity between Chinese, English and Portuguese languages’ versions.

These findings directly address the questions raised by Pięta (2019) in her article: “Where are more changes introduced: during the transition from the ultimate source text to the mediating text? Or in the passage from the mediating text to the ultimate target text?” (p. 29). Pięta discovers that in literary texts, more changes tend to be introduced in the first part of the indirect translation chain, often resulting in a situation where the ultimate target text is a faithful rendering of the mediating text, but the mediating text is an unfaithful version of the ultimate source text, possibly due to uneven power relations between the languages involved (2019).

While our study’s results confirm Pięta’s argument, they also suggest that power relations between languages are not the sole factor contributing to changes in indirect translation. In our case, cultural differences emerge as another essential cause of changes, a factor not discussed in Pięta’s justification. By examining the changes in cultural themes and concepts, our study complements Pięta’s work and provides another perspective to the question of faithfulness in indirect translation.

To better understand these findings, it is crucial to consider the role of the translator as a cultural mediator. In direct translation, like Chinese to English, the translator acts as a link between the original and target cultures, trying to balance their differences. But in indirect translation, such as English to Portuguese, this cultural balancing act becomes more complicated. The Portuguese translator must navigate not only the source culture (Chinese) and target culture (Portuguese) but also the intermediary culture (English). This process of cultural negotiation across multiple languages and contexts can result in a less faithful representation of the source culture’s heroic ideals in the final translation.

The current research makes significant contributions to our understanding of translation and its implications for culturally-loaded concepts such as heroism in three key ways. First, through quantitative and empirical approaches, this study confirms that indirect translation (Chinese-Portuguese) produced more changes in these concepts than direct translation (Chinese-English). Second, it reveals significant differences in similarities even during direct translations, specifically between Chinese-English and English-Portuguese. Third, it demonstrates that the distance between the target text and the source text can be attributed to the changes introduced by the translator of the mediating text.

In essence, when the fidelity of the mediating text to the source text is compromised, it has a spillover effect on the target text, causing the target text to become more dissimilar to the original source text. This finding challenges the simplistic notion that the target language translation in indirect translation is merely “a copy of a copy of a copy” (Landers 2001, 131). Moreover, it reinforces the earlier finding from the analysis of theme names, which indicated that the Portuguese translation shows different preferences than the English one in relation to various themes.