RSS feed Untitled Document

Quality in translation and adaptation for dubbing: Applied research in a professional setting

Giselle Spiteri Miggiani, University of Malta

ABSTRACT

Following the proposal of an assessment model that attempts to measure quality in dubbing (Spiteri Miggiani, 2022), this paper presents the data and findings of its application in a professional setting. The study aims to pin down the main translation and adaptation issues affecting quality standards in the professional practice, thereby highlighting them for the sake of translators, trainees, trainers, and other professional roles in the dubbing workflow. The goal is to address these issues through awareness, focused strategies, and dialogue with the stakeholders involved. The model was applied to two translation tasks carried out by a group of established adapters. The tasks were reviewed to identify, categorise, and quantify errors. The data analysis reveals that in one task, the top three errors were related to (1) script functionality, more specifically dubbing notations and character allocation; (2) natural sounding language, more specifically source calques; and (3) translation, more specifically unnecessary loss. In the second task, (1) synchronisation also emerged as a significant challenge, in particular, mismatched labial consonants; together with (2) functionality; and, (3) translation.

KEYWORDS

Dubbing, translation, adaptation, quality standards, assessment, established adapters, professional practice, taxonomy of errors, end product-oriented errors, process-oriented errors.

1. Quality awareness in media localisation

A recent customer impact survey carried out by Entertainment Globalization Association (EGA) in collaboration with Whip Media, reveals that 61% of the 15,000 respondents based in France, Italy, Germany, and Spain encounter poor localisation quality on streaming platforms. In most cases, this is leading to customer abandonment of movies and TV series: nearly 65% of the respondents have stopped watching at least one movie or TV show over a period of a year (EGA 2021). These survey results refer to both subtitling and dubbing. Quality in media localisation in general, but more specifically in dubbing, has recently become a hot issue. This could very well be due to the increased availability of dubbed content on OTT platforms, that also host newly emerging services such as English language dubbed streams (Hayes 2021; Sánchez Mompeán 2021; Spiteri Miggiani 2021a). Dubbing in traditional territories has always enjoyed a sterling reputation that now seems to be waning among streaming platform users.

The issue of quality deserves due consideration and research-based investigation intended at identifying enhancement strategies and potentially reversing any perceived negative trends. The matter is drawing the attention of both industry stakeholders and academic researchers. The industry is trying to address the need to enhance quality standards by offering training courses, while also reaching out to universities. The latter serves a two-fold purpose since it also addresses the so-called talent crunch (Green 2018; Deck 2021; MESA 2022), that refers to a lack of human resources in the field including translators and adapters.

Scholars in audiovisual translation are also engaging in the topic of quality from a theoretical perspective, even though quality-oriented theoretical discussions in audiovisual translation are often based on quality model proposals that include taxonomies of errors, classifications of quality standards or enhancement strategies. In fact, this study relies on a set of quality standards and an error-based method. To this end, it is worth mentioning studies that have addressed the issue of quality and quality control in interlingual subtitling (Nikolić 2021; Pedersen 2017; Robert and Remael 2016), intralingual live subtitling (Romero-Fresco and Martínez Pérez 2015) and interlingual live subtitling (Robert and Remael 2017; Romero-Fresco and Pöchhacker 2017). Among the assessment models, there is Künzli’s (2021) CIA model of interlingual subtitling quality based on Correspondence (between source product and TL subtitles), Intelligibility and Authenticity, each one having its own subset of parameters. Pedersen (2017) proposes the FAR model, an error-based assessment method focusing on: Functional equivalence (semantics, style), Acceptability (grammar, spelling, idiomaticity), and Readability (segmentation and spotting, reading speed, line length punctuation, use of italics). The NER model (Romero-Fresco and Martínez Pérez 2015) applied to intralingual live subtitling is based on error-detection, too, specifically: Number of words, Editing errors, and Recognition errors. On the other hand, the NTR model (Romero-Fresco and Pöchhacker 2017) applied to the interlingual live subtitling is based on Number of words, Translation errors, and Recognition errors.

As far as dubbing is concerned, recent emerging trends (such as English-language dubbing) are enticing scholars to question which specific approaches may possibly be adversely affecting the quality of dubbed products, and how these, among other factors, have an impact on viewer response (Sanchez Mompéan 2021; Spiteri Miggiani 2021a, 2021b). To-date, quality-oriented research pertaining to dubbing has focused on expected end-product standards. Several scholars have discussed general dubbing quality standards, among these Ávila (1997), Whitman-Linsen (1992), Chaves (2000), Chaume (2012, 2020), Spiteri Miggiani (2019). Some have focused on specific parameters, such as voice suitability or character synchrony (Bosseaux 2015; Martínez Sierra 2008), the prosodic features of dubbed speech (Sánchez Mompeán 2020) or natural-sounding dialogue (Pavesi 1996, 2016; Romero Fresco 2006; Baños Piñero 2009; Baños-Piñero and Chaume 2009).

Chaume (2007) proposes a classification of six main quality standards: acceptable lip-synch, credible and natural-sounding dialogue, fidelity to the original product, semiotic cohesion between words and images, clear sound and volume, and adequate role interpretation. This classification was further revisited to encompass two categories: textual and non-textual quality parameters (Spiteri Miggiani 2021a, 2021b). The non-textual parameters include (1) suitable voice selection; (2) convincing voice performance; (3) natural-sounding intonation; (4) appropriate sound quality, while the textual quality parameters include (1) adequate lip synchronisation; (2) natural-sounding language; (3) semiotic cohesion; (4) fidelity to source text; and (5) agreeable phonaesthetics.

