Taboo Language in Non-Professional Subtitling on Bilibili.com: A Corpus-Based Study

Abstract

This qualitative and quantitative study examines how taboo language is rendered in non-professional subtitling (NPS), how viewers react to the renderings, and how the interactions between danmu and general comments’ contributors affect the translation activities and language changes. The study draws on a parallel corpus consisting of taboo language and its translations from 18 of the most-viewed and commented upon subtitled videos on the popular video-sharing platform, Bilibili.com. Danmu comments and general comments related to the renderings of taboo language are also collected and studied. When analyzing translation activities in an NPS setting, the study adopts and modifies some mainstream subtitling strategies and techniques proposed by. The study finds that various creative approaches are adopted, such as lexical recreation and substitution by euphemism. While the strength of the taboo language is reduced in more than half of the instances, in an unexpected 17.2% of cases the effects are enlarged. The study concludes that a virtuous, collaborative mechanism for potential translation problems and language learning is formed by providing positive, neutral, and critical feedback in the comments. In addition to linguistic knowledge and cultural background, viewers also share knowledge beyond the scope of translation.

1. Introduction

Translating taboo language is a daunting task and solving it with spatial–temporal constraints makes it even more challenging. According to Díaz-Cintas and Remael, taboo language, or emotionally charged language (Díaz-Cintas and Remael 2007, 2021), is often toned down in audiovisual translation (AVT) by deleting the words and expressions, if space is limited, or by using euphemisms (Díaz-Cintas and Remael 2007; 2021, p. 189). In recent years, several studies have focused on increasingly diverse subtitling strategies and techniques in rendering taboo language in professional and non-professional subtitling (NPS). Some scholars noted the fluidity, heterogeneity, and fuzziness of NPS (Dore and Petrucci 2021; Orrego-Carmona and Lee 2017). One of the main issues is that this NPS term implies the lack of professional training and thus poor translation quality, neglecting the creative and flexible nature of this kind of subtitling practice.

However, as Orrego-Carmona (2016, p. 231) argues, NPS practices “do not necessarily contradict” the professionally established standards for subtitling. In recent years, the focus of scholarship on NPS has gradually transformed from this binary to focusing more on the characteristics of NPS practices and the blurring boundary between NPS and professional subtitling. However, very little research has been conducted to examine the translation activities and interactions on translation and language issues related to taboo language in an NPS setting.

This qualitative and quantitative study draws on a parallel corpus consisting of taboo language and its translations from 18 of the most-viewed and commented upon subtitled videos on the most popular video-sharing platform for Chinese young people under 24 years old, Bilibili.com, where NPS practices flourish (China Tech Express 2020). In addition to the corpus, danmu or danmaku in Japanese comments and general comments related to the translations of the taboo language are also collected and studied.

When analyzing translation activities in an NPS setting, the study adopts and modifies some mainstream subtitling strategies of culture-specific references. There is no widely accepted framework specifically for taboo language, which is considered part of a culture.

Two of the most widely accepted taxonomies from Díaz-Cintas and Remael (2007, 2021) and Pedersen (2011) are modified based on the instances of this study. In addition, the study analyzes the expected effects received after taboo language is rendered, based on Ávila-Cabrera’s taxonomy (Ávila-Cabrera 2017, 2020), including toning up, toning down, and maintaining.

Based on Ávila-Cabrera’s taxonomy (Ávila-Cabrera 2016) with considerations of Chinese contexts, taboo language in this study is defined as words and expressions considered as restricted or prohibited in the target culture and contexts, i.e., the Chinese culture and the guidelines of the platform of the study—bilibili.com. They may include but are not limited to swearwords, sexual references, and illegal content under Chinese laws and regulations (for instance, drugs and gambling).

The following sections review the literature on taboo language, taboo language in AVT, translation activities on Bilibili.com, and subtitling strategies and techniques and effects.

2. Literature Review

2.1. Taboo Language

According to Allan and Burridge (2006, p. 11), the definition of the word taboo is “a proscription of behaviour for a specific community of one more persons, at a specifiable time, in specifiable contexts”. Taboo language or taboo words, in this sense, are “expressions whose use is restricted or prohibited by social custom” (Díaz-Cintas and Remael 2021, p. 181). Euphemism, a term that is associated with taboo words, is usually used to avoid embarrassing or unpleasant topics as an alternative to taboo language (Hughes 2006, p. 151).

Although taboo language changes over time and contexts, it is not unstable. According to Jay (2009), a list of 10 English words (fuck, shit, hell, damn, goddamn, Jesus Christ, ass, oh my god, bitch, and sucks), which account for 80% of public swearing, has remained lasting from 1986 to 2006. Jay also notes that highly offensive words, such as cunt, cocksucker, and nigger, occur less frequently during this period. Interestingly enough, more instances of the 10 words, e.g., fuck, shit, hell, damn, etc., are observed in this study, while there is only one highly offensive word (nigger) in the 18 selected videos.

The classifications of taboo language are consistent in the literature. Allan and Burridge (2006, vii) propose four categories of taboo: (1) naming and addressing, (2) sex and bodily effluvia, (3) food and smell, and (4) disease, death, and killing. Jay (154) expanded the four categories into eight as follows and points out that it is helpful to qualify references to “taboo words” by noting what taboo category they represent:

(1)

Sexual references;

(2)

Profane or blasphemous references;

(3)

Scatological referents and disgusting objects;

(4)

Ethnic–racial–gender slurs;

(5)

Insulting references to perceived psychological, physical, or social deviations;

(6)

Ancestral allusions;

(7)

Substandard vulgar terms; and

(8)

Offensive slang;

Based on Wajnryb (2005); Hughes (2006); and Jay (2009)’s studies, Ávila-Cabrera (2016) proposes a more comprehensive taxonomy of offensive and taboo language, including abusive swearwords, expletives, animal name terms, ethnic/racial/gender slurs, psychological/physical condition, sexual/body part references, urination/scatology, filth, drugs/excessive alcohol consumption, violence and death/killing, invectives and profane/blasphemous words/phrases.

