PreprintPDF Available

'Time is money' and the value of translation

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

This article uses a multi-faceted approach to discuss the relation between time, money and different perspectives that help define the value of professional translation. It challenges the narratives created by the translation industry on post-editing as a revision of pre-translated content, confronting them with the detailed description of the task in industry standards and with the reality of translators' work. The article also addresses the different roles that time plays as an instrument of analysis and evaluation of translation, and as a fundamental factor in the definition of labour relations in the translation market. The main claim of the article is that translation is an increasingly specialised high-value work, requiring translators that are able to make complex and efficient decisions, especially when they are expected to work under time restrictions, with the support of content that has been previously processed by machine translation.
Content may be subject to copyright.
1
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
Time is money and the value of translation
Abstract
This article uses a multi-faceted approach to discuss the relation between
time, money and different perspectives that help define the value of
professional translation. It challenges the narratives created by the
translation industry on post-editing as a revision of pre-translated content,
confronting them with the detailed description of the task in industry
standards and with the reality of translators work. The article also
addresses the different roles that time plays as an instrument of analysis
and evaluation of translation, and as a fundamental factor in the definition
of labour relations in the translation market. The main claim of the article
is that translation is an increasingly specialised high-value work, requiring
translators that are able to make complex and efficient decisions,
especially when they are expected to work under time restrictions, with
the support of content that has been previously processed by machine
translation.
Keywords
Value of translation; time as a definer of value; professional translation; post-editing of
machine translation; translation productivity; economic factors in translation; translation
efficiency; translation industry.
1 Introduction
There are many ways to assess the value of a professional activity. The most immediate
indicators are how much the products from that activity cost and how long it takes to produce
them. We can also factor in the expertise and qualifications of the people that produce them,
or try to grasp the perception that the public, or a portion of it, has of the value of the
profession. The first two factors, product cost and production time, occupy the focal point of
this article. The other two will be referred to indirectly: specialised qualifications are seen as
2
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
the answer to the requirements of a complex work environment, and the perceptions of the
value of translation will be at the background of the whole article, more visible when we
analyse some of the most powerful discourses on translation.
The aphorism ‘Time is money’ is attributed to Benjamin Franklin (1748), having been
used as the moral of an anecdote which can be summed up as a man who wastes time on
idle activities when he should be working does not understand the value of time’. The
expression is a call for careful time management, so as not to waste its intrinsic value. Modern
professional lives are guided by regular and strict forms of evaluation of performance,
embedded in a global pressure for increased productivity and efficiency. As the aphorism has
not lost its relevance, it will serve as guidance to a reflection on the different dimensions of
the value of translation.
This article starts by analysing diverse descriptions of the tasks performed by
translators, based on elements of time, not only by academics and researchers, but also by
the voices that represent the industry. This is presented in Section 2, which reflects on how
the assumptions of increased productivity embedded in the description of translation tasks,
instead of contributing to rewarding gains of efficiency, result in downgrading the value of
the different services provided by translators.
Section 3 reflects on direct and indirect effects of time in professional translation, first
how its passage affects income, then how time pressure is changing translator training and
finally by looking closely at a small sample of the issues that make it difficult for translators to
control their production time.
The next section considers different dimensions of time, both as a metric and as a factor
in labour relations. The final part of the section reflects on how time is not transparent enough
to help us determine the value of translation.
The conclusions of the article highlight the claim that the demands of the business world
of translation can only be met by an expert workforce, which has the capacity to produce high
value translations in a short time, so as to respond to the demands of a technologically-
sophisticated environment.
2 Time-based descriptions of processes in translation
3
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
2.1 Translation, revision and post-editing in Translation Studies
The tasks performed by professional translators have changed over the years, most
recently due to the acceptance that MT has achieved a level of quality that make it usable as
an input for human translation. This section revisits the definitions of translation, revision and
post-editing as professional services and tasks determined by constraints of time.
The processes that are performed as part of translation services are studied by a branch
of Translation Studies known as ‘translation process research’. This branch is devoted to
helping us achieve some of the fundamental goals of the discipline, namely to explain and
predict translator’s behaviour” (Carl, Bangalore, and Schaeffer 2016, 4). To do so, translation
process research needs observable and measurable data. Records of time provide hard data,
which may be correlated not only with notions of productivity, but with difficulty and
complexity, amongst others. Because of this, records of time and user activity data play an
important role in translation process research.
Time is also a guide to conceptualise processes. The most common descriptions of the
translation process divide it into three phases, which, although not necessarily performed one
after the other, take names that describe a temporal sequence, which starts with preparing,
then writing, and finally correcting what was written. The names given to these phases by
Carl, Dragsted, and Jakobsen (2011) are orientation, drafting and revision. Revision as a
separate process can also be divided into three phases, the first and last of which correspond
to the preparation and rechecking of the product of the middle phase (Huang and Minocha
2014).
This is a clear simplification of complex processes, but we may say that the main
difference between these two tasks is determined by time spent reading and time spent
writing: when translating, translators mainly write, with interruptions to read when they face
a problem, whereas revision is essentially a reading task (Mossop 2014, 115), interrupted
by writing when the reviser faces a problem.
Although translation and revision are complementary tasks, performed in sequence in
a translation process, the industry adopted a different term when MT content entered the
production workflow: post-editing (PE). This term appeared shortly after the development of
the first MT systems in order to differentiate the process from pre-editing of the source text,
4
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
but it became more common after the appearance of systems based on statistical processing
of bilingual data (García 2012).
PE is defined in one of the most important regulation documents of the industry as “to
edit and correct machine translation output” (ISO 2015, 2). This simple and common
definition (discussed in more detail below) conveys the notion that this process is easier and
takes less time than translation, since the translation effort has already been performed by
the machine. PE is thus seen as a form of revision, an a posteriori effort that validates the
achievements of the MT production and corrects the mistakes.
This view of PE as a simple process is not often discussed, but it is at the root of
contradictions in professional translation, a world which pushes for increases in productivity
while expecting no effects on output or quality. For example, the claim that PE should be done
“in as few edits as possible” (Mesa-Lao 2013) is not often criticised. However, unless we add
productivity requirements from the industry, there is no intrinsic feature in PE that requires
this constraint. The contradictory message in PE seems to be: ‘to produce more, don’t do so
much’.
This type of constraint, in the form of operational instructions passed down to
translators, is not necessary if we consider that efficiency has always been part and parcel of
translation as a service. Juan Sager’s study of the automation of communication processes
(1993) explores a view by which translation (a secondary semiotic process) needs to be
efficient to justify its presence in an industrial environment, in which everything is set to
maximise production. This view of professional, or industrial, translation, as being driven by
time efficiency is confirmed when we look at the adaptation of professional translators to the
different waves of technology that occupied their desktops in the last thirty years. For
translators to remain active in the industry, they must embrace all forms of technology that
help speed up processes, offer more resources, or open up new markets and work
opportunities. Nevertheless, as we will see in the following sections, the focus on productivity
often comes from, or brings with it, a devaluation of the work done by translators.
2.2 Translation, revision and post-editing viewed by the industry
The term ‘localization industry’ describes a sophisticated, highly technological, highly
competitive, dynamic and future-ready area of business, according to sources such as Slator
5
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
or Common Sense Advisory. This industry determines a big part of the current practices in
translation, and the impact of its communication can be observed in the frequency by which
these sources are cited in academic papers in Translation Studies and even in translator
training programmes. This section analyses how this industry, namely through its own
regulation documents, describes the way translators work, as a contribution to a discussion
on the impact these descriptions have on the global perception of the value of translation.
The first global industry standard in translation services, ISO 17100:2015, is an
instrument that serves one main purpose: to regulate the provision of services in the
localization industry by specifying process steps and creating patterns that enable objective
measurement of quality in different levels of service. It would be worth analysing the specific
vision promulgated in the document of all the work performed by translators, but we will only
analyse what is said about revision and PE in the text.
The standard specifies that revision defined as: “bilingual examination of target
language content against source language content for its suitability for the agreed purpose”
(ISO 2015, 2) is one of the required stages in any translation process and that it must be
performed by a person different from the translator. The standard also considers the
possibility that revision may be performed as a standalone added-value service, appearing in
a list of such services in one of the annexes.