The data collection and analysis in this study focus on the textual quality parameters since these pertain to the dialogue rewriting (Spiteri Miggiani, 2019: 28) process, that is, the translation and adaptation of the text to suit both technical and non-technical demands of the dubbing script. These parameters are mainly the responsibility of translators and adapters. Lip synchronisation is a technical demand that refers to matching the target language dialogue to the lip movements on screen. This includes timing, speech tempo, pauses and lip articulatory movements. Academic terminology labels these as phonetic synch (matching bilabial consonants, labiodental consonants, and lip-rounded vowels), isochrony (matching dialogue line duration and pauses) (Chaume 2012) and rhythmic synchrony (matching mouth flap movements) (Spiteri Miggiani 2021a). Semiotic cohesion, on the other hand, refers mainly to matching the target language dialogue to the body language or kinesics (Chaume 2012: 70). Natural-sounding dialogue is also considered a quality standard and refers to the need for the target language dialogue to sound credible and realistic (Romero Fresco 2006). That said, naturalness in dubbing scripts does not necessarily mirror real spontaneous spoken discourse, but is rather a register in its own right; a prefabricated orality (Baños Piñero and Chaume 2019: 1) that is tacitly tolerated by viewers who are accustomed to dubbing. Attention to phonaesthetics, on the other hand, guarantees pleasant-sounding dialogue or lack of cacophony, and is a requirement that directors and actors insist on in certain dubbing cultures. Finally, fidelity to the source text that, in dubbing scripts, usually entails a homologous or functional approach (Nord 2005, 2014) to translation, while respecting the so-called creative intent of the original product, a main priority from a client perspective.

With this background in mind, this study aims to identify the most recurrent translation and adaptation issues affecting quality standards in the dubbing professional practice. It aims to pin down that which is often vague or elusive. Professional agents and viewers may sometimes judge a script as ‘good’, ‘average’ or ‘bad’, although it may not always be easy to pinpoint what makes it so. Besides, translators, trainees/students, trainers, and other professional roles in the dubbing workflow may find it helpful to know what to look out for while they are fully immersed in the dubbing process of a product. Being aware of the issues and specifics can also act as a first step, to then address these challenges through focused strategies and dialogue with the stakeholders involved. A secondary research aim of this study is to observe whether a taxonomy of errors based on a set of quality standards is exhaustive enough or whether other criteria should be used to determine the quality of translation and adaptation for dubbing.

2. The study
2.1 The applied model

The study seeks to identify the specific issues related to the quality of dubbing scripts. It is therefore based on the premise that the quality of a script is impacted by the presence of what will be conveniently referred to as ‘errors’, in line with quality assessment terminology. It follows on a previous paper that proposes a quality assessment model for dubbing, the Textual Parameters or TP model (Spiteri Miggiani 2022), that focuses on the translation and adaptation phase. This model encompasses a taxonomy of errors combined with an error-based formula that enables a quality controller or evaluator to mark any given translation for dubbing with a percentage score. The present study only applies the taxonomy of errors belonging to this TP model.

The model is based on the above-mentioned revisitation of commonly accepted dubbing quality standards (Spiteri Miggiani 2021a, 2021b), further adapted from Chaume (2007). These product-oriented quality standards include all those aspects that are part of the final product and reach the end user, that is, the viewers. Therefore, errors refer to issues in the following categories: synchronisation, language, visuals and sound, translation, and phonaesthetics. Issues pertaining to visuals and sound imply lack of semiotic cohesion between the target text and the images or the aural elements belonging to the original audio track. Issues tied to phonaesthetics usually imply cacophony or any element that hinders the text from sounding pleasant to the ears. Synchronisation errors emerge when the target text does not match the lip movements on screen. The translation and language categories, at first glance, may seem more straightforward, though the specifics in the TP model need to be considered with a dubbing mind-set that does not necessarily tally with other fields of translation. For instance, omission and loss are both included in the translation category. Omission is often necessary in dubbing, being one of the reduction strategies adopted, so it would be considered an error only if applied unnecessarily, or if other redundant elements could have been easily omitted instead. Loss, on the other hand, implies any meaning that is lost unintentionally and is not necessarily tied to omission, the latter being a deliberate strategy. For example, the translation of a dialogue line may fail to convey its underlying tone because the translator did not pick up on that nuance conveyed at a prosodic level in the original dialogue. This loss would then most likely be carried over during the recording phase. The language category includes lack of naturalness which is generally associated with the presence of so-called dubbese, the ‘artificial’ register of dubbing. Lack of naturalness in a dubbing script most likely leads to unnatural intonation, affecting the prosodic level, too. Naturalness therefore refers to the positioning of fictitious dialogue lines on the written-spoken continuum, since target language translations are written to be spoken, recited and recorded. The synchronisation constraints are among the main factors affecting naturalness in dubbed dialogue. The translation category, on the other hand, includes a wider micro category labelled awkward translation. This can incorporate other problematic issues that are not necessarily narrowed down to naturalness in speech utterances and are more tied to the translation before it has been affected by the synchronisation process. It could also refer to the rendering of verbal elements in the dubbed product that cannot be classified as dialogue, as such. These could include the translation of on-screen written graphics, website snippets, news features and other modes of discourse incorporated in the dubbed product, such as voice-over narration.  Other error specifics will be discussed further on during the analysis of the findings.