Similar to other culture-specific references, taboo language is culture-bound. However, significant works on linguistic taboos, including Euphemism and Dysphemism (Allan and Burridge 1991) and An Encyclopedia of Swearing (Hughes 2006), are silent on the Chinese language and culture, as Jing-Schmidt (2019) argues. In Chinese history, the idea of taboo or mystical prohibition had existed for over two millennia when it was mentioned in The Book of Rites from the Western Han (202-8 BCE):

入境而问禁，入国而问俗，入门而问讳 （《礼记.曲礼上》）

Inquire about the legal prohibitions when entering a country, inquire about the customs when entering a metropolis, inquire about the unmentionables when entering a residence.

(Translated by Jing-Schmidt 2019)

Unlike Western scholarship on taboo, which comes from various disciplines such as linguistics, sociology, psychology, history, and archaeology, Chinese language scholarship mainly comes from folklore and cultural linguistics. Folklorist Ren (1991, p. 6) classified Chinese taboos into six categories: (1) names (e.g., of ancestors, senior family members, rulers, and the deceased); (2) age and zodiac; (3) death and calamity; (4) vulgarity, especially sexual activities and body parts; (5) wealth decline and bankruptcy; and (6) animal names relevant to certain professions. Another Chinese scholar, Chan (2016, p. 380), draws on Allan and Burridge’s study (Allan and Burridge 2006) and describes Chinese taboo in five thematic categories, including (1) bodies and their effluvia; (2) organs and acts of sex, and sexuality; (3) diseases, mental and physical defects, and death; (4) naming and addressing; and (5) other aspects of human existence. Despite the nuanced ways of classification, the motivations of Chinese taboo are usually associated with “uncleanness, filthiness and obscenity, privacy, fear, bad luck, and respect” (Chan 380). It should be noted that both the abovementioned classifications, including diseases and death and naming and addressing, build on ancient Chinese culture and thus those categories do not apply to this study.

Jing-Schmidt noticed a more substantial interest in examining “curse words in historical texts than in contemporary usage”, and she believes the reason may be the historical texts’ “comfortable distance from the embarrassment of contemporary language uses” (Jing-Schmidt 2019). Whatever the reasons, the lack of study of contemporary Chinese taboo leads to outdated classifications, especially under the successful development of new media and ongoing globalization.

While many instances fall into the previously mentioned categories, some language use that should be considered taboo either falls into the categories proposed by Allan and Burridge or Jay or does not fall into any preceding categories. Examples of the former are racial slurs, such as nigger and ching chang chong, while an example of the latter includes smoking weed. This example is considered as violating Chinese laws and regulations.

Since this study examines how English taboo language is rendered into Chinese, the literature on both languages and cultures is considered. Based on the literature, taboo language is defined as words and expressions that are considered as restricted or prohibited in the target culture and contexts, i.e., the Chinese culture and the guidelines of the study’s platform—Bilibili.com. They include but are not limited to swearwords, sexual references, and illegal content under Chinese laws and regulations, for instance, drugs and gambling. According to the cases of this study, Ávila-Cabrera’s taxonomy (Ávila-Cabrera 2016) is adopted, and minor adjustments are made.

2.2. Taboo Language in AVT

Translating cultural references related to taboos is challenging and adding the spatial–temporal constraints makes it seem impossible to render. According to Díaz-Cintas and Remael, taboo language, or emotionally charged language (Díaz-Cintas and Remael 2007, 2021), is often toned down in AVT either by deleting the language if spatiotemporal constraints are limited or by using euphemisms (Díaz-Cintas and Remael 2007; 2021, p. 189). Some studies (such as Lie 2013; Han and Wang 2014; Yuan 2015; Ameri and Ghazizadeh 2015; Stephan 2016; Trupej 2019; Al-Yasin and Rabab’ah 2019; and Liang 2020) that examine how taboo language is rendered in AVT confirm this argument.

These previous studies examine taboo language—including swearwords, sexuality, and sexual references—found in professional and non-professional subtitles and dubbed versions of movies, reality shows, and TV series in Arabic, Brazilian-Portuguese, Chinese, Norwegian, Persian, and Slovenian languages translated from English. The conclusions consistently reveal that euphemism and omission are primarily used to translate taboo words. Some scholars argue that self-censorship and subtitlers’ incompetency may be the reasons (Trupej 2019), while some link it to an increasingly improved identity in the target context (Liang 2020).

In more recent years, some studies (Ávila-Cabrera 2015, 2016, 2020; Pratama 2017; Giulia Magazzù 2018; Díaz-Perez 2020; Alsharhan 2020; Valdeón 2020; MicKayla Wilkinson 2021) have focused on increasingly diverse subtitling strategies and techniques in rendering taboo language. In Ávila-Cabrera’s descriptive studies (Ávila-Cabrera 2015, 2016, 2020), he examines the offensive and taboo language in the subtitled Quentin Tarantino films into European films, and among the 645 cases analyzed, 61.2% were transferred and 38.8% not transferred (30.1% omitted, 8.7% neutralized). This phenomenon is probably because of technical constraints and the fact that Tarantino’s films are full of offensive and taboo words. Thus, many of them cannot be transferred to the subtitles.

Some studies deviate from Díaz-Cintas and Remael’s claim. For example, Valdeón (2020) analyzed 412 pairs of swearwords in four series (The IT Crowd, Chicago PD, Brothers & Sisters, and Eyewitness) and found that the number of swearwords increases in over half of the instances, while toning down and omission only account for 13.88%. Alsharhan (2020) also summarized a variety of subtitling strategies used to render taboo language into Arabic in addition to omission and euphemism.