The description of revision in section 5.3.3 of the standard determines that “The reviser
shall examine the target language content against the source language content for any errors
and other issues, and its suitability for purpose. This shall include comparison of the source
and target language content for the aspects listed in 5.3.1(ISO 2015, 1011). Section 5.3.1
refers to the translation stage of the translation process, and the eight aspects that are listed
there include: compliance to terminology and style guides, semantic accuracy, appropriate
syntax and conventions, cohesion and phraseology, locale, formatting, audience and purpose.
The section on revision concludes by adding that the process should be repeated “until the
reviser and the TSP (Translation Service Provider) are satisfied” and it adds in a note that
“corrections can include retranslation” (ISO 2015, 11).
So, it appears that time spent on revision is justified because it is required to repeat the
reading part of the translation process: a ‘second pair of eyes’ rethinks what the translator
did and checks the same details the translator focused on. In this description of the work of a
6
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
reviser, their intervention is not complementary to translation, but a reinforcement of
translation, a repetition performed until the reviser and the TSP are happy. In the standard,
there are no procedures specific to revision, such as whether and when the reviser should
communicate with the translator (to handle interpretation difficulties or divergences of
opinion), the need for the reviser to read the target, the source text or both more than once,
or the need to do particular checks and use particular tools, all tasks which are specific to
revision, and complementary to translation.
The standard clarifies that it does not encompass PE. In a note under the definition of
‘post-edit’, it is admitted that a translator may independently resort to using MT output, but
when translators do so, they are still understood to be translating (ISO 2015, 2). PE is,
however, included in the list of added-value services that may be offered by the TSP. This
means that PE only exists when the TSP expressly determines that this is the service offered;
in any other situation, although the translator may be editing and correcting MT output (as
PE is defined in the standard), they are not post-editing, but translating.
Two years after the publication of the standard on translation, a specific standard for
PE was created: ISO 18587:2017 (2017). The reasons for the creation of this standard are
presented as follows:
The use of machine translation (MT) systems to meet the needs of an
increasingly demanding translation and localization industry has been gaining
ground. Many translation service providers (TSPs) and clients have come to realize
that the use of such systems is a viable solution for translating projects that need
to be completed within a very tight time frame and/or with a reduced budget.
(ISO 2017, v my emphasis)
Concerns about the viability of utilising MT output may have been discarded because of
a general acceptance that PE was the only solution to answer the need to reduce time frames
and budgets. Thus, time and money are presented as the defining factors of PE, the
differentiators from previous forms and modes of translation production. But, as the
following analysis of the contents of the standard shows, PE is not as simple a solution as its
7
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
definitions tend to show. When we compare ISO18587:2017 with ISO17100:2015, the added
complexity of PE becomes evident.
ISO18587:2017 uses the same definition of PE as the previous standard: “to edit and
correct machine translation output” (ISO 2017, 2). It appears that PE is a simpler task than
revision, as defined in ISO17100:2015, since it does not require a “bilingual examination of
two types of content for suitability to an agreed purpose.
A note placed after this definition explains that the 2017 document does not include
the reference which appeared in ISO17100:2015 to the cases in which translators resort to
MT by choice. The standard does not clarify how this type of situation should now be
considered, and by which of the two documents it is covered: by the translation or by the PE
standard. This is a serious omission in a document that aims at providing unequivocal
descriptions of services, so that these can be offered in a comparable way.
The introduction to Section 4 of the standard on PE shows a major difference between
translation and PE, which is the fact that PE “involves three texts: the source text, the MT
output and the final target text” (ISO18587:2017, 5). The purpose of this task was also
extended to include more complex forms of reading and writing, including “checking [the MT
output] accuracy and comprehensibility, improving the text, making the text more readable,
and correcting errors” (2017, 5).
Section 4.3 of ISO18587:2017 contains three sub-sections, one for objectives, another
for requirements, and a third one for tasks. The first two sub-sections contain lists of three
objectives and nine requirements (of which three are optional), which expand the eight
‘aspects’ included in section 5.3.1 of ISO17100:2015. New items include “comprehensibility
of the post-edited MT output” and “compliance with post-editing guidelines”, which appear
next to repeated ones, such as compliance with client style guides (2017, 6). Section 4.3.3
presents three sets of tasks that should be performed by the post-editor: reading the MT
output and evaluating whether a reformulation of the target is necessary; using the source
language content as reference in order to understand and, if necessary, correct the target
language content; and producing target language content, either from existing elements in
the MT output or providing a new translation” (2017, 7). These tasks show a very particular
form of reading: reading, evaluating, and using a reference to understand and decide if
corrections are necessary. Besides, the list clarifies that the content delivered at the end of
8
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
the process is produced, not by the MT system, but by the post-editor, who is the one
responsible for the result of the process.
Furthermore, this standard includes an extra section, where eight different
requirements for full PE are presented. The aim of this level of PE is “to produce an output
which is indistinguishable from human translation output”, but it is recommended that “post-
editors use as much MT output as possible” (2017, 8). The latter expression, “using as much
MT output as possible” also appears in the requirements of ‘light PE’, in annex B of the
standard (2017, 10).
This description of the PE process and its requirements contradicts the notion that PE is
a simple, fast and cheap task. In fact, it describes a more complex form of work than revision
of human translation.
Industry standards are associated with production metrics, but the standards of the
localisation industry do not provide a safe foundation for the creation of metrics, which could,
for example, allow providers to assess thresholds, possibly time-based, for the classification
of PE as viable. One of the illusively quantifiable recommendations in the standard is the
expression related to the use of MT output not only ‘possible’ is a vague word, but it has no
applicability in reality. In fact, the industry has not attempted to implement methods to
control the amount of MT output left unedited, as there is no clear correlation between this
constraint and time saved or efficiency gained. Besides this, the standards add items from
which it is very difficult to create objective control metrics, such as comprehensibility,
readability, or indistinguishability from human translation.
ISO standards are not the only industry documents that send a mixed message about
simplicity or complexity of PE, with a negative impact on the perception of the value of work
done by translators.
On the website of TAUS (Translation Automation User Society), a think-tank that has
driven the automation of processes in the translation industry, there is a longer definition and
description of PE. The definition includes details of the tasks involved in the correction of MT
content, such as the correction of “recurring and predictable errors” (TAUS 2016a).
Furthermore, it presents a list of strategies which, although seen as time-reduction
techniques, add to the requirements of the tasks. Examples of these are: “keeping changes to
9
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
a minimum”, “applying quick fixes”, “automating certain word-processing operations”, or
“providing regular feedback to the MT team”.
TAUS has also been a champion of a flexible and dynamic quality-oriented view of PE,
namely by referring to the purposes of full PE and light PE, one aiming at “human-like quality”,
the latter at “good enough quality” (TAUS 2016a). Since the aim of PE is to save time, and
since human translators need only to be given enough time to produce human-like
translation, the focus of PE guidelines is mostly on the need to avoid over-correcting, and not
waste time on details. TAUS PE guidelines always include “as much MT output as possible”,
no matter what the intended level of quality is (Massardo et al 2016).
Definitions and descriptions of tasks, processes, strategies and levels of quality contain
or reveal assumptions about time required in production. As we have seen in this section,
time constraints for translators became especially visible with PE. In the next section, we
explore more effects of time in this professional context.
3 Effects of time in professional translation
This section focuses on direct consequences of the pressure of time over remuneration,
training and productivity control by professional translators.
3.1 Evolution of prices over time
The most reliable data on product and service prices is collected by official finance
departments in central governments. In 2008, at a conference of the Association of
Translation Companies of the UK, Douglas Lawrence presented an analysis of the relation
between global inflation and translation rates using such data (Lawrence, 2008). Figure 1
presents an update of his analysis, by collecting similar data for a period of twenty years.
1
1
The data was collected from the Office for National Statistics of the UK (2018). The data was normalised
with the base values (100) set in year 2000.
10
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
Figure 1: Inflation rates in the UK (1998-2018).
