Translation errors as tools for quality assessment of subtitles

This study concentrates on assessing subtitling quality through error analysis. For the purposes of the present study, two types of errors have been distinguished. These are

translation errors and subtitling errors. Translation errors, according to Hansen (2010) and Vehmas-Lehto (2005), arise from the existence of a relationship between the source text and the target text. Translation errors usually occur because something goes wrong when the source text is transferred into the target language. Translation errors can be, for instance, pragmatic, semantic or linguistic, and they can be caused by various reasons, such as

misunderstandings of the content of the source text, not rendering the meaning of the source text accurately enough or differences in the cultural backgrounds of the source text and the target text. In the scope of this study, translation errors are classified according to Hansen’s

28 (2010) and Vehmas-Lehto’s (2005) definitions. Subtitling errors, in turn, refer to deviations from the technical conventions of subtitling discussed in Chapter 2.1.

The concept of a translation error has previously been utilised in the context of literary and journalistic texts (see for example the aforementioned Vehmas-Lehto 1989; 2005) and technical texts (see for example Schmitt 2002) but the concept is fairly unused in the context of subtitling. Still, Abdallah (2007) and Tuominen (2013) share a similar view that subtitling is usually surrounded by critical discussion which can be noticed, for instance, in the casual subtitle-related conversations by viewers. While the studies discussed in Chapter 3.3. indicate that viewers’ attitudes towards the quality of subtitles is positive, the “real-life”

conversations have a much more critical tone (Tuominen 2013). To provide a comparative perspective to focus group discussions, Tuominen (2013: 302–309) looked at some

spontaneous comments on subtitles on the internet. Tuominen (2013: 308) found that the comments on the internet express a more negative attitude towards subtitles and imply that viewers pay more attention to translation errors than the focus groups did. The discussions on the internet often revolve around the shortcomings and errors the viewers have noticed in subtitles and sometimes they even lead to questioning the translators’ professionalism and intelligence (Tuominen 2013: 305; Abdallah 2007: 276). Alternatively, translation errors in subtitles can cause amusement in viewers and, as Hansen (2010: 386) points out, in some contexts errors are accepted and regarded as “fun”. This view applies well to subtitles where translation errors and typos are at times considered more entertaining than serious. For example, Paakkinen, mentioned at the end of the previous chapter, has collected numerous translation errors into his books Yhdeksänmetrinen maasika – Käännöskukkasten parhaita (2003) and Agentti Appelsiini – Käännöskukkasten parhaita 2 (2005) and has created a whole website devoted to amusing mistranslations collected from Finnish subtitles (Jouni Paakkinen:

http://www.jounipaakkinen.fi/kaannos.html). However, the viewers seem to be quite up to date about the translators’ working conditions (for more see Chapter 2.3.). Tuominen (2013:

305) noticed that the commenters speculated whether tight timetables, lack of decent dialogue lists or other factors prevent translators from producing acceptable quality. Thus, while criticism towards subtitles if often harsh, the viewers also express sympathy towards the translators and indicate that they are aware of the factors which might make it difficult for translators to produce good quality (Tuominen 2013: 305–306).

Hansen (2010: 386) reports that definitions and categorisations of errors have originally been created for the assessment of translation quality depending on different purposes and situations. Some researchers (see for example Nord 1999) have presented a

29 categorisation and gradation of errors for translator training, while others (see for example Mertin 2006) have created criteria for the gradation of errors in professional translations.

Vehmas-Lehto (2005), alternatively, offers a thorough analysis of different types of

translation errors and categorises them according to different aspects of translations. Vehmas-Lehto (2005: 52) bases her discussion of translation errors on the assumption that the aim is to create a printable translation which corresponds as well as possible to the reader’s

expectations and needs. Vehmas-Lehto (2005: 52) argues that this kind of translation should not present any problems in understanding the text for the reader and has a natural linguistic form. In the scope of this study, the most relevant aspects are the severity of errors and the obviousness of errors. On one hand, when assessing the severity of errors, the evaluator analyses how severely the translation errors degrade the quality of subtitles. On the other hand, when assessing the obviousness of errors, the evaluator analyses how easy it is to notice the translation errors from the text. On the basis of these two aspects, Vehmas-Lehto (2005:

53; 63–69) divides translation errors into two categories which are overt errors (in Finnish

“avovirheet”) and covert errors (in Finnish “piilovirheet”). Again, Vehmas-Lehto’s (2005) definitions are based on House’s (1997) model.