Comparative studies (including Beseghi 2016; He 2018; Khakshour Forutan and Modarresi 2018; Al-Jabri et al. 2021; Dore and Petrucci 2021) and reception studies (such as Briechle and Eppler 2019) have also been conducted to help better understand taboo language in audiovisual settings. He’s study (He 2018, p. 80) examined Chinese subtitles in 51 English films and concluded that the fansubbed version “transferred a higher severity of swearing” than the version rendered by the professional subtitlers. Dore and Petrucci (2021) jump out of this “toning down or not” dichotomy and noted the different treatments of coarse language after analyzing and comparing the Italian dubbed, subtitled, and fansubbed versions of the same American TV series. They believe that the fluidity of professional and non-professional subtitling is the main reason.

In the edited volume Non-Professional Subtitling (Orrego-Carmona and Lee 2017), the editors argue that the fluidity and heterogeneity of non-professional translation practices pose challenges for “proposing an exclusive and clear typology that applies to all instances of non-professional translation” (4).

Thus, the volume was entitled Non-Professional Subtitling to “shed light on the fuzziness and organic nature” of this field. One of the main problems with the term NPS is that it implies the lack of professional training and therefore poor translation quality, neglecting the creative and flexible nature of this kind of subtitling practice. Terms that are closely associated with NPS include fansubbing (Nornes 1999), user-generated translation (O’Hagan 2009), volunteer translation (Pym 2011), community translation (O’Hagan 2012), collaborative translation (Costales 2012), and social translation (Jiménez-Crespo 2017). These nuanced terms are much overlapped and refer to subtitling or translation practices conducted by non-professionals who do not receive and do not require monetary remuneration for the activities they perform and the translations they provide (Orrego-Carmona 2015).

This study adopts the term “non-professional subtitling” as the selected videos are mostly stand-up comedies rather than anime, whose contributors would be fansubbers by definition. Terms such as user-generated translation or collaborative translation are not used because one of the research goals is to examine whether there are activities related to translation in the subtitles, danmu, and general comments on Bilibili.com; in other words, if and how user-generated translation or collaborative translation exists. The rest of the terms are not relevant to the thesis of this paper and thus are not further discussed.

In recent years, as Orrego-Carmona (2016) argues, NPS practices do not rebel against professionally established standards for subtitling. The focus of scholarship on non-professional subtitling has gradually transformed from the dichotomy of professional and non-professional, the excellent quality vs. the poor, to focusing more on the latter’s characteristics and even the blurring boundary between the two.

More specifically, Jiménez-Crespo (2017) summarizes in his book, two of the seven most common “deviations” from professional norms identified in fansubbing, including: (1) more creative and individual renditions of source texts and (2) different approaches to taboos and improper language. Khoshsaligheh, Mehdizadkhani, and Ameri’s study (Khoshsaligheh et al. 2016) demonstrates these “deviations”. By looking at the amateur AVT, or fansubbing and fandubbing, in the Iranian context, the study selects ten American movies to examine how taboo language is rendered into Persian. The paper concludes that fansubbers tend to keep taboo language in the subtitles, deviating from Iran’s current official translation norms.

2.3. Translation Activities in Bilibili.com

Many studies look at various sites that contain enormous NPS activities, such as Viki.com, Facebook and Twitter, Ted.com, and Bilibili.com. According to QuestMobile’s research (China Tech Express 2020), Bilibili.com, or B Site (B站, B Zhan) in short, has been the most popular application among Chinese young people under 24 years old in 2020, whose monthly active users reached 223 million in the first quarter of 2021 (Bilibili 2021). The B Site allows users to upload videos on various themes with or without subtitles as a video-sharing platform. Since 2018, viewers have been allowed to contribute to any subtitles when approved by their uploaders (Bilibili 2018). Another highlight that makes this site famous among young people is danmu or danmaku in Japanese. This practice displays users’ real-time comments directly on the videos, usually cutting through from right to left. These comments are anonymous but are regulated and sometimes censored, following the site’s guidelines. One must pass a test (Bilibili 2020) to register for an advanced (or VIP) account to post danmu, where users are asked about the details of the guidelines, such as whether viewers can post personal information and whether one should post swearwords in danmu, etc. According to the guidelines, viewers should not post spoilers or private information and should only post positive content (compliments or some funny comments) on danmu. In addition, the guidelines explicitly state that users cannot post anything related to sexual references, illegal content, etc., and should refrain from using vulgar language.

Nevertheless, users still find ways to “squeeze” taboo language into the videos, either in subtitles, danmu comments, or general comments, where viewers usually comment on the content of the video, which includes the translation style and quality.

There is some literature on fansubbing and danmu comments using Bilibili.com as a research site. For example, Zhang and Cassany (2020) examine danmu from a discourse analytical perspective; Yin and Fung (2017) argue that Bilibili has become a space for democratic discussion; and Yang’s research (Yang 2021a) on danmu subtitling also uses Bilibili.com as the site of study. But her focus was only on the danmu subtitling in untranslated English videos. By investigating some untranslated BBC documentaries on the website, her study reveals the essential characteristics of danmu subtitling. It describes this new translation phenomenon as enabling knowledge sharing and amateur-led linguistic collaboration. Although one of the research goals in Yang’s most recent study (Yang 2021b) was to broaden the scope of non-professional translation studies, she mainly investigates danmu comments posted in untranslated English documentaries through a social semiotic lens.

Although this project also looks at danmu, the selected 18 English videos are already translated into Chinese when uploaded to the website. The main goals are to examine how the taboo languages are rendered by the uploaders or the non-professional subtitlers who help the uploaders, and how danmu contributors and video viewers react to the subtitlers’ renderings of the taboo languages. In other words, the study examines the translation activities and interactions on translation, language, and cultural issues related to taboo language on a popular video-sharing platform, Bilibili.com. So far, very little research has been done on the abovementioned intersection.