Figure 1 shows that the average item bought with 100 GBP in 2000 cost around 130 GBP
in 2018, while the money a translator earned was virtually the same. Put another way, if their
production has been consistent since 2000, translators can only buy two thirds of what they
could 18 years ago. This loss of buying power is down only to the policy of the industry of
selling based on cost control, rendering it incapable of accompanying natural inflation with
growing rates.
For translators, the effect of the mere passing of time is a real loss of earnings. It is
important to note that this data has been collected across the industry, not just at specific
points in the supply chain. This means that, because of the complexity and dependency of
relations in the supply chain, many translators, being links with reduced agency (Kinnunen
and Koskinen 2010), do not have the capacity to negotiate increases to their own rates.
Productivity increases brought by PE have been regularly associated with a reduction in
rates (DePalma 2013). What makes it worse, there are constant signs of a tendency not to
increase but to reduce the industry rates (Slator 2019, 23).
The pressure to increase earnings in line with inflation when rates are static may only
be compensated for by increases in productivity and efficiency, for which technological aids
11
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
may be beneficial. But this, in turn, brings with it a pressure towards constant spending, to
update to new technologies and to upgrade skills.
Other effects of the passage of time are not so visible, or easy to analyse. Translator
specialisation should be accompanied by a growing recognition of their skills, a progression
in their careers and incomes that would compensate for growing requirements and
investments. Can the industry claim that it is offering these opportunities of growth to a
skilled workforce? Or should the industry be concerned with how growing disappointment
and loss of talent across translator communities may affect the sustainability of the business?
Many other questions are worth looking at, like whether translation providers are accessing
the advantages of accumulated knowledge, only achievable with a continuous provision of
the same services. Some of these are related to the next topic, training.
3.2 Training for saving time
Although most of the literature on training for PE is focused on language issues, it also
includes references to training for “rapid and minimal PE” (Depraetere 2010), to the
usefulness of students to “examine their edits and PE times” (Koponen 2015), to the
requirement for students to know how to do PE “according to the quality and productivity
objectives” (EMT Expert Group 2017), or to help students “identify the point beyond which
editing is not necessary” (Guerberof Arenas and Moorkens 2019). PE training is also
increasingly inspired by industry practices, with, for example, in-class project management
tasks or PE guidelines (Guerberof Arenas and Moorkens 2019).
While time management is incorporated into translator and PE training, it creates
conflicts with traditional skills, like giving attention to detail and always questioning simple
solutions. Flanagan and Christensen (2014), for example, identified a major contradiction
between industry guidelines that call for keeping as much output as possible, and the ones
that require attention to detail. Their suggestion is that the requirement to respect the MT
output must be placed upfront, in opposition to the others. However, there are no studies on
the implications of this strategy in practice.
This contradiction in the instructions and the need to manage contradictory tensions
justifies much of the resistance of translators to PE (Moorkens et al. 2015). To oppose that
resistance, translator trainers suggest that PE should be incorporated into training as part of
12
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
a strategy to empower translators, as opposed to devaluing their work (Kenny and Doherty
2014). All this means investing time and money in specialising translators, but it is easy to see
why this is necessary.
To respond adequately to the requirements of the industry, a translator needs specific
skills. Post-editors can only achieve the productivity gains expected by the industry when they
are able to anticipate the difficulties they will face. For this, they need to devote time to
understanding and adjusting to different MT systems, since their outputs are different. They
also need to acquire skills to automate and accelerate the use of different translation tools,
and to adapt to time-consuming systems of communication with development teams. On top
of that, they need to develop quick reading and error identification skills, something which
can only be achieved with long practice, confidence, and a good knowledge of the technical
areas in which they work. The contradiction is in the fact that, for processes to be simple,
translators need to make complex decisions and control a demanding environment. All these
factors should be taken into consideration when measuring productivity gains, as we see next.
3.3 Do translators save time with PE?
As explained in section 2.2, PE has been proposed as a time-saving strategy. Various
studies have focused on productivity gains from PE, confirming this, in one way or another,
(Vasconcellos 1987; Zhechev 2014; Koponen 2016; Castilho et al. 2017). Most of these studies
either describe very specific work environments, or they isolate the production task from its
usual industrial environment, giving test participants a text to post-edit, from beginning to
end, with very simple and restricted instructions. However, as one may see in the examples
discussed in this section, professional PE is frequently done in circumstances in which internal
and external factors negatively affect any increase in productivity brought by technological
advances.
Krings seminal study on PE reveals that this may be considered as a more complex task
than translation itself. For example, medium-quality MT implies an increase in reading time,
because a decision must be made on whether or not to retain and edit the MT output (Krings
2001, 319). The technical conditions of Krings study were not the same translators use today,
and subsequent studies have analysed how new tools inundate translators with simultaneous
13
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
sources of data that interfere with mental processes and slow down decisions (Christensen
and Schjoldager 2011; Austermühl 2013; Alves et al. 2016).
PE is often done by combining MT with matches from Translation Memories (TM).
Several studies address the balance between the compared quality and usefulness of these
two resources, but most conclude that it is not possible to automate the decision on which to
prioritise, especially for intermediate levels of TM matching (Guerberof Arenas 2008;
Moorkens and Way 2016).
Localisation projects (in which, for example, strings of software are translated) have
very steep learning curves. One of the factors contributing to this is interface design: each
tool has different keyboard shortcuts and automation techniques that translators need to
command. The second factor is project documentation and support resources, which may
include instructions on language-bound conventions and corporate communication
guidelines that translators need to learn and follow, besides several glossaries and
terminological resources that need to be consulted in the same project (Désilets et al 2009).
Figure 2 is an illustration of a translation tool used by professional translators in a real
work scenario.
2
This screenshot illustrates part of a localisation project of software strings
from English into European Portuguese. Translators view all strings in a sequence. For some
of the segments they have MT suggestions, while for others they have matches from TMs.
The reliability of these resources varies a lot, and so do the reading and editing times
associated with them.
When there are conflicting sources of information and the tool does not provide enough
context, translators need to balance the amount of time they spend on each string,
sometimes devoting time to editing just one word, and sometimes recognising that the string
needs to be fully retranslated. As in any translation task, this is a very subjective decision, and
when time is the most important decision factor, experienced translators and novice alike are
2
Note: The picture was collected by asking a translator to capture a random screenshot of his desktop.
There was no intention to capture a specific example or particular problem. This collection method does not
give indications, by itself, of whether the example discussed here represents a very typical or a very special
situation. So, there are no claims of the value of the example in terms of representing typical work conditions.
For confidentiality reasons, portions of the screenshot have been erased and anonymised.
14
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
bound to accept reasonable suggestions, just because they cannot afford to wonder whether
they would find the correct alternative, or just a better one.
Furthermore, in many modern collaborative production environments, translation
decisions are less likely to be individual, with real-time updates to the contents being
translated, in a more or less visible coordination with other remote workers. One of the tasks
translators need to perform in these production environments is the provision of structured
feedback to MT teams, for system improvement, as mentioned in section 2.2. Modern
workflows also incorporate quality evaluation procedures, which follow different error
categorisation models, such as TAUS Quality Dashboard (TAUS 2016b). All these factors affect
the decision process, sometimes each word choice, as the example illustrated in Figure 2
allows us to discuss.
Figure 2: Post-editing MT and TM strings.
One may appreciate the complexity of the interface illustrated, with different panes,
each with several columns of information, over which the eyes of the translator wander, using
each for its different purpose, as they move through the assignment, translating or editing
one short string after another. In the specific situation caught in this screenshot, the translator
had received a suggestion for the translation into Portuguese of the string “Great! Your PIN
has been changed.” [(1) in the editor pane, centre of screen]. The translation suggestion is:
“Ótimo! O PIN foi alterado.” Although this is a perfectly acceptable translation, they decided
to use the concordance feature to search the expression “your PIN”. One may wonder why
devote valuable time to researching such a straightforward expression. However, as one may
15
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
see in the three resources provided by the tool, there are several inconsistencies in the
suggested translations into Portuguese. The results of the concordance search [(2) above,
right], the sequence of strings they were translating [(3) above, centre] and the results of the
TM [(4) below] all show strings with only the definite article (“o PIN” the PIN), while others
use the possessive (seu/sua) after the article (“o seu PIN” your PIN).