In addition to the end product quality parameters, the TP model includes a process-oriented quality parameter: script functionality (Spiteri Miggiani 2022) which refers to those practical aspects that can possibly disrupt or slow down the dubbing workflow when the script is being handled by the other professional roles involved. These elements are part of the script (such as notations for the actors, layout, format, text actor-friendliness, character allocation, pause markers, etc.) and are functional to the dubbing workflow and more specifically the recording process. For example, one of the errors that belongs to this category is missing or inadequate ‘walla’ that refers to crowd murmur or background dialogue. Background dialogue lines need to be singled out and attributed to different characters to support the recording workflow. Failing to do so would constitute a functional error. Likewise, missing dialogue could refer to the adapters who skip speech utterances by mistake, or fail to incorporate them when these are heard in the audio but do not feature in the original script deliverable provided. Other specifics include orthographic or typographic errors that belong to this category rather than the language category. Dubbing scripts are written to be spoken and recorded and will not be seen or read by viewers. Hence, a spelling mistake will constitute a practical issue if it somehow disrupts the recording flow. For instance, if there is inconsistent spelling across character names, this could possibly lead to a dialogue line being missed in the case of recording workflows that rely on automated methods to single out each character’s dialogue. Orthographic mistakes could be considered a practical issue if they are so frequent that the actors find them distracting, to a point where they might have to stop recording to adjust the text. Since such errors belong to the functionality rather than the language category, evaluators adopting this model can choose to ignore them if they do not cause any disruption.

The model’s error-based formula is intended to grade translations and rank translators in a professional or training setting, but for the purposes of this study, the percentage scores of each adapter are not relevant. In this context, focus lies on the errors outlined in the assessment rubric illustrated in Table 1. The main error categories correspond with the main textual quality standards outlined earlier, and are simplified as follows: synchronisation, language, visuals and sound, translation, phonaesthetics, and functionality. These categories are then broken down further into 37 error specifics. Therefore, evaluators, trainers, and professionals can choose between a simplified and detailed version of the same assessment rubric. Each error category and error specific are assigned a code for ease of reference during an evaluation or feedback process. In this context, the error codes were used to identify and label the errors encountered. Table 1 illustrates the detailed variant that was adopted for the purposes of this study. The next section outlines the research design and how the model was applied.

Quality parameter
(Textual parameters)

 Generic
error code

Error
category

 Specific
error
code

Error
specifics

Adequate lip synchronisation

[S]

Synchronisation

[…]

Too short

[--]

Too long

[R]

Rhythmic issues
(mouth flaps mismatch)

[L]

Labial consonants mismatch

[V]

Vowels or semivowels mismatch

Natural-sounding language

[L]

Language

[GR]

 Incorrect grammar

[SC]

Source calque

[REG]

Unsuitable register

[COMP]

Lack of clarity & comprehension

[NAT]

Lack of naturalness

[FLOW]

Lack of flow & cohesion
between dialogue exchanges

Semiotic cohesion

[VS]

Visuals & Sound

[VIS]

Lack of cohesion between words & visuals (such as body language

[SND]

Lack of cohesion between words & sound belonging to the original audio track (music & effects, lyrics, noise)

Fidelity to
source text

[T]

Translation

[MIS]

Mistranslation

[OM]

Unnecessary omission

[ADD]

Unnecessary addition

[LOSS]

Unnecessary loss (semantic)

[AWK]

Awkward translation
or rendering

[IMP]

Improper translation (undue non-inclusive, offensive, derogatory terms that are not functional to the plot
or characterisation)

Phonaesthetics

[PH]

Phonaesthetics

[CAC]

Cacophonic utterances

[REP]

Annoying repetition

[RHY]

Unintended rhyme

Script functionality

[F]

Functionality

[CON]

Lack of consistency
(non-compliance with glossary sheets; inconsistent use of names/nicknames, forms of address & terminology within same script or across serial production scripts)

[REAC]

Missing or wrong reaction

[NOT]

Missing or wrong notation

[/]

Missing pause marker

[FOR]

Layout or format issues

[DS]

Unsuitable dialogue segmentation

[OR]

Orthography mistakes

[CH]

Wrong character allocation

[D-?]

Missing dialogue

[B-?]

Missing or inadequate
background walla

[PUN]

Misleading punctuation

[TC]

Missing or wrong time code

[G/P]

Non-compliance with guidelines & policies

[PRON]

Tricky articulation
or pronunciation

[MISC]

Miscellaneous

Table 1. Quality assessment rubric drawn from Spiteri Miggiani (2022)

2.1 Methodology

The findings presented in this paper are based on a group of twenty-three established freelance adapters based in Italy. The requirements to participate included the ability to translate from English as a source language and long-term experience (five+ years) in adaptation for Italian-language dubbing. A brief questionnaire was handed out to gather more information about the adapters, who in this case turned out to have on average twelve to twenty-five years of professional experience.