2.4. Subtitling Strategies and Techniques and Effects

When analyzing translation activities in an NPS setting, the study adopts and modifies some mainstream subtitling strategies and techniques of culture-specific references, as taboo language is considered part of a culture. There is no widely accepted taxonomy specifically for taboo language. Two of the most commonly accepted taxonomies come from Díaz-Cintas and Remael (2007, 2021) and Pedersen (2011). The former includes loan, calque, explicitation, substitution, transposition, lexical recreation, compensation, omission, and addition. The latter has seven baseline strategies: retention, specification, direct translation, generalization, substitution, omission, and official equivalence. These taxonomies are mainly for professional subtitling, including DVD production and TV broadcasting. Although some categories, such as omission and direct translation, work for NPS, some highly creative strategies that appeared only in NPS are not included in the existing taxonomies. This study attempts to modify the aforementioned taxonomies to describe NPS practices more accurately.

In addition, the study analyzes the expected effects received after taboo language is rendered, or “techniques” defined by Ávila-Cabrera as “the result of a translation” (Ávila-Cabrera 2020, p. 129). The instances in this study demonstrated three general effects: toning down, toning up, and maintaining. The study simplifies Ávila-Cabrera’s taxonomy (Ávila-Cabrera 2020) based on the characteristics of the instances.

More specifically, either partial or complete omission of the taboo language in the target text (TT) tones down its effect. Sometimes, the foreign word or expression is retained in the TT for several reasons. The subtitler may find it difficult or unnecessary to find an equivalence in the target culture and thus retains the source expression. Due to the various levels of English proficiency, it is likely that the viewers cannot understand the retained version. The load of the taboo language is thus reduced.

When euphemism (or substitution by euphemism) or neologism (or lexical recreation in the TT) is used in the Chinese context, the strength of the taboo language in the source text (ST) is often reduced. As Jing-Schmidt and Hsieh argue in “Cursing, Taboo, and Euphemism”, a section from The Routledge Handbook of Chinese Applied Linguistics (Jing-Schmidt and Hsieh 2019), the euphemistic strategy “remains popular in modern Chinese where high-frequency taboo items are graphically and semantically disguised” (p. 401). Some of the examples provided by the authors also frequently appear in our study, such as 屄 bi ‘cunt’ as 逼 bi ‘to force’, 肏 cao ‘fuck’ as 操 cao ‘to wield’(p. 401), and the onomatopoeia 啪啪啪 papapa as euphemisms for sexual intercourse (p. 400).

Jing-Schmidt also provides a comprehensive overview of Chinese neologism in this edited volume and the “Linguistic Studies of Chinese Internet Neologisms” section is highly relevant to this study. According to Jing-Schmidt, the five most notable morphological processes include compounding, derivative information, homophony, loanwords from Japanese, English, and Korean languages, and alphabetic words that are Chinese words written fully or partially in Roman letters (pp. 519–20), and many renderings in this study fall into these categories, such as the pinyin acronym TM.

Similar to the effects of euphemism, these Chinese neologisms also tone down the load of the taboo language. As argued by Jing-Schmidt, Chinese millennials, who happen to be the primary users of Bilibili.com, embrace these neologisms for their concise, creative, playful, fashionable, and in some cases, satiric potential (pp. 519–20). For this study, one more function of euphemism and neologism is to avoid potential violations against the guidelines of the video-sharing platform or the rules and regulations of the target culture, which usually leads to being censored or upload failures.

The only strategy observed in this study that is likely to maintain the effects of the taboo language is direct translation, where all features of the ST are kept in the TT.

It is worth noting that the strategy addition proposed here is different from the one proposed by Díaz-Cintas and Remael (2007, 2021). Here, it refers explicitly to the instances that do not have taboo language in the ST, but subtitlers add it to the TT. This is the only strategy that may tone up the load of the ST.

Table 1. Subtitling strategies, definition, and expected effects.

No.	Strategies	Definition	Expected Effects
1	Partial omission	Parts of taboo language are omitted in the TT.	Toning down
2	Complete omission	No taboo language is omitted in the TT.	Toning down
3	Retention	The foreign word or phrase is retained in the TT.	Toning down
4	Lexical recreation in TT	Neologism is observed in the TT.	Toning down
5	Direct translation	All features of the ST are kept in the TT.	Maintaining
6	Substitution by euphemism	Taboo language is substituted by euphemism in the TT.	Toning down
7	Addition	Taboo language is not observed in the ST but is added to the TT.	Toning up

lists the subtitling strategies observed in the instances of this study and their expected effects:

Overlapping or missing instances are reduced to the minimum but still possible. As Díaz-Cintas and Remael (2021) note, overlapping in these classifications is inevitable, and “no existing classification can cover all the translation strategies” (p. 207).

This taxonomy mainly represents the scope of this study, whose focus is NPS practices in China and requires major or minor modifications when applying to other settings or scenarios.

3. Materials & Methods

Triangulation, defined as “a mix of procedures to grasp complex phenomena” by Hansen (2010, p. 207), is implemented in this study by adopting a mixed method that combines quantitative with qualitative data (Robson 2002). The quantitative data are from a parallel corpus consisting of taboo language and its translations from 18 of the most-viewed and commented upon subtitled videos on one of China’s most popular video-sharing platforms, Bilibili.com. The corpus is mostly unidirectional and bilingual (from English to Chinese), with many cases of non-Chinese characters in the Chinese translations. In addition to the corpus, relevant danmu comments and general comments related to the taboo language translations are also collected and studied as the qualitative data.

More specifically, the study selects videos based on the following criteria:

• The original language of the video is English.
• The subtitles are either in Chinese only or in Chinese and English.
• The video has more than one taboo language.
• The view count is more than one million.
• The number of danmu comments is more than 1000.

As listed in Appendix A, most of the videos are clips from stand-up comedies and late-night TV shows, and it is no coincidence that these kinds of genres usually contain more taboo language than others. The language instances were manually collected due to the various sort of taboo language and its strong connection with multiple contexts. Thus, methods such as keyword searches are not feasible.

