Perspectives

Studies in Translation Theory and Practice

ISSN: 0907-676X (Print) 1747-6623 (Online) Journal homepage: http://www.tandfonline.com/loi/rmps20

How much would you like to pay? Reframing

and expanding the notion of translation quality
through crowdsourcing and volunteer approaches

Miguel A. Jiménez-Crespo

To cite this article: Miguel A. Jiménez-Crespo (2017) How much would you like to pay? Reframing
and expanding the notion of translation quality through crowdsourcing and volunteer approaches,
Perspectives, 25:3, 478-491, DOI: 10.1080/0907676X.2017.1285948

To link to this article: http://dx.doi.org/10.1080/0907676X.2017.1285948

Published online: 08 May 2017.

Submit your article to this journal

Article views: 94

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=rmps20

Download by: [National Taipei University of Technology] Date: 17 July 2017, At: 01:12
PERSPECTIVES, 2017
VOL. 25, NO. 3, 478–491
https://doi.org/10.1080/0907676X.2017.1285948

How much would you like to pay? Reframing and expanding

the notion of translation quality through crowdsourcing and
volunteer approaches
Miguel A. Jiménez-Crespo
Department of Spanish and Portuguese, The State University of New Jersey, New Brunswick, U.S.A.

ABSTRACT ARTICLE HISTORY

This paper examines the reconceptualization of translation quality Received 28 January 2016
since the emergence of new paid and volunteer crowdsourcing Accepted 25 November 2016
approaches both in the industry and in Translation Studies (TS). It
KEYWORDS
explores how economic considerations have led to a dynamic Translation quality; quality
conceptualization of translation quality. Its main focus is the impact levels; crowdsourcing; paid
of innovative crowdsourcing workflows and practices that currently crowdsourcing; economic
offer on free and paid models different quality tiers. In this process, turn
translation quality has been reconceptualized from a desirable,
static and high cost commodity that can be certified through
international standards, to a new dynamic construct in which the
fitness for purpose of the translation product, rather than quality, is
negotiated by different actors and through a wide range of process-
based factors that directly correlate to different prices charged for
the translation. The evolution of quality from static to dynamic
approaches will be explored. The article will end with a discussion
on whether the expansion of the notion of translation quality can
help consolidate the so-called economic turn in the discipline.

Introduction
Quality represents one of the most controversial issues in Translation Studies (TS). While
scholars often contend that a theoretical foundation is a prerequisite for assessment activi-
ties (House, 1977, 2014), industry approaches have been moving towards a dynamic per-
spective in which all participants in the translation event, such as clients, managers,
translators, and end users, can jointly decide what is a good enough level of quality or
the fitness for purpose of any translation (Drugan, 2013; Göroj, 2014a, 2014b; Gouadec,
2010). Many factors are behind this evolution. Firstly, the industry has long ago acknowl-
edged and accepted that it is impossible to provide top quality in all situations due to econ-
omic, resource and time restrictions. Secondly, the expansion of new technology dependent
phenomena, such as free and open machine translation (FOMT) and a range of crowdsour-
cing and volunteer translation initiatives, have also introduced new dimensions beyond the
classic literary, professional and training quality evaluation settings found in TS literature.
This flexible approach has irrevocably shifted the paradigm from the search of the desirable
absolute quality to a dynamic real-world conceptualization that takes into consideration

CONTACT Miguel A. Jiménez-Crespo miguelji@rci.rutgers.edu

© 2017 Informa UK Limited, trading as Taylor & Francis Group
 PERSPECTIVES 479

monetary, time and resource constraints (Wright, 2006), as well as new forms of MT and
volunteer production models. The shift from the constant top quality mantra to the adop-
tion and widespread use of different levels of adequacy under specific circumstances has led
to industry discourses that include different tiers of fitness for purpose with different price
points depending on the production model. Thus, these different levels of quality are often
associated to (1) how translations are produced, professionally, crowdsourced, community-
driven, machine translated and post edited by the crowd, or (2) the type of content, rather
than to indicators related to internal and external properties of translation products. This
paper explores how the impact of innovative crowdsourcing workflows and practices and
related economic considerations have further consolidated a dynamic conceptualization
of translation quality.