The participants were asked to carry out two tasks that involved both translation and adaptation from English into Italian. Therefore, the model was applied twice in the same setting and with the same group of adapters. It is useful to note that in professional practice the translation and adaptation process is sometimes split into two distinct phases and carried out by two different professionals. This is quite common in traditional dubbing territories. The adapters may not necessarily have any knowledge of the source language, but they would have the adaptation skills to suit the text for dubbing purposes. Although the participants involved in the study were mostly accustomed to this type of traditional dubbing practice, they were asked to carry out both translation and adaptation so that all quality parameters could be considered. In any case, the participants will be referred to as ‘adapters’ throughout this paper, in line with industry jargon. The terms ‘translators’ and ‘adapters’ are sometimes used interchangeably to refer to the entire textual workflow, from translation to adaptation, but they are also used to highlight the separation of roles.

2.1.1 The adaptation tasks

The two tasks will be referred to as Task 1 and Task 2. The participants had to complete Task 1 in a couple of hours, while they were less under pressure for Task 2 which was completed in their own time over a few days. Task 1 was a 1 minute, 30 second excerpt drawn from a US TV series. The level of difficulty was low to average, although it still presented quite a few challenges that were identified through an analysis of the source audiovisual text. Technical challenges included (1) varying speech tempo, rhythm, pauses, voice intensity; (2) alternating off and on-screen shots and shot/reverse shot throughout the same speech utterance; (3) bilabial consonants on close-up shots; (4) fast, almost-overlapping speech in some instances; (5) hesitation; (6) subtle paralinguistic features; (7) kinesics. Translation and linguistic challenges included (1) slang expressions and mild insults; (2) source interference traps (lexical and syntactical source calques, use of pronouns); (3) redundancy and sentence structure; (4) sarcasm and varying emotional intensity; (5) forms of address and vocatives. Script functionality challenges included (1) use of notations due to continuous alternating OFF/ON screen shots; (2) deliberate errors in character allocation in the original script; (3) deliberate missing dialogue bits in the original script; (4) original script time codes that did not correspond with the video file time codes. Task 2 was of a higher degree of difficulty. This excerpt was also drawn from a US TV series, therefore, same genre, but longer in duration (3 minutes 35 seconds). Technical challenges included (1) tricky rhythm; and (2) several close-up shots. Functionality challenges included (1) constant alternation between off and on-screen within the same utterance; (2) filtered effects; (3) several paralinguistic features; and (4) background dialogue not signalled in the original script. On a linguistic level and translational level, the original dialogue encompassed (1) specialised jargon, mainly legal; (2) intertextual references; (3) sarcastic tone; (4) vulgar language; (5) slang; (6) homophobic language; (7) gender pronoun challenges; and (8) varying registers.

2.1.2 The script review process

When both tasks were completed, data were collected by means of a thorough review of the target language scripts and subsequently analysed by the researcher, who was also the evaluator. The review process involved spotting, labelling, and quantifying both generic and specific errors encountered in both tasks in accordance with the assessment rubric illustrated in Figure 1. These were totalled, and percentages subsequently drawn to reveal the most recurrent errors in each task. The results of both tasks were then combined and an average was drawn. It is important to note that the review process was carried out consistently across the scripts. In other words, when a specific element was treated as an error, or otherwise ignored, this was applied across all adaptations.

Despite the use of the term ‘errors’, some of the issues encountered may not necessarily be errors, as such. The term also encompasses those issues that may simply not be permissible in a specific dubbing project or that may imply additional time or costs. This, of course, depends also on territorial norms and client expectations. Certain restrictions deriving from local territorial policies may be a case in point. Some issues may be considered as blockers or major issues, and in such cases, changes in the text prior to recording would be required, while other issues may be considered as minor and do not necessarily disrupt the workflow. Such considerations were made during the review of the tasks, which was therefore carried out from an industry and practitioner perspective. This implied highlighting only those errors that could potentially be an issue during the recording workflow. The scripts were voiced during the review process to test each dialogue line against the video before signalling any issues, both technical and linguistic. The review was not based solely on the written target text or on a comparison between the source text and the target text. For example, if a target version dialogue line featured five mismatching bilabial consonants based on the written text, this did not necessarily imply five flagged errors. The dialogue line could have been considered as having no errors at all depending on the technical needs of the specific line also in relation to the other filmic codes. Choosing which errors to highlight or ignore naturally implies a small degree of subjectivity. Such review choices were based on know-how acquired during long-term professional practice, regular exchanges with dubbing directors, assistants, actors, and sound engineers as well as feedback, continuous observation, auditing, and active participation in different dubbing recording sessions. Finding other Italian-native long-term practitioners (ideally also researchers and academics) who could assess dubbing scripts from a recording studio perspective, while being willing to quantify and label errors according to a specific model, for research purposes, was no easy feat. Hence, the review was carried out by the researcher alone, despite the limitations that this implies.

3. Findings
3.1 Recurrent generic errors in Task 1 and Task 2

Figures 2 and 3 reveal the error percentages in Task 1 and Task 2, respectively. In Task 1, the language category had the highest error rate (26.6%). Close runners-up were functionality (24.8%) and translation (24.5%). With reference to the taxonomy of errors illustrated in Table 1, language errors are intended as lack of naturalness, incorrect grammar, unsuitable register or style, lack of flow and cohesion between dialogue exchanges, source language interference and lack of clarity. Translation errors include mistranslation, unnecessary omission or addition, unnecessary loss, awkward rendering, and improper translation which refers mainly to sensitive and inclusive language. An example of improper translation could imply an instance where the target language text sounds more offensive than its original counterpart. These errors are most likely attributable to lack of translation practice because as mentioned earlier, the workflow is often split between two professionals. Many established adapters delegate the raw translation to a professional translator so that they can focus their attention on the adaptation phase. This could also be due to lack of proficiency in the source language.