By manually checking the subtitles, danmu comments, and general comments, 66 examples of taboo language and comments that discussed the translations of taboo language are collected, mainly based on Ávila-Cabrera’s taxonomy (Ávila-Cabrera 2016). While Ávila-Cabrera distinguishes between offensive and taboo language, there is no such distinction in this study, as the term taboo is used instead. Eight categories of taboo words are identified, while instances of the rest of the five groups are not found in the corpus. One instance, 新guan病毒 (the COVID virus), may or may not be a taboo word and thus it is classified as a new category “other” in this study. It is possible that the subtitler believed that this term was sensitive and thus used pinyin to replace the correct Chinese characters. Another possibility was that the uploader and subtitler, who is a non-native Chinese speaker, was not certain of the correct characters and thus decided to leave the pinyin version in the subtitle. There is no evidence for either possibility in the danmu comments and general comments. In addition to the taboo language identified in the ST, the study also collects taboo language that does not exist in the ST but is added to the subtitles, danmu comments, or general comments.

Table 2. Categories, numbers, and examples of taboo language.

Category	Number (Source)	Number (Added)	Example
Abusive swearwords	26	3	Fuck the police
Expletives	9	2	Damn!
Profane/blasphemous		0
Animal name terms		0
Ethnic/racial/gender slurs	7	1	nigger
Sexual/body part references	11	1	Barack Obama giving the sex talk
Urination/scatology	1	1	pee
Filth		0
Drugs/excessive alcohol consumption	1	0	Smoking weed
Violence		0
Death/killing		0
Invectives	0	2	我这副德性 (de xing)
Other	0	1	新guan病毒

summarizes the number of taboo language instances collected in each category and provides examples.

The data is compiled into four categories: ST, subtitle, danmu comments, and general comments using an Excel file.

Table 3. An instance of corpus and comments.

	ST	TT	Danmu Comments	General Comments
1	Smoking weed	吸DM	Weed = 杂草	抽DM(暗物质) 可还行
			说杂草的认真的吗哈哈哈	突然让我想起来老友记里甘瑟抽钱德勒那跟烟的台词
			Da ma可还行hhh	特朗普:没有人比我更懂DM！吸溜～吸溜～吸溜

is an instance from Video 1 (Appendix A).

It is worth mentioning that not every instance has relevant danmu comments or general comments, but almost all the cases have the former.

After collecting the data, each instance with its danmu comments or general comments (if any) is analyzed, subtitling strategies and techniques, and effects using the modified taxonomy based on the ones proposed by Díaz-Cintas and Remael (2007, 2021), Pedersen (2011), and Ávila-Cabrera (2017, 2020) are identified, and translation activities that occurred in an NPS setting are examined.

This study attempts to answer the following two research questions:

(1)

Taboo language is usually toned down in professional subtitling. Given the creative and individualistic nature, what are some of the most used subtitling strategies and techniques in rendering taboo language in non-professional settings?

(2)

How do subtitle viewers react to the renderings of taboo language provided by non-professional subtitlers? How do the interactions between danmu and general comments contributors potentially affect the translation activities and language changes?

To address these questions, the study analyzes the translation strategies and their expected effects of the cases in the parallel corpus, based on the previously mentioned taxonomies after minor adaptations. Relevant interactions regarding the renderings of taboo language are also examined.

4. Results and Discussion

To answer the first research question, 66 instances that contain taboo language are drawn from 18 of the most viewed videos on Bilibili.com. Based on the comments contributed by the viewers and the author’s analysis, two instances are considered mistranslations due to subtitlers’ overly literal translation or misinterpretation, accounting for about 3% of the total cases. For the rest of the 64 instances, as shown in

Table 4. Results I.

No.	Strategies	Expected Effects	Number of Instances (in Percentage)	Example
1	Partial omission	Toning down	3 (4.7%)	ST: fuck TT: 你X的 BT: fXk
2	Complete omission	Toning down	8 (12.5%)	ST: What the hell is laser? TT: 镭射是什么？ BT: What is laser?
3	Retention	Toning down	2 (3.1%)	ST: it’s made gay TT: gay息太浓了
4	Lexical recreation in TT	Toning down	15 (23.4%)	ST: What the hell! TT: TM的
5	Direct translation	Maintaining	10 (15.6%)	ST: big butt TT: 大屁股 BT: big butt
6	Substitution by euphemism	Toning down	15 (23.4)	ST: sex TT: 啪啪啪 papapa
7	Addition	Toning up	11 (17.2%)	ST: bad fall TT: 摔个狗吃屎 BT: fall like a dog eating shit

, two of the most used subtitling strategies are lexical recreation in the TT and substitution by euphemism, which account for 23.4%, respectively. The strength of the taboo language is reduced in about 67.2% of instances by using various strategies such as partial (4.7%) and complete (12.5%) omission, and retention (3.1%), in addition to the two main ones, as

Table 5. Results II.

No.	Expected Effects	Number of Instances (in Percentage)	Example
1	Toning down	43 (67.2)	ST: fuck TT: 你X的 BT: fXk
5	Maintaining	10 (15.6%)	ST: big butt TT: 大屁股 BT: big butt
7	Toning up	11 (17.2%)	ST: bad fall TT: 摔个狗吃屎 BT: fall like a dog eating shit

illustrates.

The features of 15.6% of the ST are kept in the TT, thus maintaining similar effects of the taboo language. The result also indicates that taboo language is added to the TT in up to 17.2% of instances, while no such expression is observed in the ST. Examples and some back translations (BT) are also provided for better illustration for each category.

In terms of the second research question, active interactions on translation issues are observed in 15 of the 18 selected videos, providing various textual examples for better understanding of non-professional, collaborative, user-generated translation activities in major video-sharing platforms. The comments can be loosely organized into three categories: positive feedback, neutral feedback, and constructive criticism.