Defining crowdsourcing and online collaborative translations

From the many factors that have helped to widen the notion of translation quality, this
article focuses on the irruption of crowdsourcing and volunteer online collaborative trans-
lations as one of the most significant issues behind this change. Online collaborative trans-
lations represent activities in which the locus of control resides within the community
itself and it tends to respond to horizontal, rather than vertical hierarchies and control
structures (see for example Gambier, 2014; Jiménez-Crespo, 2015, 2017; Pym, 2011, for
a more detailed treatment of the differences). This last set of phenomena is defined as col-
laborative translation processes in the web initiated by self-organized online communities
in which participants collaborate with motivations other than monetary ones (Jiménez-
Crespo, 2017). Projects are initiated and selected by the community that, at the same
time, self controls and monitors the translation processes. These are cases such as the fan-
subbing of TV series or movies, scanlations or the translation of comic books, fan trans-
lation of videogames, etc.
Translation crowdsourcing has been used by businesses, organizations, institutions and
other groups to harness the wisdom of the crowd, generally a large group of amateurs,
experts, volunteers, professionals, fans, and citizens in order to accomplish any given
task (Brabham, 2013). In TS, it can generally be defined as collaborative translation pro-
cesses performed through dedicated web platforms that are initiated by companies or
organizations and in which participants collaborate with motivations other than strictly
monetary ones (Jiménez-Crespo, 2017). Crowdsourcing in translation has now expanded
from the models that assume the free nature of the contribution in cases such as social
networking sites (Jiménez-Crespo, 2015) or user forums (i.e. Mitchell, O’Brien, & Rotur-
ier, 2014), which could be considered as free translation crowdsourcing, to include
approaches in which participants are compensated, i.e. the so-called paid crowdsourcing
(García, 2015). In the paid crowdsourcing paradigm, initiatives often compensate partici-
pants depending on their qualifications or performance: from way below market rates to
non-professional participants, such as language learners, all the way to variable higher
rates to vetted communities of generalist or specialized translators in which professionals
with different skills collaborate to produce translations (see for example García, 2015).
This last notion can be considered as professional paid crowdsourcing (Jiménez-Crespo,
forthcoming), and it is represented by companies that crowdsource translations to pro-
fessionals using crowdsourcing workflows.
480 M. A. JIMÉNEZ-CRESPO

Despite the fact it is often argued that economic considerations are the main drive
behind crowdsourcing initiatives, cost estimates vary widely. In some cases, such as
social networking sites, it has been reported that the cost of creating and managing a
crowdsourcing platform might be the same as the cost of using professional services
(DePalma & Kelly, 2011). In other cases, crowdsourcing has been reported to cost approxi-
mately 20% of professional translators (Munro, 2013). Some non-profit initiatives some-
times virtually run at zero cost for the organization. Paid crowdsourcing initiatives also
vary widely and compensation to participants and price and prices to clients vary
widely. This range of variation opens up the business transaction to negotiation on how
the process is produced, and obviously to a range of fuzzy and ill-defined quality level
in the translation product.

Crowdsourcing and the reconceptualization of translation quality

Following the disruption caused by MT to the language industry at the beginning of the
twentieth century, crowdsourcing production models started to challenge whether top or
professional quality was ideal for all situations. It was then evident that in some cases
access, speed or user involvement were more important than high quality. The alarm in
professional circles came about in 2007 when the influential case of Facebook emerged,
and it took center stage after LinkedIn suggested that only professional translators
should participate in its volunteer translation initiative (Kelly, 2009). It quickly became
clear, despite a constant stream of complaints by professionals, that the highly aspirational
notion of top quality could also be produced by a highly motivated collective of users, with
careful management and workflows, leaving aside highly remunerated professionals
(Jiménez-Crespo, 2013b). If some of the most profitable companies in history would
turn to free labor, maybe the trend would expand and undermine the struggle of transla-
tors worldwide for higher remuneration. The debate was out, if volunteers were providing
translation services for free or below market rates, how would that impact the professional
world of translation that had been for decades striving for recognition and higher rates
related to the higher quality they produce? The main contention from scholarly and pro-
fessional associations was the inability to produce sufficient quality if professionals were
not involved (see Kelly, 2009), often pointing at the prevalent number of errors in non-
professional translations (i.e. Bogucki, 2009). Meanwhile, a segment of the industry
quickly saw potential to expand and develop a different business model that could capi-
talize on the wisdom of the crowd, leaving behind preconceived ideas about translation
quality. New business models emerged, stretching the quality range offered to span
from free online MT, volunteer crowdsourcing, volunteer post editing of MT, paid crowd-
sourcing, all the way to high quality human translation. Quality then started to be concep-
tualized in a continuum, ‘with the enduring and critical at one extreme, and the ephemeral
and inconsequential at the other’ (García, 2015, p. 340). Content prioritization became a
reality in the industry (O’Brien, 2012): does the translation of an internal email or a tweet
from a famous person always needs to be of high quality? The emergence of new highly
dynamic approaches to quality evaluation witness to this evolution, with several proposed
translation quality models, such as the Dynamic Framework of the Translation Auto-
mation User Society (TAUS) (Göroj, 2014a; O’Brien, 2012), and theoretical proposals
such as Jiménez-Crespo (2013a, pp. 127–131). These proposals implicitly acknowledge
 PERSPECTIVES 481