Functionality errors, on the other hand, could possibly be attributed simply to lack of attention or the short span of time provided for the task (even though, it may be argued that the amount of time given mirrored that of the professional practice. In any case, increasingly short turnaround times in the industry demand fast script deliveries). As mentioned earlier, these refer to the practical aspects in the script that can slow down or disrupt the dubbing workflow in the production phase. They include issues with format, layout of dialogue segmentation, lack of consistency, missing reactions or sounds, missing dialogue or background conversation, misleading or missing dubbing notations or pause markers, misleading punctuation, orthography mistakes, wrong character allocation, missing or wrong time code and non-compliance with guidelines and policies.

A plausible reason that can account for the significant number of functionality issues can be referred to as the ‘adapter bubble’. The adapters’ role contributes to the initial phase of the workflow, before starting the recording process. Their presence in the studio is not a requirement; they work at home in isolation and therefore are often detached from the rest of the dubbing workflow reality. There is almost always no follow-up on their submissions and therefore little or no verification from their end as to whether their submitted script is adequate on a functional level or whether it meets client expectations and demands. This detachment is clearly reflected in the diverse approaches and styles of the participants’ scripts, and this also highlights the lack of a standardised approach. One such example is the fact that some scripts contained an overabundance of dubbing notations while others provided the bare minimum. Adapters who are very familiar with dubbing workflows and recording dynamics are aware that it is best to keep notations to a required minimum since overabundance can possibly hinder the recording process and increase costs. The extreme approaches that emerged in both tasks reveal that some adapters are more aware than others. Moreover, some scripts lacked consistency when it came to layout, formatting, fonts, and so on, and not enough attention was always given to certain practical important details, such as character allocation. A more detailed discussion on the error specifics will ensue in Section 3.2.1.

The top three rated errors in Task 2 were functionality (34.3%), synchronisation (29.2%), and translation (21.4%), therefore functionality and translation were common denominators in both tasks, while language issues (12%) were fewer in the second task and were surpassed by synchronisation issues (29.2%). The higher level of difficulty of Task 2 could have contributed to the increase in synchronisation errors. The lack of time pressure, given that they had more time to complete Task 2 compared to Task 1 did not seem to contribute to a better performance.

Figure 4 and Table 2 merge the error occurrence in both tasks. The average final outcome reveals that the highest number of errors fall into the functionality category, followed by synchronisation and translation. Section 3.2 will investigate the errors in further detail and examine the specifics of each error category.

3.2 Error specifics

Tables 3a, 3b, 3c, 3d, 3e and 3f illustrate the participants’ performance when it comes to error specifics within each main error category. These tables outline the overall percentages reflecting the presence of specific errors in each task. It is important to note that, as mentioned earlier, the two tasks differed in length (duration and number of words); this explains the use of percentages to draw a comparison. The comparison between tasks is intended to establish common threads and possibly strengthen the findings thanks to the fact that the assessment model was applied twice with the same group of adapters.

Table 3a reveals the presence of source calques as a major issue in the language-related category, followed by lack of naturalness in the target language. Next in line are unsuitable register and lack of clarity, that is, dialogue lines that may not be understood immediately by a wide target audience. This is probably due to the way in which the dialogue line is formulated or perhaps due to ambiguity resulting from the missing subject or object in the clause. Section 3.2.1 provides specific types of errors for each category drawn from both Task 1 and Task 2.

Table 3b illustrates the translation-specific errors. Unnecessary loss accounts for more than 50% of the errors, followed by mistranslation and awkward rendering, though to a lesser extent. Unnecessary loss refers to those instances in translation where there is an unintentional missing element that contributes to the plot or characterisation as opposed to omission (most likely intentional) to suit the technical demands of the script. Section 3.2.1 provides examples, such as underlying sarcasm expressed through intonation that went unnoticed. Hence, loss also refers to the prosodic features, in other words a loss in the delivery of dubbing actors who would not be able to reproduce the original intent because this is lost in the written target language text. Improper translation errors in Task 1 refer to utterances which came across as far more emotionally charged, vulgar or offensive when compared to the original utterances, while also considering target culture sensitivity, vulgar language TV norms, client and viewer expectations, and also the medium and type of production at hand.

Table 3c focuses on the so-called technical errors, that is, those related to the synchronies. Almost 50% of the errors consist of labial consonant mismatches. It is important to note that all along the target text review process, client and viewer norm expectations and demands were considered. In other words, only the necessary lip articulatory movements were treated as errors. Labial mismatches mainly constitute missing bilabial stops or labiodentals in the target text when these are very evident in the visuals. The reverse situation, that is, bilabial stops added in the target text only (possibly causing an evident aesthetic mismatch), was also kept in view, though to a lesser extent.

Other common errors were dialogue lines that were too long, and mismatches between target language and source language vowels and semi-vowels. There were also some rhythmic issues, although, in comparison they were less frequent. That said, it would be interesting to replicate the experiment with a different language pair, perhaps with English as a target language. It is easy to hypothesize that rhythmic issues would prevail in this case.