4.1. Positive Feedback

Positive feedback refers to comments that compliment the high quality of the subtitles or subtitlers’ creativity. In the 21 loosely organized threads of interactions, seven of them are positive comments regarding relevant taboo language.

Compliments can be more specific in some cases than others, as one comment describes the overall translation style as “very vivid and flexible” (这个翻译好灵气), while some only briefly comment that “the translation is so good” (翻译得太好了).

Usually, in those more specific comments, more playful language is observed. An American uploader with over 500 thousand followers on the B Site, Meiguo Laozhang (ST: 美国老张; TT: American Old Zhang), posted Video 6 (Appendix A) expressing his disappointment and despair because he could not go back to China to see his family from the United States due to the pandemic. In his video, he used the taboo expression “what the fuck” and interestingly, in the video that he made and subtitled, he beeped himself when saying that expression and this is what the viewers see on the screen (Figure 1).

The screenshot shows that the English subtitle is “What the beep”, while the Chinese subtitle uses a commonly used pinyin acronym TM to refer to the taboo expression. In the meantime, an enlarged WTF is shown on the screen. An enormous number of comments talk about this. Many of them use playful or ridiculing language such as “hahaha, wtf sounds a little cute” (ST: 哈哈哈wtf有点可爱) or “hahaha, I finally understand an English word” (ST: 哈哈哈我终于看懂了一个英语单词). Some provide another creative rendering, such as 王德发 (wang de fa). The Chinese characters read like the taboo language, which is considered a lexical recreation strategy in rendering taboo language.

Some positive feedback is brief and yet illustrates a solid foundation of language and cultural knowledge. In Video 2 (Appendix A), Jo Koy joked about the mysterious amount of water being added to make perfect rice, and he used a pun as shown below:

ST: All you need is rice, a pot, and this fucking line right here.

TT: 你所需要的就是大米、一个锅、以及中指上的这条线

BT: All you need is rice, a pot, and this line on your middle finger.

Nine danmu comments are made right after this punchline, and five of them compliment this brilliant rendering:

Danmu 1: 中指好评

Danmu 1 BT: Great job on the “middle finger”

Some feedback is even more in-depth and inspiring. In Michael Che’s video (Video 12, Appendix A), he talked about racism:

ST: We all need Asian people because they make shit affordable.

TT: 我们都需要亚洲人，因为他们的东西物美价廉

BT: We all need Asian people because their stuff is good quality and reasonably priced.

Here, the four Chinese characters in the TT rendered from the taboo language make a common expression in the Chinese language. Many danmu comments respond to this rendering, and one of them specifically points out that “this translation indeed avoids a war” (ST: 这翻译真是避免了一场战争); another comment makes a connection between “affordable” and the word “cheap” and points out that the actor used “affordable” instead of “cheap”, showing that he’s being meticulous”. (ST: 用affordable不是cheap很小心)

The first viewer is aware of the potential racial slurs if rendered inappropriately, and the other can tell the nuanced difference between “affordable” and “cheap”, demonstrating their knowledge of the two languages.

One crucial fact to note is that in general comments, one way to show support to uploaders and possibly also subtitlers is called “一键三连” (yi jian san lian ‘one key three connections’; see Figure 2), referring to doing all three things consecutively, including “like the video” (点赞), “monetary support” (投币), and “save as favorite” (收藏).

In this particular comment, in addition to an emoji that resembles a cartoon bear, the contributor specifically said they did the three things in a row because of the excellent quality of the subtitle translation. Uploaders are usually motivated to produce videos that receive popularity and based on this comment it seems that good quality of translation is one crucial indicator.

Thus, a quasi-peer review system is formed, where viewers evaluate the translation quality, pick up potential errors, and provide feedback to uploaders and subtitlers via danmu comments or general comments on this video-sharing platform. Positive feedback matters to the uploaders as it will bring more views and potentially more positive feedback. Thus, it can be inferred that the uploaders will take comments seriously and improve their translation/subtitles accordingly.

Another way to support NPS providers is also observed in the comments, which usually happens when specific comments complain about the arguably poor quality of the subtitle translation because they do not understand many punchlines (ST: 老实说，这一版翻译不行，好多梗没翻译出来, BT: Honestly speaking, this translation version is no good. Many punchlines are not translated). Some viewers would defend the subtitler and respond harshly and impolitely by saying, “this is a decent subtitle translation. If you don’t understand, it’s probably you are not using your eyes”. (ST: 字幕做的很好，说没看懂估计是没用眼睛看)

4.2. Neutral Feedback

Almost half of the comments are neutral feedback, which refers to no negative or positive opinions towards the rendering of taboo language in the subtitles. In some neutral feedback, contributors provide linguistic knowledge, cultural background, and associations to help other viewers better understand the taboo language and its contexts. Sometimes this leads to interesting discussions, creating a language and culture learning platform.

For the only drug reference in this study, when “smoking weed” in the ST is rendered as a pinyin acronym “DM” (see Figure 3), some made a connection to the word “dark matter”, which shares the acronym “DM”; some mentioned a scene from an American TV series Friends that is relevant to smoking (ST: 突然让我想起来老友记里甘瑟抽钱德勒那跟烟的台词, BT: this suddenly reminds me of Gunther’s line when he takes a drag off of Chandler’s cigarette in Friends).

Take a video that focuses on one taboo language as another example. In this video clip, Conan O’Brien interviewed Quentin Tarantino on Late Night, and he mentioned a Chinese word that means “super cool”, which is one of the most used Chinese abusive swearwords 牛逼 (niu bi). When Quentin explains to Conan what that means, he says niu bi means cow’s vagina, which is correct, and in Chinese, the word vagina is euphemized to the word 逼 (bi), which means to force literally.