Firstly, Vehmas-Lehto (2005: 64; 1989: 27–28) defines overt errors as any deviations from the content of the source text as well as any violations against the system and norms of the target language. Thus, overt errors can be further divided into two subcategories. These are semantic errors and linguistic errors. Semantic errors, sometimes also called

communication errors, can be caused by the translator’s misunderstanding or

misinterpretation of the content of the source text. In other words, semantic errors change the original meaning of the source text. In the scope of subtitling however, semantic errors do not include only mistranslations but also failures to convey the atmosphere and style of the source text. As was discussed in Chapter 2.1., subtitles always represent someone’s speech and this is why they should not be “stripped too bare” (Vertanen 2007: 153). While it is clear that

subtitles cannot translate everything word for word, unnecessary omissions and excessive condensation can result in watered-down translations that do not convey the tone or

characteristics of the speaker. According to Vehmas-Lehto (2005: 53), semantic errors can be considered to be the most severe types of errors since they change the original meaning of the source text or, in the case of subtitling, fail to convey the style or atmosphere of the source text to viewers. Linguistic errors, in turn, violate the system and norms of the target language and the target language usage. Vehmas-Lehto (2005: 53, 64) notes that there are various types of linguistic errors, such as lexical errors, grammatical errors, textual errors and

30 pragmatic errors. Lexical errors refer to unsuitable word choices, whereas grammatical errors (solecisms) are issues in the grammar of the target language, such as incorrect noun and verb inflections or errors in clause and sentence structures. Textual errors then relate to

shortcomings in text structure and cohesion, while pragmatic errors relate to the ignorance of the function of the text and the expectations of the viewer. Vehmas-Lehto (2005: 53) regards linguistic errors less severe than semantic errors, but notes that they can make the text difficult to understand or simply annoy readers. Vertanen (2007: 155) also emphasises that grammatically incorrect language usage might affect language learning. Vertanen (2007: 155) reports that teachers of Finnish find it a difficult task to teach pupils, for instance, the correct use of punctuation marks when subtitling does not seem to follow the same rules. Both types of errors, i.e. semantic and linguistic errors, can have an effect on the quality of the target text and are, additionally, rather easy to perceive from translations.

Secondly, as already discussed in Chapter 3.3., Vehmas-Lehto (2005: 64–69) defines covert errors as systemically and grammatically correct words or phrases but they violate the conventions of language usage in target culture. They are words and phrases a native speaker of the target language would rarely use. Thus, covert errors can make the text sound

unidiomatic and unnatural and affect the readability and fluency of the target text (Vehmas-Lehto 2005: 65). An example of a covert error is the use of premodifiers and postmodifiers when translating from English to Finnish. In texts originally written in Finnish, premodifiers are more common but when translating from English to Finnish, postmodifiers, especially relative clauses, appear more often in the translated text because the conventions and

structures of source language affect the target language. According to Vehmas-Lehto (2005:

65), in most cases covert error are caused by the interference of the source text. Vehmas-Lehto (2005: 64) argues that covert errors can be rather difficult to notice from translations and thus, they do not affect the quality of the translation as severely as semantic errors. It is common for the viewers to notice that something is not quite right in the text but they might be unable to explicitly explain what. Similar to linguistic errors, covert errors can sometimes make the text difficult to understand but in addition, they can also affect the fluency and readability of the target text (Vehmas-Lehto 2005: 65). Tuominen (2013: 279–280) agrees with Vehmas-Lehto by noting that covert errors can momentarily draw unnecessary attention to themselves and away from the audiovisual text as a whole as well. In her study, the focus group informants brought up some elements of the target text during the discussion, which could be classified as covert errors (Tuominen 2013: 162). An example of this kind of element is the word “lomanen” which is a creative translation for the English word

“mini-31 break”. As Tuominen (2013: 279–280) points out, the word itself is not a mistranslation or incorrect Finnish but it draws viewers’ attention by not being very familiar. At the same time, the one overt error that occurred in the subtitles went unnoticed among the informants

(Tuominen 2013: 166). The overt error was a mistranslation of the word “proposition” as

“kosinta” [proposal]. Thus, Tuominen (2013: 279–180) supports Vehmas-Lehto’s view that covert errors can “diminish the fluency and readability of the translated text without being actual mistakes” and affect the text’s reception.

An alternative categorisation of translation errors is presented by O’Brien (2012).

O’Brien’s (2012: 57–58) categorisation of translation errors is based on quality evaluation models that are in active use by companies operating in the translation industry. O’Brien (2012: 57–58) reports that most of the evaluation models used in the industry are error-based and examine translation quality by identifying errors, classifying them, allocating them to a severity level and applying penalty points. By comparing the different error-based quality evaluation models, O’Brien (2012: 60) found the common denominators, and divides translation errors into three categories accordingly. These are language errors, Desk-Top Publishing (DTP) errors and software user interface (UI) errors. DTP errors and UI errors are mostly to do with layout, formatting, graphics and alignment (O’Brien 2012: 61–62), and they could be applied to subtitling errors, that is errors referring to deviations from the technical conventions of subtitling. Language errors are specifically to do with the language and content of the translation. O’Brien (2012: 60) further divides language errors into four subcategories. These subcategories are language, terminology, accuracy and style (O’Brien 2012: 60–61). Language category includes grammar, syntax, punctuation and spelling of the target language. Some quality evaluation models also include typos, fluency and cultural references in the language category. Terminology category includes errors that are to do with the translator failing to comply with client-specific glossaries or industry-specific terminology or is not consistent in term usage. Accuracy category consists of unnecessary omissions, additions and inaccurate cross-references. Lastly, style category includes errors that arise from failures to adhere to client style guide. In the context of subtitling, this might mean the individual company guidelines which were discussed in Chapter 2.3. Criteria for style errors include tone, register, language variants, slang, literal translations and awkward syntax.