Table 3d presents the most common functionality issues encountered in both tasks. These are mainly related to dubbing notations, that is, missing notations — where considered indispensable from a dubbing studio perspective — or overabundance, which was mostly the case in this experiment. As mentioned earlier, the review process was carried out consistently across all scripts. When a missing notation was treated as an error, this was applied across all adaptations, while other missing notations were ignored because these were deemed less important. The possible reasons behind functionality issues were put forward in Section 3.1. Dubbing notations were followed by wrong character allocation. Almost all adapters failed to notice the wrong names allocated to specific dialogue lines and carried them over to their target language script despite the video material. Other functionality errors featuring in the scripts, as can be seen in Table 3d, include miscellaneous errors. Indeed, the application of this assessment rubric revealed that some errors are not catered for in the detailed taxonomy. These include errors such as missing translation for written graphics and other types of errors such as typos, the repetition of the same word in the sentence or a missing word altogether.

Table 3e reveals that there was only one error occurrence when it came to the visuals and sound category. This refers to the temporal and semantic correspondence between a specific word in the target language and the visuals, or any sound element that belongs to the music and effects track.

Table 3f reveals phonaesthetic issues, in particular cacophonic utterances, such as consonant clusters or the hissing sound produced by too many occurrences of the letter ‘s’, or unnecessary repetition within the same dialogue block. This can make the text sound heavier and unpleasant in the target language, irrespective of whether this same repetition is present in the original dialogue.

3.2.1 Specific types of errors across both tasks

This section presents specific types of errors related to each of the six main categories to further illustrate what has already been discussed.

Functionality:

  • Wrong character allocation: the original script featured a deliberate error. Half the group did not notice the error and the wrong character name featured also in the dubbing script;
  • Missing dialogue: the original script featured missing dialogue as is often the case in original dialogue lists, and some adapters failed to notice this;
  • Missing notations and pauses featured throughout, as well as overabundance in the second task.
  • There were several inconsistencies in typeface and font size for notations. Even though there is no set universal font or typeface, consistency across the same script is considered important. These were also observed in the use of “..” or “…” to indicate a short pause;
  • Typos and punctuation errors were observed throughout;
  • Missing distinct or indistinct background dialogue, e.g., clients ordering food in a diner.

Synchronisation:

  • Issues with labial consonants and vowel articulatory movements especially at the end of dialogue lines in close-up shots, e.g., most adapters failed to find matching lip articulatory movements;
  • Issues with timing and rhythm, especially in fast-paced lines with several mouth flap movements;
  • Rhythmic issues, such as elongated and stressed monosyllabic words.

Language:

  • Several instances of source interference and source calques;
  • Several instances of literal translation;
  • Unnecessary pronouns, e.g., Io (I) or tu (you) at the beginning of dialogue lines;
  • Lack of coherence between dialogue exchanges, e.g., intentional use or repetition of words in an exchange that have been omitted in the target language;
  • Lack of clarity/comprehension, e.g., the subject being referred to is not clear;
  • Unsuitable register, e.g., a lower register and informal jargon spoken by a lawyer, or a childish register adopted by an adult criminal in the target version.

Translation:

  • Unnecessary loss, e.g., underlying sarcasm in the tone of voice of a charcter was lost. Also, some intertextual references went unnoticed, e.g., biblical references;
  • Stronger language compared to the original intent, e.g., the use of terms to translate insults that sound more derogatory in the target language;
  • Lack of naturalness, e.g., affirmative statements instead of interrogatives that would sound more natural in the target language.

Phonaesthetics:

  • Target language cacophony due to repetition, consonant clusters, the hissing ‘s’, unnecessary rhyme, and long winding sentences.
4. Discussion and conclusions

This research study aims to pin down the most recurrent translation and adaptation issues featuring in the dubbing professional practice, and which possibly affect the overall quality of a dubbed product. Undoubtedly, there are other non-textual factors that impinge on the final product and that depend on the other professionals in the workflow. However, these too lean heavily on the adapted script, therefore the quality of the script can undeniably influence viewer experience in a positive or negative way. In addition to product-oriented issues, there are also process-oriented issues that may disrupt the dubbing workflow. This paper, therefore, also looks at quality from a localisation company standpoint.

To achieve the aims mentioned, the study set out to apply a dubbing quality assessment model, the Textual Parameters or TP model (Spiteri Miggiani 2022) in a professional setting. Two translation and adaptation tasks carried out by well-established adapters were reviewed in accordance with the taxonomy of errors incorporated in the chosen model. Data pertaining to the main generic error categories and the more detailed error specifics were drawn and analysed. The findings reveal that the top error categories were functionality, as a process-oriented category and translation as a product-oriented category. Also, one of the tasks featured several language-related issues, while the other featured many more synchronisation issues. The main specific errors that featured in the tasks include issues with dubbing notations and character allocation; unnecessary loss; source calques; and labial consonant mismatches. As will be elucidated in the concluding remarks, the product-oriented errors can possibly be avoided through awareness and focused training, while the process-oriented errors reveal the need for more feedback, script follow-up and communication between clients and adapters.