Since this less than two-minute clip is centered around this taboo language and nothing else, many comments in both places also discuss this taboo language. Topics include various creative ways to render this and other similar, also commonly used taboo language (such as 屌 diao as euphemisms for male genitalia), and different schools of the etymology of this Chinese expression. The enormous number of creative renderings, including neologism and euphemism, provide rich resources, information, and knowledge for language, culture, and media learners and scholars alike.

When users participate in this kind of collaborative translation and language learning activities, knowledge sharing goes beyond the scope of translation. Linguistic knowledge is observed in comments explaining the two primary meanings of the word gay and providing background information, while explaining the pun intended. In Video 11 (Appendix A), a kid cannot distinguish between “uranus” and “your anals”, leading to knowledge sharing on astronomy and Roman mythology in danmu and general comments. When Ronny Chieng complains about the dry taste of turkeys in Video 17 (Appendix A), he says, “Fuck the turkeys! It’s dry”. And the TT 去他妈的 (qu ta ma de) is vulgar and offensive slang in Chinese, which maintains a similar strength to the taboo language fuck. While pork, beef, and lamb remain staples for Chinese consumers, turkey is not popular. Thus, in some danmu comments (see Figure 4), viewers firstly doubt “if turkeys are really yucky” (ST: 火鸡这么难吃么?) and then other viewers “seconded” (ST: 附议) or explain the taste of it and introduce proper ways to eat it (ST: 火鸡肉又干又柴, 只能炖汤, BT: Turkey is dry and chewy and can only be used to make soup).

A racial slur ching chong chang is used in another racism joke. While the translation (中国佬 zhong guo lao) renders a similar discriminating tone, some people do not understand what this means. Thus, the background knowledge of racism and relevant culture is provided by some viewers in the general comments.

Apart from providing background knowledge, some other comments provide more creative renderings without criticizing the quality of the translation. This kind of behavior is often much appreciated by other viewers and there are two such instances. In Video 1 (Appendix A), Trever Noah imitates the then-President Obama in a sex talk with his kids and says, “When a man loves a woman, he wants to stimulate her economy”, while the subtitle adopts a relatively conservative strategy, direct translation, and renders the second half of the ST into “他想刺激她的经济” (BT: he wants to stimulate her economy). One danmu comment chooses a bolder approach and renders it into “刺鸡经济” (ci ji jing ji). Here, 刺激 ci ji and 刺鸡 ci ji are homonyms, but the character 鸡, which means chicken, is an informal way of saying “prostitute”. By replacing a non-taboo expression with its homonym that has strong sexual references, addition is used as a subtitling strategy.

Another instance concerns some parts of the taboo language in the ST “Donald J Mother-fucking Trump”—fucking is partially omitted in both the source and the target subtitles, and in the TT “唐纳德 J mother-fxxxking trouble” (BT: Donald J mother-fxxxking trouble), the retention strategy is adopted for parts of the ST rendered as “唐纳德 J mother-fxxxking trouble”. The partial omission and the retention strategies bring confusion to some viewers. Thus, some ask if someone can provide a more straightforward and more accurate rendering in the general comments, followed by a rendering “唐纳德.我双倍的脏你妈” (BT: Donald. I double fuck your mother), where lexical creation strategy is adopted.

As the results show, about 16% of the taboo language is partially or entirely omitted in the TT. Based on the observations, not many viewers specifically comment on these strategies unless the omitted parts cause confusion or misunderstanding. For example, in a video clip of Michael Che’s stand-up comedy show, he talks about converting to Muslim only if two things are not in the way: bacon and titties, and the partial omission strategy is adopted when rendering the latter into Chinese (see Figure 5).

ST: (I actually liked the religion). I would convert if it would not for the policies on titties and bacon.

TT: 要不是因为他们对奶*和培根的特殊规定，我都想加入他们了.

BT: I would convert if it would not for the ti** and bacon policies.

Danmu 1: 奶油含有猪油

Danmu 1 BT: Cream has lard in it.

Danmu 2: Titty.

Danmu 3: 前面的不是奶油好吗，是那个奶

Danmu 3 BT: It’s not cream, okay? It’s boobs.

Confusion does not only happen when omission strategies are used. Back to the “smoking weed” example, a pinyin acronym is used to render this expression into Chinese. This lexical recreation strategy not only tones down the strength of the taboo language but also confuses some. Again, other viewers explain what this is either by providing the full pinyin of that word or posting a directly translated version:

Danmu 1: 草（植物）

Danmu 1 BT: Grass (a plant)

Danmu 2: 说杂草的认真的吗哈哈哈哈

Danmu 2 BT: You said grass? Are you being serious, hahaha

Danmu 3: 特朗普开始抽大麻了

Danmu 3 BT: Trump is smoking weed.

Danmu 4: weed = dama

Similar to professional subtitling, partial and complete omission strategies are also quite common in NPS due to the spatial–temporal constraints. Lexical recreation strategy, which produces neologisms such as new acronyms, is also often adopted to follow corresponding guidelines and rules, make playful and fashionable comments, or save space for the NPS practices.

It is unavoidable for these kinds of strategies to confuse viewers who do not have enough knowledge of the source language and culture. Unlike professional settings, viewers still have a chance to have their questions answered or clarify confusion via danmu comments and general comments.

4.3. Constructive Criticism

About 19% of the threads focus on renderings that the viewers deem inappropriate or wrong, and thus, heated debates usually ensue. This is considered constructive criticism in this study.

The only two mistranslations in this study are pointed out in the comments, where viewers propose renderings that they consider correct or more appropriate. For example, in a video clip of Joe Wang’s talk show at the White House (Video 4, Appendix A), he commented on the then-President Obama—“President Obama has always been accused of being too soft. But he was conducting two wars. And they still gave him the Nobel Peace Prize. And he accepted it”. Then he says the following sentence:

ST: You can’t be more badass than that.

TT: 还有比这更操蛋的事情么？！

BT: Can it be more fucked up?!