3
After they identified
this inconsistency, they must make the decision to insert or not to insert the possessive into
this string.
The factors that weigh on this decision are varied, one of them being the time the
translator can spend editing this string. But the translator also knows they are only working
on a fraction of the whole project, so consistency within the projects is also a concern.
Furthermore, they need to consider error feedback and evaluation. On the left of the screen
(5), there is a box to classify each edit in terms of different quality criteria. If they decide to
insert the possessive, they may classify the original MT suggestion as containing a
‘grammatical error’, since it omits a function word. But they may consider that the omission
of the possessive has no serious impact, as it is disambiguated by the context of use of the
string, and thus decide to maintain the translated string as suggested by the MT engine.
Finally, they must consider the odds that other translators working on the project make
the same type of change, and, most importantly, how the quality evaluator will consider this
issue. If the translator’s decision does not agree with the evaluator’s, they may be penalised.
It may sound like the use of possessives is a very small and subjective detail, which does
not really affect the quality of a project. This could even be considered a good example of
how translators can save time by not dealing with such details in light PE projects. However,
even light PE projects can be subject to style guides (ISO 2017). The use of possessives and
articles is often covered in Portuguese style guides, because of differences in their use in the
European and Brazilian varieties of the language. In fact, when MT engines are trained with
texts from both varieties, as they so often are, or with large amounts of data from inconsistent
legacy TMs, these are exactly the details that post-editors need to check. A divergence over a
word, in a situation like the one we describe here, may result in a negative classification,
3
In European Portuguese, the possessive is always used together with the article. In Brazilian Portuguese,
there is the option to only use the possessive.
16
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
according to different criteria: besides grammar, an evaluator may classify the issue as a
failure to consult (or interpret) corporate style guides, as a translation which is inconsistent
with the rest of the project, or as a disrespect to local conventions.
This example shows a typical situation in which spending too much time may mean
losing money, but spending too little time may lead to a lower quality evaluation. In real life,
translators need to find the balance between identifying the devil in the detail and the
profitability of each decision. If translators privilege not wasting time as a fundamental
decision factor, the whole notion of translation quality naturally shifts from ‘language quality’
to a more dynamic type of quality, but this dynamic quality does not necessarily reflect all
dimensions of the decision process in a professional context.
In these demanding professional contexts, translators see themselves as the sole
guarantors of language quality. They tend, therefore, to focus more on language issues, while
expecting that the tools that they use provide an efficient support for the decisions they need
to make, more than just adding functionality and complexity to each decision.
Replacing TM fuzzy matches with MT output for PE, does not, per se, guarantee an
increase in productivity, especially when each decision depends on so many factors. In fact,
as we have seen, it may even be associated with added tasks and increased complexity.
Translators who are experienced with localization projects have been working for years with
fragmented content, such as out-of-context fuzzy matches, and it is their accumulated
expertise that allows them to deliver high quality translations in a short time, even when they
post-edit.
4 The different faces of time in translation
This section first zooms in on uses of time as a metric of productivity, in translation
research and the industry. Then, it zooms out to factors which are not often used to moderate
the conclusions taken from the study of these metrics, because they cannot be observed at
the production desks.
4.1 Time as a metric of productivity
Translation Studies has researched the effects of time at different levels. One of these
is constraints on translation performance or translated text quality (DeRooze 2003;
17
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
Khalzanova 2008; Bayer-Hohenwarter 2009; Jiménez-Crespo 2012). Time also plays an
important role in describing translation effort (Krings 2001; Koponen 2016), and current
translation evaluation systems combine temporal effort with error typologies (Daems et al
2017; van Egdom et al 2018; Moorkens et al 2018).
The focus on productivity and efficiency in academic reflections on professional
translation has been driven by the discourse of the industry. The localisation industry
identifies itself as a “data-driven industry (Joscelyne 2018), with strategies and definitions of
quality being supported by solid and reliable data. On its website, TAUS advertises average
productivity rates that raise the demands on translators to a hard-to-reach bar. In one of its
most recent reports, it presents an average productivity for the industry that has fluctuated
from 848 to 2011 words per hour, with a global average of 1575 words/hour (TAUS 2020). As
a comparison, commonly-accepted figures for translation output are between 2000 and 3000
words for a whole day (Common Sense Advisory 2019). The global average figure of 1575
words per hour is presented on the opening page of the TAUS Quality Dashboard, with no
notes or comments, but the reality is not so simple.
These productivity rates refer to all words processed by a translator, using a TM or a MT
system, even if they do not edit them. So, it includes all TM full matches that are just read and
validated, and all MT suggestions that are considered as having ‘good enough’ or ‘human-like’
quality. In the same site, if you filter the results to include only edited segments, the average
reduces to 665 words per hour. Beyond time, the only other factor considered in the TAUS
report is ‘edit density’, or the ratio of edited words per total words, another type of data that
is easy to collect from keyboard logs. TAUS considers that this data is reliable enough to be
associated with the purposes of standardization, efficiency gains, and risk management. It
also claims that this data provides the business intelligence necessary to understand
translation effort and even to manage quality: “From our interaction with the field, we get a
better understanding of the KPIs [Key Performance Indicators] that quality managers would
like to track. Edit density and productivity turn out to be meaningful data points to estimate
translation effort.” (TAUS 2019, 1)
So, it seems that the metrics of time are turning the industry into a race against the
clock, with competitors comparing each other’s throughput and quality scores. This
competitive environment takes its toll on the production lines of the industry.
18
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
4.2 Time as a definer of labour
TAUS has also pushed the narrative of the wide availability of translation, based on MT
as a technology that enables the offer of free translation. This narrative is known as
“translation as commodity” (TAUS 2012), but this discourse is also contaminated by the
‘commoditisation’ of human work, as one of the features of modern digital labour (Wood et
al 2019). This section reflects on tensions in the labour market of the localisation industry.
The expression “time is of the essence” has been used to describe the competitive
localisation industry (Mazur 2007, 341). This competitiveness requires a flexible workforce,
which explains why this industry’s workforce is mainly freelance (Pielmeier and O’Mara 2020),
vulnerable to time and price pressures, especially when working in virtual marketplaces
(García 2015).
Freelance translators feel the need to make decisions quickly for several reasons: to
maintain their positions in a very competitive market, to guarantee the reputation of efficient
service providers, and, most important, to make sure their output converts into reasonable
income. Adapting in order to maintain the productivity that is required for a financially viable
life means accepting new tasks, like PE, and performing them in a way that maximises
efficiency while not deteriorating the quality of outputs. As freelancers, they are the only ones
managing the value of their time, navigating among lower paid but time-consuming tasks,
such as editing high fuzzy matches, revising or post-editing MT text, dedicating themselves to
administrative and managerial tasks, while also considering labour-related rights, such as
holidays and sick leave (Moorkens 2017).
As we have seen, it is very hard for freelance translators to negotiate their rates. This
extends to the difficulty in establishing a proportional relation between the rates and the time
dedicated to each task, which could be achieved with time-based charges. But the discussion
about advantages and disadvantages of this method of payment is still an open one (TAUS
2009; Milan 2016; FIT 2017).
On top of the difficulties in changing the habits of a whole industry, translation
companies may resist this change because they need to be able to anticipate costs, so as to
enable swift quotations and applications for tenders. Besides, time-based prices reflect a
trust- and expertise-based relationship between the provider and the buyer, something which
19
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
is difficult to maintain in an industry that is increasingly defined by its reliance on data and
automation (Joscelyne 2018).
A recent and extensive report of the translation supply chain identified only 13% of
translators with an in-house position, only half of which in localisation companies (Pielmeier
and O’Mara 2020). The fact that most translation work is outsourced is an indication that its
perceived market and growth value is low. In other words, the intrinsic value of translation
time is not worth the investment in in-house skills acquisition.
Time spent on the job is, together with task and autonomy, one of the factors that
defines a labour relation, one that used to have the form of a work contract. (Manyika et al.
2016) But things have changed and work relations in the ‘gig economy’ are more flexible.
Beyond full-time and part-time jobs, ‘freelance work’ is gradually becoming ‘gig work’.