This research study also has its limitations. Apart from being restricted to one language combination, another limitation of the study is the fact that the researcher and the reviewer of the target texts were the same person. Moreover, having more than one evaluator would have reduced the degree of subjectivity during the review process. Another possible limitation is the fact that the tasks handed out were not intended for recording and broadcast as in a real case scenario. This could have possibly influenced the performance and the effort put in the given tasks. Another minor limitation that emerged belongs to the model itself. As mentioned earlier in Section 3.2, the error specifics pertaining to the functionality category do not cater for all the errors that were encountered, therefore some were simply labelled as miscellaneous. Moreover, there were a few instances where a given error could possibly be classified under more than one category, hence occasional overlaps occurred, though this was rare.

A secondary research aim put forward at the beginning of this paper was to observe whether a taxonomy of errors based on a set of quality standards is exhaustive enough or whether other criteria should be used to determine the quality of a translation and adaptation for dubbing. In the first instance, it is important to mention that this type of quality assessment model does not reward brilliant or creative solutions or holistically ‘better’ scripts. However, this is in line with the industry approach: scripts need to ‘function’ and translators are not usually rewarded for coming up with better solutions. They would at most be assigned further projects. Those subtle differences that may lead dubbing directors and viewers to perceive a target dialogue list as ‘better’ or ‘worse’ are not quantified in the assessment model applied, which is limited to singling out the issues. Though as mentioned earlier, the intention is precisely to try to pin down that which is often vague or elusive when determining the quality of a dubbing script. Therefore, despite being aware of the limitations of an error-based model approach, if quality can be said to be impacted by errors, this is a first necessary step to start filtering out such issues and move towards enhanced quality. Moreover, this approach can possibly help distinguish between objective errors and negative judgement based on personal style or subjective approaches to translation, in other words, simply disliking a translation choice. Despite necessarily involving a minimum degree of subjectivity, the ‘rigidity’ of the assessment rubric is what ensures further objectivity on behalf of quality evaluators rather than being driven by feeling, perception or subjective approaches. Ideally, an assessment rubric should be combined with further evaluation criteria or a more holistic approach that would also focus on quality at other levels, e.g., taking creativity into account. That said, this was beyond the scope of this research study.

In conclusion, these findings may possibly contribute to enhancing both viewer and client satisfaction if the recurrent issues are addressed by stakeholders and the adapters themselves. The first step toward this achievement is awareness of the specific issues accompanied by dialogue and collaboration between academia and the industry. Moreover, dialogue between adapters and their clients is also necessary.

Once the recurrent issues are identified, a possible way to address them is to consider focused professional development training. Providing feedback to adapters would also undoubtedly help improve quality standards, especially in those contexts where quality control is not integrated in the workflow. Assessment models, such as the one applied in this research study, can be used to provide feedback as part of the dubbing workflow. Inserting error codes in the dubbing scripts can be a feasible and quick way of drawing the adapters’ attention towards certain issues without the need for lengthy explanations. Translators can refer to the assessment rubric and use it as a legend for the codes appearing in their script. These can signal which specific issues are present and where in the dialogue. Applying this randomly to script samples could also suffice.

Another important aspect to consider is the ‘adapter bubble’ mentioned earlier, that is, the detachment between adapters and the dubbing studios where the entire workflow unfolds. Moreover, there is isolation among the adapters themselves who often do not have the possibility to communicate and exchange ideas and solutions, be it whether they are working on the same production or otherwise. Most adapters submit their text without any follow-up, and this invariably leads to the delivery of scripts with recurrent issues. Finding ways to enhance communication between adapters and dubbing directors or dubbing managers, and involving the adapters further in the workflow would go a long way towards addressing quality issues. Dubbing workflows that have migrated to cloud platforms seem to facilitate communication and feedback, albeit remotely (Spiteri Miggiani, 2023). However, this is not yet the widespread reality in traditional dubbing countries.

Further research is necessary to apply the same assessment model to other language pairs, other territories, and other groups of well-established adapters to corroborate the findings and investigate quality from a global perspective.