Here the subtitler seems to over-interpret the meaning of “badass”, which should be understood as the opposite of being too soft. Some danmu contributors identify this issue (see Figure 6).

Danmu 1: 翻译错了，badass应该翻译成还有比他更狠的人

Danmu 1 BT: The translation is wrong. Badass should refer to someone tough.

Danmu 2: Badass有厉害的意思

Danmu 2 BT: Badass can mean fierce.

Moreover, badass in the ST is not considered in English as taboo language. Still, the subtitler uses the addition strategy to add taboo language in the TT, regardless of appropriateness.

With the danmu and general comments features, the platform provides a collaborative virtual discussion board to learn language, cultures, and beyond. It is, however, not always clear enough if the subtitler is making a mistake or not, and thus rigorous discussions happen in the threads of the comments. In Video 3 (Appendix A), Jimmy Fallon interviews Trevor Noah on The Tonight Show and when Fallon asks Noah how he can keep the energy, Noah says the following sentence that leads to controversial opinions in the comments:

ST: It helps you if you don’t have a family, Jimmy!

TT: 可能单身的人可以节省一些“精”力吧

BT: Maybe single people can have more “jing” li.

While jing li in Chinese means “energy” in general, using quotation marks around jing implies a sexual reference, as jing also means sperm (精子 jing zi). Some viewers compliment this rendering by posting comments such as “the translation is perfect!” and “this translation is accurate, fluent, and elegant”. But others believe that the ST simply means having a family can be distracting and exhausting, which does not have a sexual connotation. This open-ended discussion provides excellent language, culture, and communication learning, and translators’ training resources.

5. Conclusions

One of the study’s research questions is to examine subtitling strategies and techniques used in rendering taboo language in NPS settings. According to the data collected, the study confirms the findings of some scholars that a variety of creative approaches are being used. The most common ones are lexical recreation in TT (23.4%) and substitution by euphemism (23.4%), while partial and complete omission account for a smaller percentage of 17.2%. Although the load of the taboo language is reduced in more than 67% of the instances, the study also observes an unexpected 17.2% of the taboo language that is missing in the ST but added in the TT, strengthening the effects of the taboo language.

NPS practices have long being accused of lacking guidelines or standards, thus the translation quality of NPS has been inaccurately depicted as substandard and unsatisfactory. Instances with mistranslation account for 3% in this study, which is an arguably adequate amount in professional subtitling. There are many reasons for a relatively low number of translation errors. Based on the analysis, it is noted that the viewers, who are usually young and multilingual, are familiar with the content and multicultural elements. Thus, they can identify the mistakes in the subtitles, as proven by a significant number of comments.

Based on this study, it is found that when the subtitler makes a mistake, which is rare in this study, the viewers identify the subtitler’s errors in danmu and general comments. Since uploaders are usually motivated to produce videos that receive popularity, they typically take the comments contributed by their viewers very seriously. Thus, a quasi-peer review system is formed. Viewers evaluate the translation quality, pick up potential errors, and provide feedback for uploaders and subtitlers via danmu comments or general comments on this video-sharing platform.

The Bilibili.com platform provides a virtuous, collaborative mechanism for potential translation problems and language learning purposes, which is also demonstrated in some comments that provide more creative, less conservative renderings, without criticizing the quality of the translation, and help viewers clear up confusion and answer questions regarding specific taboo language.

Contributors not only provide linguistic knowledge, cultural background, and associations to help other viewers better understand the taboo language and its contexts, but they also share what they know beyond the scope of translation, such as knowledge in sociology and astronomy.

According to the study, it is found that subtitle viewers in NPS settings usually react positively in their comments, where they show interest in translation, language, and culture; sometimes they ask questions, which are typically answered by other peer viewers using their solid knowledge, skills, and expertise. The constructive criticism, which accounts for about 19% of the comments, mainly focuses on the inappropriate renderings that they consider mistakes, rather than criticizing subtitlers’ translation style and quality.

The study concludes that social conventions, community platform guidelines, and “unspoken rules” are important factors to consider in NPS when taboo language is rendered. These so-called constraints, on some level, encourage and produce more creative and non-conventional approaches (which would not be allowed in a strictly professional setting) in rendering taboo language.

It is also undeniable that some subtitlers and viewers may have received professional training and gained various levels of knowledge on languages, cultures, and translation, blurring the boundaries between professional subtitling and NPS and calling for a redefinition of NPS, that is still associated with poor quality and lack of supervision by some.

5.1. Significance and Future Directions

By examining various creative subtitling strategies and techniques used in NPS settings and analyzing the viewer-contributed comments, this qualitative and quantitative study helps understand current non-professional, collaborative, user-generated translation activities on a video-sharing platform. The conclusion echos with previous research that emphasizes the creative and individualistic nature of NPS (Jiménez-Crespo 2017) and calls for more attention to examine the blurry boundaries between professional subtitling and NPS. This study also contributes to translators’ training and reception studies in AVT. The subtitles and comments themselves are valuable resources for translators’ trainers and novel translators. Since the comment contributors are also viewers, this study looks at how taboo language renderings are received by some viewers, especially those with more relevant knowledge; as Chiaro (2009) argues, reception studies are essential in AVT because audiences from both source and target cultures should have similar audiovisual experiences.

5.2. Limitation

The scope of the study is limited as only 18 videos are selected, and most of the genres are limited to stand-up comedy shows. A comparative study would be ideal for comparing the differences between professional subtitling and NPS in an in-depth manner. However, since the selected videos are not seen with professional subtitles, such as on state-owned TV broadcasting channels, it is not feasible to conduct such a study with the selected videos that contain taboo language.

In addition, although the theoretical framework and relevant taxonomies help reduce the subjectivity of the study, only comprehensive reception studies can better solve this potential problem. Thus, the study calls for more comprehensively conducted reception studies to reduce subjectivity.