The vendor model that characterises the localisation industry (Moorkens 2017) shares
tensions with other professions in the gig economy. A debate recently started in the US about
the impact of labour legislation on new forms of work, and unsurprisingly this affected the
definition of work relations in the localisation industry (Marking 2019). This debate promises
to be long and controversial.
4.3 Time as an opaque metric for value
Research that aims to be rigorous may be enchanted by hard data. However, we cannot
be blinded by numbers. Numbers, as hard data, may not be fully transparent they may be
interpreted as showing something, but sometimes they just make themselves visible,
obscuring other ways of looking at reality.
When we look at very objective values, such as the comparison of the amount of words
an MT system such as Google Translate produces in a day and the number of words all human
translators produce in a year (Pym 2019), we must take a step back and ask what this means
in terms of the value of translators’ work.
With metrics and figures like these, it is not surprising that human translation is seen as
a time-consuming and inefficient task. Nevertheless, translation has kept its relevance in the
professional world, having even expanded its presence, due to a progressive transformation
into a highly technical and specialised task. This is consonant with Juan Sager’s description of
the role of translation in industrial settings, as described in the introduction to this article:
20
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
translation’s value is intrinsic to its role and it is reinforced by the intrinsic need to be efficient,
to overcome its secondary role in a communication act.
Technology seems to play a double role in this context: by adding technological
expertise and accompanying the speed of evolution of the industry, translation shows its
value, but since technology is used to reduce costs of production, it creates the perception
that it is the technology that contains that value. Both views are related to a one-dimensional
view of the time-saving function of technology: faster means cheaper, while faster also
implies greater effort and complexity.
For time to be a good metric for value, it must be related not only to cost, but to other
factors, such as the complexity of the task, the qualifications of producers, not to speak of
notions of effort, risk and quality. All of these factors are very difficult to observe and measure
in an objective way, but all of them contribute to the global perception of the value of
translation.
5 Conclusions
This article covers different perspectives of professional translation, guided by the
interaction between time and money and their relation to perceptions of the value of
translation.
The article starts by challenging some of the narratives that have been built around task
definitions which appear to be simple, but which are intrinsically complex. A main focus of
interest is placed on PE, a task that reveals its intricacies when we look at the documents the
industry produced and we reflect on the processes happening on the translators’ desks.
The article proposes that PE cannot go on being seen essentially as a time- and cost-
saving strategy, because this is hampering the general perception of its specialised and expert
dimension. As a new type of translation, PE represents a step forward in achieving the
efficiency that human translation has always envisaged. For it to be effective, PE demands
complex decision-making, which reinforces the need for recognition of translators as
specialised knowledge workers.
Guided by the reflection on the value of time, the article calls into the foreground data
that shows that, as time goes by, and prices for translation services do not increase, the value
of time spent in translation is inevitably depreciated. The news from the industry show that
translation producers are the ones paying the price of this depreciation. In fact, work at
21
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
translator’s desk, instead of getting faster because it is simplified, tends to grow ever more
complex. A critical view of the solution to gaining back value (by resorting to technology) also
brings up problems: the disruption created by MT accelerated even further the devaluation
of human translation. The narrative of productivity gains achieved with PE is confronted with
the delays in micro-decisions and the impact these may have on the expected gains or on the
control of the quality of the outcomes. For a workforce that manages its income on its own,
time constraints and productivity requirements often create an unbearable pressure.
By talking so much about the value of time, we run the risk of overestimating it. That is
often the case, when research uses time as the most objective metric available, leaving other
external and internal factors aside to measure productivity and efficiency. Useful as these
approaches may be, it is important to highlight how time control can be misused and
misinterpreted as a simple solution to complex cases.
The analysis of different dimensions of time in professional contexts brings to light
aspects that contribute to the perception of value of work. This article does not exhaust the
theme, because, as we have stressed, this is just one of the elements in such an evaluation,
but it reveals its importance to the understanding of the evolution of translation work.
Acknowledgements
This Project has received funding from the European Union’s Horizon 2020 research and
innovation programme under the EDGE COFUND Marie Skłodowska-Curie Grant Agreement
no. 713567. This publication has emanated from research supported in part by a research
grant from Science Foundation Ireland (SFI) under Grant Number 13/RC/2077.
List of references
Alves, Fábio, Arlene Koglin, Bartolomé Mesa-Lao, Mercedes García Martínez, Norma B.
de Lima Fonseca, Arthur de Melo Sá, José Luiz Gonçalves, Karina Sarto Szpak, Kyoko Sekino,
and Marceli Aquino. 2016. “Analysing the Impact of Interactive Machine Translation on Post-
Editing Effort.” In New Directions in Empirical Translation Process Research, edited by Michael
Carl, Srinivas Bangalore, and Moritz Schaeffer. Heidelberg: Springer International Publishing
Switzerland. https://doi.org/10.1007/978-3-319-20358-4_4.
Austermühl, Frank. 2013. “Future (and Not-so-Future) Trends in the Teaching of
Translation Technology.” Tradumàtica: Tecnologies de La Traduciò, no. 11: 32637.
22
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
Bayer-Hohenwarter, Gerrit. 2009. “Methodological Reflections on the Experimental
Design of Time-Pressure Studies.” Across Languages and Cultures 10 (2): 193206.
https://doi.org/10.1556/Acr.10.2009.2.2.
Carl, Michael, Srinivas Bangalore, and Moritz Schaeffer, eds. 2016. New Directions in
Empirical Translation Process Research. Heidelberg: Springer International Publishing
Switzerland. https://doi.org/10.1007/978-3-319-20358-4.
Carl, Michael, Barbara Dragsted, and Arnt Lykke Jakobsen. 2011. “On the Systematicity
of Human Translation Processes.” In Tralogy 2011. Translation Careers and Technologies:
Convergence Points for the Future. Paris, France. http://research.cbs.dk/en/publications/on-
the-systematicity-of-human-translation-processes(0225f368-7eea-4ed2-8a35-
9a7d9efbce6c).html.
Castilho, Sheila, Joss Moorkens, Federico Gaspari, Iacer Calixto, John Tinsley, and Andy
Way. 2017. “Is Neural Machine Translation the New State of the Art?” The Prague Bulletin of
Mathematical Linguistics 108 (1). https://doi.org/10.1515/pralin-2017-0013.
Christensen, Tina Paulsen, and Anne Schjoldager. 2011. “The Impact of Translation-
Memory (TM) Technology on Cognitive Processes: Student-Translators’ Retrospective
Comments in an Online Questionnaire.” In Proceedings of the 8th International NLPCS
Workshop (Special Theme: Human-Machine Interaction in Translation), edited by Bernadette
Sharp, Michael Zock, Michael Carl, and Arnt Lykke Jakobsen. Copenhagen: Samfundslitteratur.
Common Sense Advisory. 2019. “Translation Industry Headed for a ‘Future Shock’
Scenario.” Common Sense Advisory. https://csa-research.com/More/Media/Press-
Releases/ArticleID/38/Translation-Industry-Headed-for-a-“Future-Shock”-Scenario.
Daems, Joke, Sonia Vandepitte, Robert J. Hartsuiker, and Lieve Macken. 2017.
“Translation Methods and Experience: A Comparative Analysis of Human Translation and
Post-Editing with Students and Professional Translators.” Meta: Journal Des Traducteurs. 62
(2): 245.
DePalma, Don. 2013. Post-editing in Practice. tcworld e-magazine.
http://www.tcworld.info/e-magazine/translation-and-localization/article/post-editing-in-
practice/
Depraetere, Ilse. 2010. “What counts as useful advice in a university post-editing
training context? Report on a case study.” In François Yvon and Viggo Hansen (eds) EAMT
2010: Proceedings of the 14th annual conference of the European association for machine
translation. EAMT.
Désilets, Alain, Christiane Melançon, Geneviève Patenaude, and Lousie Brunette. 2009.
How translators use tools and resources to solve translation problems: An ethnographic
study. In Machine Translation Summit XIIWorkshop: Beyond Translation Memories.
Ottawa, Canada. http://www.mt-archive.info/MTS-2009-Desilets-2.pdf
De Rooze, Bart. 2003. “La Traducción, Contra Reloj [Translation, against the Clock].” In
Actas Del I Congreso Internacional de La Asociación Ibérica de Estudios de Traducción e
Interpretación, edited by Ricardo Muñoz Martín, 5966. AIETI.