References
  • Ávila, Alejandro (1997). El Doblaje. Madrid: Cátedra.
  • Baños Piñero, Rocío (2009). La oralidad prefabricada en la traducción para el doblaje. Estudio descriptivo-contrastivo del español de dos comedias de situación: Siete Vidas y Friends. PhD thesis, Universidad de Granada, Granada.
  • Baños-Piñero, Rocío and Frederic Chaume (2009). “Prefabricated Orality: A challenge in Audiovisual Translation.” Special issue of inTRAlinea.
  • Bosseaux, Charlotte (2015). Dubbing, Film and Performance: Uncanny Encounters. Oxford: Peter Lang.
  • Chaume, Frederic (2007). “Quality standards in dubbing: a proposal.” TradTerm 13, 71-89.
  • Chaume, Frederic (2012). Audiovisual Translation: Dubbing. Manchester: St. Jerome Publishing.
  • Chaume, Frederic (2020). “Dubbing.” Łukasz Bogucki and Mikołaj Deckert (eds) (2020). The Palgrave Handbook of Audiovisual Translation and Media Accessibility. London: Palgrave Macmillan, 103-132.
  • Chaves, Maria José (2000). La Traducción Cinematográfica: El Doblaje. Huelva: Universidad de Huelva Publicaciones.
  • Deck, Andrew (2021). “Lost in translation: The global streaming boom is creating a severe translator shortage.” Rest of world journal. https://restofworld.org/2021/lost-in-translation-the-global-streaming-boom-is-creating-a-translator-shortage/
  • EGA (2021). Localization consumer impact study: France, Italy, Germany and Spain. Glendale, California: EGA press.
  • Green, Stuart (2018). “How digital demand is disrupting dubbing.” M&E journal.
  • Hayes, Lydia (2021). “Netflix disrupting dubbing. English Dubs and British Accents.” Journal of Audiovisual Translation 4(1), 1-26.
  • Künzli, Alexander (2021). “From inconspicuous to flow — the CIA model of subtitle quality.” Perspectives 29(3), 326-338.
  • Media and Entertainment Services Alliance (MESA) (2022). “The Talent Crunch - Does it Exist and can it be Addressed?” Content Workflow Management Forum, 22 March 2022.
  • Martínez Sierra, Juan José (2008). Humor y Traducción: Los Simpson Cruzan la Frontera. Castellón: Universitat Jaume I.
  • Nikolić, Kristjan (2021). “Quality Control of Subtitles: A Study of Subtitlers, Proofreaders, Quality Controllers, LSPs and Broadcasters/Streaming Services.” Journal of Audiovisual Translation 4(3), 66-88.
  • Nord, Christiane (2005). Text Analysis in Translation. Theory, Methodology, and Didactic Application of a Model for Translation-Oriented Text Analysis. Amsterdam/New York: Rodopi.
  • Nord, Christiane (2014). Translating as a Purposeful Activity: Functionalist Approaches Explained. London/New York: Routledge.
  • Pavesi, Maria (1996). “L’allocuzione nel doppiaggio dall’inglese all’italiano.” Rosa Maria Bollettieri Bosinelli and Christine Heiss (eds) (1996). Traduzione multimediale per il cinema, la televisione e la scena. Bologna: Clueb, 117-130.
  • Pavesi, Maria (2016). “Formulaicity in and across film dialogue: Clefts as translational routines.” Across Languages and Cultures 17, 99–121.
  • Pedersen, Jan (2017). “The FAR model: assessing quality in interlingual subtitling.” The Journal of Specialised Translation 28, 210-229.
  • Robert, Isabelle and Aline Remael (2016). “Quality control in the subtitling industry: An exploratory survey study.” Meta 61(3), 578-605. 102.
  • Robert, Isabelle and Aline Remael (2017). “Assessing quality in live interlingual subtitling: A new challenge.” Linguistica Antverpiensia NS: Themes in Translation Studies 16, 149-167.
  • Romero-Fresco, Pablo (2006). “The Spanish Dubbese: A case of (un)idiomatic friends.” The Journal of Specialised Translation 6, 134–151.
  • Romero-Fresco, Pablo and Juan Martínez Pérez (2015). “Accuracy rate in subtitling: The NER model.” Rocío Baños Piñero and Jorge Díaz Cintas (eds) (2015). Audiovisual Translation in a Global Context: Mapping an Ever-Changing Landscape. Basingstoke: Palgrave Macmillan, 28-50.
  • Romero-Fresco, Pablo and Franz Pöchhacker (2017). “Quality assessment in live interlingual subtitling: A new challenge.” Linguistica Antverpiensia NS: Themes in Translation Studies 16, 149-167.
  • Sánchez-Mompeán, Sofía (2020). The Prosody of Dubbed Speech. Beyond the Character’s Words. Basingstoke: Palgrave Macmillan.
  • Sánchez-Mompeán, Sofía (2021). “Netflix likes it dubbed: Taking on the challenge of dubbing into English.” Language and Communication 80, 180–190.
  • Spiteri Miggiani, Giselle (2019). Dialogue Writing For Dubbing. An Insider’s Perspective. London: Palgrave Macmillan.
  • Spiteri Miggiani, Giselle. (2021a). “English-language dubbing: challenges and quality standards of an emerging localisation trend.” The Journal of Specialised Translation 36, 2-25.
  • Spiteri Miggiani, Giselle. (2021b). “Exploring applied strategies for English-language dubbing.” Journal of Audiovisual Translation 4(1), 137-156.
  • Spiteri Miggiani, Giselle (2022). “Measuring quality in translation for dubbing: a quality assessment model proposal for trainers and stakeholders.” XLINGUAE 15(2), 85-102.
  • Spiteri Miggiani, Giselle (2023). “Cloud studios and scripts: Evolving workspaces and workflows in dubbing.” Carmen Pena-Díaz (ed) (2023). The Making of Accessible Audiovisual Translation. Oxford/New York: Peter Lang, 145-175.
  • Whitman-Linsen, Candace (1992). Through the Dubbing Glass. Frankfurt am Main: Peter Lang.
Note:

The data that support the findings of this study are not publicly available and the information contained in this article has been published with the courtesy and permission of the third party involved. Restrictions apply to the availability of these data due to information that could compromise the confidentiality of the research setting and the privacy of research participants.

Biography

Giselle Spiteri Miggiani, Ph.D. is a tenured lecturer at the University of Malta where she set up a postgraduate audiovisual translation specialisation stream. She is also an audiovisual translator and dubbing adapter since 2006. She is regularly invited to deliver lectures and talks at other universities and provides training and consultancy services to global leading stakeholders. She is the author of the book Dialogue Writing For Dubbing. An Insider’s Perspective (Palgrave Macmillan, 2019), among other journal articles and chapters.

E-mail: giselle.spiteri-miggiani@um.edu.mt

ORCID: 0000-0002-6382-9938