Similar to Vehmas-Lehto’s (2005) evaluation of the severity of errors, O’Brien (2012: 62) divides translation errors into three severity levels. These are minor, major and critical. Minor errors are errors that are noticeable but they do not have a negative impact on meaning. In other words, minor errors do not change the meaning of the source text in any

32 way and they do not confuse or mislead the reader. Major errors are considered to have a negative impact on meaning because they change the meaning of the source text. Major errors provide the reader false information. Critical errors, in turn, are considered to have “major effects not only on meaning, but on product usability, company liability, consumer health, safety and behaviour”. O’Brien (2012: 63) notes that the translation error-based quality evaluation models have a very low tolerance to errors. On average, three minor errors per thousand words and one major error per thousand words are accepted, after that the translation fails (O’Brien 2012: 58).

Both Hansen (2010) and Vehmas-Lehto (2005) emphasise that the translator’s theoretical approach to translation has to be taken into account when defining and evaluating errors as translation errors. For instance, in theories based on the concept of equivalence, a translation error is regarded as some kind of non-equivalence between source text and target text. All kinds of deviations from the theoretical approach can be defined as translation errors (Vehmas-Lehto 2005: 49). Based on the translator’s solutions the evaluator can speculate about the translator’s theoretical approach but it is quite impossible to know with certainty what the translator’s actual approach is. Furthermore, the evaluation of translation errors always depends to some extent on the evaluators’ subjective criteria (Vehmas-Lehto 2005:

51). This means that the evaluator and the translator might have a different view on what constitutes an acceptable translation, and the translator might feel that their work is unduly criticised (Vehmas-Lehto 2005: 51, Hansen 2010: 386). In the case of subtitling, the evaluator might come up with “better” translation solutions but the other case is whether or not the corrected version would fit the screen. At times, evaluators are keen to present more creative solutions but they are not always aware of the restrictions and conventions involved in the process of subtitling.

Additionally, who is doing the evaluation might have an effect on translation quality assessment as the study by Suokas (2014) suggests. Suokas (2014) compared the assessments by language experts and ‘ordinary’ users of a Finnish translation of an instructional text on guitar playing. An excerpt of the book was assessed by three language experts on the basis of usability heuristics (for more see Suojanen et al. 2012), while a group of users were asked to learn a guitar-playing technique with the help of the same text excerpt and comment on the instructions in an interview (Suokas 2014: 38–47). The results indicate that clumsy language and translationese bothered the language experts, while the ordinary readers mainly criticised the lay-out of the book as not being reader-friendly and found the language to be quite good (Suokas 2014). I agree with Jääskeläinen (2016) in that the study by Suokas (2014) suggests

33 that evaluators who are experts in the field of translation and language, such as translation teachers and translation students, might be more demanding and more critical towards the text than ordinary readers. In the scope of the present study, this could mean that I, as an

evaluator, might be harsher in classifying errors as I am an expert in the field of translation.

There is a possibility that ordinary viewers are happy with less but in order to acquire more insight on how ordinary viewers evaluate translations, reception studies should be conducted.

In the scope of this study, subtitles are expected to give a semantically adequate account of the original dialogue and convey the style and atmosphere of the source text while complying with the grammatical rules of Finnish. The translation errors will be perceived and evaluated based on this approach as this study mainly concentrates on the quality of the target language and the content of the subtitles. As was described in Chapter 2.1., the audiovisual environment of subtitling involves various constraining elements, such as the size of the screen and the timing of the programme, which might lead to condensation and omissions of information. Due to the constraining elements, condensation and omissions are warranted as long as they do not leave out any information that the viewer needs in order to understand the plot. As one of the purposes of this study is to assess the usability of error analysis in the research of subtitles, the aspects of subtitling conventions might have some interesting effects on the analysis but these will be addressed as they occur.

Even though studies (Vehmas-Lehto 1989; Lång 2010; Tuominen 2013) show that fluency and readability of the translated text as well as the synchronisation with the

audiovisual text are more likely to affect the viewing experience, it is clear that viewers do notice translation errors in subtitles. In my opinion, translation errors can have an effect on the overall quality of subtitles as well as prevent the subtitles from fulfilling their main functions as a support for understanding and a learning method. The fact that translation errors in subtitles have not yet been studied poses a risk that that viewers begin to regard errors as common; something that just happens in subtitling.

34