23
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
van Egdom, Gys-walt, Lucas Nunes Vieira, and Jakub Absolon. 2018. “Towards Testing
Post-Editing Performance: A Futureproof Diagnostic.” Tradumàtica: Tecnologies de La
Traducció. no. 16: 11424.
EMT Expert Group. 2017. European Master’s in Translation - Competence Framework
2017. European Master’s in Translation (EMT).
https://ec.europa.eu/info/sites/info/files/emt_competence_fwk_2017_en_web.pdf
Flanagan, Marian, and Tina Paulsen Christensen. 2014. “Testing Post-Editing Guidelines:
How Translation Trainees Interpret Them and How to Tailor Them for Translator Training
Purposes.” Interpreter and Translator Trainer 8 (2): 25775.
https://doi.org/10.1080/1750399X.2014.936111.
FIT. 2017. “FIT Position Paper on the Future for Professional Translators.” Paris.
Franklin, Benjamin. 1748. “Advice to a Young Tradesman”, In Works of the Late Doctor
Benjamin Franklin: consisting of his life, written by himself, together with essays, humorous,
moral & literary, chiefly in the manner of the Spectator (1793). P. Wogan, P. Byrne, J. Moore,
and W. Jones, Dublin. Consulted online at:
https://play.google.com/store/books/details?id=g-I4AQAAMAAJ&rdid=book-g-
I4AQAAMAAJ&rdot=1.
García, Ignacio. 2012. “A Brief History of Postediting and of Research on Postediting.”
Revista Anglo Saxonica 3: 293310.
———. 2015. “Cloud marketplaces: Procurement of translators in the age of social
media.” The Journal of Specialised Translation 23, 1838.
Guerberof Arenas, Ana. 2008. “Productivity and Quality in the Post-Editing of Outputs
from Translation Memories and Machine Translation.” Universitat Rovira i Virgili.
Guerberof Arenas, Ana, and Joss Moorkens. 2019. Machine translation and post-
editing training as part of a master’s programme. Jostrans Journal of Specialised
Translation. 31, 217238.
Guzmán, Rafael. 2007. “Manual MT Post-Editing: ‘If It’s Not Broken, Don’t Fix It.’”
Translation Journal 11 (4): 2126. http://translationjournal.net/journal/42mt.htm.
Huang, Jin, and Akshay Minocha. 2014. “Cognitive Process of Revision: The Behaviours
and the Motivations Behind.” In Translation in Transition - Between Cognition, Computing and
Technology. Copenhagen.
ISO. 2015. “ISO 17100:2015 Translation Services - Requirements for Translation
Services.” Geneva: International Organization for Standardization.
https://www.iso.org/standard/59149.html.
———. 2017. “ISO 18587:2017 - Translation Services - Post-Editing of Machine
Translation Output - Requirements.” Geneva: International Organization for Standardization.
https://www.iso.org/standard/62970.html.
Jiménez-Crespo, Miguel Ángel. 2012. “Translation under Pressure and the Web: A
Parallel Corpus-Study of Obama’s Inaugural Speech in the Online Media.” The International
Journal for Translation & Interpreting Research 4 (1): 5676. https://doi.org/10.12807/t.
24
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
Joscelyne, Andrew. 2018. “Translators in the Algorithmic Age: Briefing Based on the
TAUS Industry Leaders Forum in Amsterdam in June, 2018.” Amsterdam.
https://info.taus.net/translatorsin-%0Athe-algorithmic-age-full-version.
Kenny, Dorothy, and Stephen Doherty. 2014. “Statistical Machine Translation in the
Translation Curriculum: Overcoming Obstacles and Empowering Translators.Interpreter and
Translator Trainer 8 (2): 27694. https://doi.org/10.1080/1750399X.2014.936112.
Khalzanova, Serafima. 2008. “Revision and Time Constraints in Translation.”
Dissertation (Diploma of Advanced Studies). Universitat Rovira i Virgili.
Kinnunen, Tuija, and Kaisa Koskinen, eds. 2010. Translators’ Agency. Tampere
University Press.
Koponen, Maarit. 2015. “How to teach machine translation post-editing? Experiences
from a post-editing course.” In Sharon O’Brien and Michel Simard (eds). Proceedings of 4th
Workshop on Post-Editing Technology and Practice (WPTP4). Association for Machine
Translation in the Americas, 2-15.
Koponen, Maarit. 2016. “Machine Translation Post-Editing and Effort Empirical Studies
on the Post-Editing Process.” University of Helsinki.
Krings, Hans P. 2001. Repairing Texts: Empirical Investigations of Machine Translation
Post-Editing Processes. Edited by Geoffrey S. Koby. Kent, Ohio & London: The Kent State
University Press.
Lawrence, Doug. 2008. “Inflation Everywhere except in Our Pricing. Does It Matter
and What Can We Do about It?” ATC Annual Conference 2008 44 (0).
Manyika, James, Susan Lund, Jacques Bughin, Kelsey Robinson, Jan Mischke, and Deepa
Mahajan. 2016. Independent Work: Choice, Necessity, and the Gig Economy. McKinsey &
Company.
Marking, Marion. 2019. “Inside the Contractor Debate Affecting US Freelance
Translators, Interpreters, and LSPs.” Slator. 2019. https://slator.com/industry-news/inside-
the-contractor-debate-affecting-us-freelance-translators-interpreters-and-lsps/%0D.
Massardo, Isabella, Jaap van der Meer, Sharon O’Brien, Fred Hollowood, Nora
Aranberri, and Katrin Drescher. 2016. MT Post-Editing Guidelines. TAUS.
Mazur, Iwona. 2007. “The Metalanguage of Localization.” Target. International Journal
of Translation Studies 19 (2): 33757. https://doi.org/10.1075/target.19.2.11maz.
Mesa-Lao, Bartolomé. 2013. “Introduction to Post-Editing.” In SEECAT - Speech & Eye-
Tracking Enabled CAT. Copenhagen Business School.
Milan, John M. 2016. “An Hourly Fee for Translation?” ATA Chronicle, no. June: 28–29.
Moorkens, Joss. 2017. “Under Pressure: Translation in Times of Austerity.” Perspectives:
Studies in Translatology 25 (3): 46477. https://doi.org/10.1080/0907676X.2017.1285331.
Moorkens, Joss, Sharon O’Brien, Igor A L da Silva, Norma B. de Lima Fonseca, and Fábio
Alves. 2015. “Correlations of Perceived Post-Editing Effort with Measurements of Actual
Effort.” Machine Translation 29 (34): 26784. https://doi.org/10.1007/s10590-015-9175-2.
25
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
Moorkens, Joss, and Andy Way. 2016. “Comparing Translator Acceptability of TM and
SMT Outputs.” Baltic Journal of Modern Computing Modern Computing 4 (2): 14151.
Moorkens, Joss, Sheila Castilho, Federico Gaspari, and Stephen Doherty, eds. 2018.
Translation Quality Assessment : From Principles to Practice. Springer International Publishing
AG.
Mossop, Brian. 2014. Revising and Editing for Translators. 3rd ed. London & New York:
Routledge.
Office for National Statistics of the UK. 2018. “Inflation and Price Indices.”
https://www.ons.gov.uk/economy/inflationandpriceindices#timeseries.
Pielmeier, Hélène and Paul O’Mara. 2020. The state of the linguist supply chain.
Common Sense Advisory. www.csa-research.com
Pym, Anthony. 2019. “How automation through neural machine translation might
change the skill sets of translators.” http://usuaris.tinet.cat/apym/on-
line/translation/2019_NMT_socio_web.pdf.
Sager, Juan. 1993. Language Engineering and Translation: Consequences of Automation.
Amsterdam/Philadelphia: John Benjamins Publishing Company.
Schwarz, Marie-Claire Cruz. 2015. “What Kind of People Are Translators?” The Open
Mic. 2015. https://theopenmic.co/what-kind-of-people-are-translators/.
Slator. 2019. “Slator 2019 Language Industry Market Report.” Slator
TAUS. 2009. The innovation and interoperability roadmap for the translation industry.”
TAUS.
———. 2012. Moses: Commodity Creates Opportunity.”
https://www.taus.net/knowledgebase/index.php/Post-editing.
———. 2016a. “Post-Editing.” https://www.taus.net/knowledgebase/index.php/Post-
editing.
———. 2016b. “TAUS Quality Dashboard.”
https://evaluate.taus.net/component/rsfiles/download-
file/files?path=Reports%2FFree+Reports%2FQualityDashboardDocument-+March2016.pdf.
———. 2019. “DQF BI Bulletin - Q4 2018.” https://www.taus.net/dqf.
———. 2020. “TAUS Quality Dashboard.” https://qd.taus.net/ (consulted in 24 February
2020).
Vasconcellos, Muriel. 1987. “Post-Editing on-Screen: Machine Translation from Spanish
into English.” In Proceedings of the Conference Translating and the Computer 8, edited by
Catriona Picken, 13346. London: Aslib. http://www.mt-archive.info/Aslib-1986-
Vasconcellos.pdf.
Wood, Alex J. et al. 2019. ”Networked but Commodified: The (Dis)Embeddedness of
Digital Labour in the Gig Economy”, Sociology, 53(5), pp. 931–950. doi:
10.1177/0038038519828906.
Zhechev, Ventsislav. 2014. “Analysing the Post-Editing of Machine Translation at
Autodesk.” In Post-Editing of Machine Translation: Processes and Applications, edited by
26
-------------------------------------------------------------------------------------------------------
do Carmo, Félix. (2020). ‘Time is money’ and the value of translation. Translation Spaces, 9(1), 3557.
https:/doi.org/10.1075/ts.00020.car
Sharon O’Brien, Laura Winther Balling, Michael Carl, Lúcia Specia, and Michel Simard.
Newcastle upon Tyne: Cambridge Scholars Publishing.
https://doi.org/10.2507/26th.daaam.proceedings.070.
... The proposed frameworks assume MT and LLMs are sources of truth that sometimes fail and require humans as error fixers who need to compensate for this failure (do Carmo, 2020). This view places humans and machines in contending positions. ...
Article
Full-text available
The media localisation industry, as audiovisual translation (AVT) is commonly known, has undergone radical changes with the accelerated deployment of AI-powered solutions transforming production processes. This article examines how different stakeholders navigate this changing landscape: an industry promoting automation to meet growing demands, translators concerned about working conditions and professional sustainability, academia developing future professionals while researching technological impact, and audiences whose evolving expectations shape industry practices. Drawing on stakeholder theory, the analysis reveals how the sustainability of AVT requires balancing technological efficiency with human expertise. While AI tools promise faster turnaround times and reduced costs, their implementation must ensure long-term professional viability and translation quality. The article argues that successful integration of AI depends on creating value for all stakeholders through collaborative approaches that recognise translators’ agency and expertise. This requires conceptualising AI not merely as a cost-reduction tool but as part of a broader ecosystem where human expertise and technological capabilities complement each other to serve diverse global audiences. Lay summary Audiovisual translation is undergoing drastic changes due to the consolidation of an AI-powered media localisation industry. Escalating demands for faster content delivery, simultaneous releases in multiple languages, and changing production dynamics have been bolstered by an accelerated deployment of machine translation (MT) and AI technologies. This article assesses the impact of technology on media localisation by revising the perspectives of key stakeholders: industry representatives, professional translators, and academia. The review allows us to analyse the motivations and priorities of these stakeholders and, most importantly, the rationale guiding their positions. While the industry has embraced AI as an efficient response to the accelerated growth of media production and a solution to a declared shortage of professionals, professional associations have responded with strong warnings against post-editing and AI in media localisation. The magnitude and speed of recent changes have prompted primarily negative or cautious reactions from professionals. For them, the real issue lies in the deteriorating working conditions resulting from automated processes that ultimately lead to lower quality standards. As educators and researchers, we, as academics, are also in the midst of this evolving debate, grappling with the responsibility to quickly upskill ourselves to understand the nature and reach of the changes and to equip students with skills aligned with market demands; all this while instilling in them a critical awareness of these changes to ensure they can develop sustainable careers. After laying out the positions of these key stakeholders, the article proposes to encourage mutual recognition and awareness to foster understanding in an ever-evolving landscape. The potential of technology is evident, and its integration should not merely aim to replicate human work but should explore innovative ways to solve problems to address the complexities of creative translation processes. Ensuring the agency of human translators is recognised, the article constitutes a call to conceptualise processes that benefit from the expertise of human translators and successfully combine technological efficiency and human expertise to augment translation and expand the boundaries of media localisation.
... Heilinger, Kempt and Nagel (2023) have similarly questioned whether sustainable AI is possible in an economy targeted at perpetual growth. (2021), Moorkens (2020) and others about the sustainability of the language industry due to AI-enabled employment practices in which work is rigidly controlled and monitored, work processes (such as post-editing within limited translation platforms) are unilaterally imposed on translators and remuneration is dropping or not increasing in line with inflation (Vieira 2020;do Carmo 2020). There are anecdotal reports of translators leaving the industry and claims of a talent crunch in subtitling, one of the areas of the market most affected by these practices. ...
Chapter
Full-text available
Widespread disruption to the language industry from artificial intelligence (AI) such as machine translation (MT) has been predicted for many years, but now that these technologies are being deployed, the effects are varied and, at times, unexpected. Neural MT, in particular, can produce output of greater quality compared to previous MT paradigms, but not without errors, and the best way to interact with MT to produce quality translation is not entirely clear. The use of MT and other forms of AI in the language industry necessitates consideration of risk, of value and of environmental and social sustainability. In this chapter, we introduce definitions of AI and automation, follow developments in AI within the language industry, and then consider the direction in which these developments need to go and how we might get there.
Article
Full-text available
This article investigates how translation memories (TMs) can be created by translators or other language professionals in order to compile domain-specific parallel corpora, which can then be used in different scenarios, such as machine translation training and fine-tuning, TM leveraging, and/or large language model fine-tuning. The article introduces a semi-automatic TM preparation methodology that primarily leverages translation tools used by translators, in the interests of data quality and control by translators themselves. This semi-automatic methodology is then used to build a cardiology-based Turkish → English corpus from bilingual abstracts of Turkish cardiology journals. The resulting corpus, called TRENCARD Corpus, has approximately 800,000 source words and 50,000 sentences. Using this methodology, translators can build custom TMs in a reasonable time and use them in tasks requiring bilingual data.
Thesis
Full-text available
As Sundar Pichai, the CEO of Google, properly stated, “Artificial intelligence will have a more profound impact on humanity than fire, electricity and the internet.” As the mountainous wave of artificial intelligence (AI) and its related technologies lapped all domains, and translation is not an exception, a comparative analysis of AI-assisted translation and full human translation is of particular importance. The present research explores the differences between AI-assisted and full human translation in terms of time, quality, and cognitive load to specify the most optimal translation method. Waddington’s (2001) Method D was used to assess the quality of translations, and electroencephalography (EEG) which is a non-invasive neuroimaging technique was employed to measure cognitive load. Six translation studies students were recruited through convenience sampling to participate in the experiments at the National Brain Mapping Laboratory and translate two news texts from Persian into English, once by means of AI and once without AI or other translation machines. Results were the indication that AI reduced translation time, improved quality, and decreased cognitive load. However, when relative cognitive load—the amount of cognitive load in proportion to one unit of time and one unit of quality—was calculated, it was unexpectedly found that relative cognitive load was higher for AI-assisted translation (0.35 units). These insights do not deny the usefulness of AI in translation industry, they only suggest that its efficiency is limited under this research design.
Article
Full-text available
The language services industry has enjoyed consistent economic growth over the past 15 years, yet not all participants have reaped its benefits. Individual language professionals have faced persistently low rates of pay, a lack of social benefits, and reduced job security. This predicament has been exacerbated by disruptions brought by technologies such as machine translation and artificial intelligence, and the questionable business practices of language services providers. Existing research primarily focuses on economic and technological aspects, overlooking the experiences of linguists. This article seeks to broaden this area of study by compiling a comprehensive list of the bad business practices that individual practitioners experience. A survey of 682 freelance translators provides the foundation, yielding a catalogue of 17 detrimental practices along with their prevalence rates. The article also outlines proposals for addressing these issues and identifies potential avenues for future research in this domain.
Preprint
Full-text available
This article investigates how translation memories (TM) can be created by translators or other language professionals in order to compile domain-specific parallel corpora , which can then be used in different scenarios, such as machine translation training and fine-tuning, TM leveraging, and/or large language model fine-tuning. The article introduces a semi-automatic TM preparation methodology leveraging primarily translation tools used by translators in favor of data quality and control by the translators. This semi-automatic methodology is then used to build a cardiology-based Turkish -> English corpus from bilingual abstracts of Turkish cardiology journals. The resulting corpus called TRENCARD Corpus has approximately 800,000 source words and 50,000 sentences. Using this methodology, translators can build their custom TMs in a reasonable time and use them in their bilingual data requiring tasks.
Chapter
Globalisation and technological innovation have sparked significant structural changes in the language industry in the last decade. Faced with this reality, we find ourselves at a critical moment to reflect on the technological needs and trends of this dynamic sector. In order to provide as accurate a picture as possible, we have reviewed leading industry stakeholders’ publications on the state of the language industry, with a twofold aim: to identify (1) the existing professional profiles that require technology-related skills and know-how and (2) the current technological trends. All the publications reviewed confirm that the language industry is highly technologised and that traditional profiles now require high levels of technology-related skills. However, there is yet to be a clear answer as to where the profession is heading, given the relentless process of technological change taking place in the sector. Language service providers and independent professionals are well aware that new technologies may be developed that will have a disruptive effect on the way translation services are performed and delivered.
Thesis
Full-text available
Kumulative Habilitationsschrift zur Rolle der neuronalen maschinellen Übersetzung (NMÜ) als Werkzeug des fachübersetzerischen Handelns im modernen digitalisierten und datafizierten Fachübersetzungsprozess. Es handelt sich hierbei um die Mantelschrift, die die kumulative Leistung umschließt. Im vorderen Teil des Mantels wird zunächst ein größerer Kontext für die Betrachtung der neuronalen maschinellen Übersetzung als Werkzeug fachübersetzerischen Handelns aufgebaut und die vorliegende Schrift translationswissenschaftlich verortet. Im hinteren Mantelteil fließen die im vorderen Teil und in der kumulativen Leistung angestellten Überlegungen, ergänzt durch einige zusätzliche Gedanken, in einem Faktorenmodell des situierten NMÜ-gestützten Fachübersetzungsprozesses zusammen. In diesem Modell wird die „Passung des Werkzeugs, des Werkstücks und des Werkmeisters“ (Holz-Mänttäri 1984:136) in einem gegebenen soziotechnischen und sozioökonomischen Umfeld betrachtet.
Article
Full-text available
This article presents a description of a machine translation (MT) and post-editing course (PE), along with an MT project management module, that have been introduced in the Localisation Master’s programme at Universitat Autònoma de Barcelona in 2009 and in 2017 respectively. It covers the objectives and structure of the modules, as well as the theoretical and practical components. Additionally, it describes the project-based learning approach implemented in one of the modules, which seeks to foster creative and independent thinking, teamwork, and problem solving in unfamiliar situations, with a view to acquiring transferable skills that are likely to be in demand, regardless of the technological advances taking place in the translation industry.
Article
Full-text available
本文概述了一个用于评估译后编辑实践的诊断工具。尽管有许多诸如此类的例子, 但用实证研究作为评价基础并不常见。我们希望我们的工具能够帮助选择适合译后编辑项目的专业译员或学生,并检测译者的知识水平、能力,预测译者未来的工作态度。
Book
Full-text available
This is the first volume that brings together research and practice from academic and industry settings and a combination of human and machine translation evaluation. Its comprehensive collection of papers by leading experts in human and machine translation quality and evaluation who situate current developments and chart future trends fills a clear gap in the literature. This is critical to the successful integration of translation technologies in the industry today, where the lines between human and machine are becoming increasingly blurred by technology: this affects the whole translation landscape, from students and trainers to project managers and professionals, including in-house and freelance translators, as well as, of course, translation scholars and researchers. The editors have broad experience in translation quality evaluation research, including investigations into professional practice with qualitative and quantitative studies, and the contributors are leading experts in their respective fields, providing a unique set of complementary perspectives on human and machine translation quality and evaluation, combining theoretical and applied approaches.
Article
Full-text available
Machine-translated segments are increasingly included as fuzzy matches within the translation-memory systems in the localisation workflow. This study presents preliminary results on the correlation between these two types of segments in terms of productivity and final quality. In order to test these variables, we set up an experiment with a group of eight professional translators using an on-line post-editing tool and a statistical-based machine translation engine. The translators were asked to translate new, machine-translated and translation-memory segments from the 80-90 percent value range using a post-editing tool without actually knowing the origin of each segment, and to complete a questionnaire. The findings suggest that translators have higher productivity and quality when using machine-translated output than when processing fuzzy matches from translation memories. Furthermore, translators' technical experience seems to have an impact on productivity but not on quality.
Article
Full-text available
This paper proposes an approach to teaching translation technology that focus less on exposing students to ever more types of CAT tools than on two sets of meta-competences—revising skills and documentary research skills—and on the technologies that allow students to optimize these skills.
Article
Full-text available
This paper discusses neural machine translation (NMT), a new paradigm in the MT field, comparing the quality of NMT systems with statistical MT by describing three studies using automatic and human evaluation methods. Automatic evaluation results presented for NMT are very promising, however human evaluations show mixed results. We report increases in fluency but inconsistent results for adequacy and post-editing effort. NMT undoubtedly represents a step forward for the MT field, but one that the community should be careful not to oversell.
Article
Full-text available
The profession of translation is undergoing enormous change, expedited by the global recession that began in 2007-200­8. Government policies and an intensive focus on cost has resulted in a large scale move to freelance or contingent work, leaving the worker in a precarious position with regard to rights and undermining his or her agency. This shift is exemplified particularly by the vendor model widespread in specialised translation work. Related to the downward pressure on costs and productivity is the technologisation of translation, with translation tools becoming a necessity and new use cases being found for post-edited and raw machine translation. Despite the recession, continued growth has been reported for the language industry, and the outlook for employment in translation is positive. This paper looks at the background to the economic and technological changes to translation, attempts to put them into a wider context, and looks to the options available to translators to maximise their agency within the ‘global value chain’. Translators have little option but to embrace new competences, but also need to focus on their expertise to maximise leverage and agency.
Article
This article investigates the (dis)embeddedness of digital labour within the remote gig economy. We use interview and survey data to highlight how platform workers in Southeast Asia and Sub-Saharan Africa are normatively disembedded from social protections through a process of commodification. Normative disembeddedness leaves workers exposed to the vagaries of the external labour market due to an absence of labour regulations and rights. It also endangers social reproduction by limiting access to healthcare and requiring workers to engage in significant unpaid ‘work-for-labour’. However, we show that these workers are also simultaneously embedded within interpersonal networks of trust, which enable the work to be completed despite the low-trust nature of the gig economy. In bringing together the concepts of normative and network embeddedness, we reconnect the two sides of Polanyi’s thinking and demonstrate the value of an integrated understanding of Polanyi’s approach to embeddedness for understanding contemporary economic transformations.
Book
This volume provides a comprehensive introduction to the Translation Process Research Database (TPR-DB), which was compiled by the Centre for Research and Innovation in Translation and Technologies (CRITT). The TPR-DB is a unique resource featuring more than 500 hours of recorded translation process data, augmented with over 200 different rich annotations. Twelve chapters describe the diverse research directions this data can support, including the computational, statistical and psycholinguistic modeling of human translation processes. In the first chapters of this book, the reader is introduced to the CRITT TPR-DB. This is followed by two main parts, the first of which focuses on usability issues and details of implementing interactive machine translation. It also discusses the use of external resources and translator-information interaction. The second part addresses the cognitive and statistical modeling of human translation processes, including co-activation at the lexical, syntactic and discourse levels, translation literality, and various annotation schemata for the data.