the wide range of quality scenarios, including the potential to encompass non-professional
translation and MT output. Dynamicity and adaptability thus has become the norm rather
than the exception.
As a result, the twentieth century has been marked by a move away from a static notion
of quality, exploring approaches beyond top-down models in which error typologies,
quality control and quality management procedures are established a priori and applied
across the board. Top-down approaches have started to be countered by the benefits of
bottom-up approaches ‘because they extend access. […] Users drive supply, so there is
less waste: material is only translated when needed’ (Drugan, 2013, p. 160). Quality is
thus differently conceptualized when disassociated from economic pressures and market
demands. In volunteer scenarios, translations are driven by user demand, and in turn,
users establish the expectancy norms or the type of quality that might suit them depending
on several factors, such as (1) their needs, (2) the speed at which the translation is needed or
(3) the relative permanence or sentimental value of the translation. Translation quality can
even become secondary to other considerations, such as cost, speed, usability or access to
content in long tail languages. Bottom up approaches also stress that in real life, the
notion of quality is user and context-dependent. That is, the user(s) and the relation
between the user and the situation of reception, among other factors, are ultimately what
can help establish what an appropriate or adequate translation might be, whether the trans-
lation is good enough for the purposes intended or how much anyone might be able and
willing to pay for a translation in any specific context and situation.

Quality in translation: a concept in constant evolution

As House (2013, p. 546) observes, ‘translation quality assessment has always been and still
is a challenge for translation studies’. At the same time, industry experts claim that their
view ‘on quality is highly fragmented, in part because different kinds of translation pro-
jects require very different evaluation methods’ (Lommel et al., 2014, p. 455). Quality
nevertheless continues to be the focus of research both in the industry and TS because,
no matter how fuzzy the notion of quality might be, the lack of consensus does not
mean that quality evaluation is not implemented around the world. Right now, thousands
of reviewers are editing human translations, volunteers are voting on their preferred ren-
dition of a translated segment on Facebook or Twitter, a volunteer might be grading an
admission test for a crowdsourcing non-profit initiative, while others are post-editing
the output of a free MT system.
Nevertheless, all these real world scenarios where evaluation occurs are guided by
implicit notions of quality that frame practices and discourses. Quality evaluation tasks
are ‘built up upon internalized frameworks of quality that guide subjects’ decisions,
even when they might lack operative theoretical foundations’ (Jiménez-Crespo, 2013a,
p. 103). These internalized frameworks of what an ideal or adequate translation should
look like diverge between different communities of practice, geographic locations, users’
expectations or evaluation contexts. They also vary according to how dynamic the
notion of quality is understood, as well as and the constraints that operate during the
translation process (and review or evaluation). These models have been described as anec-
dotal or experiential (Colina, 2009), based on the accumulated knowledge base (or lack
thereof) of subjects or organizations involved. Often, discourses revolve around notions
482 M. A. JIMÉNEZ-CRESPO

such as faithfulness to the source text, equivalence, lack of errors, does not feel right, etc.
(Angelelli, 2009; Colina, 2009). They can be considered the prevailing ones among trans-
lation practitioners, non-professionals, philosophers or industry experts (ibid). Real world
approaches immersed in a business environment are additionally framed within econ-
omic, time or situational constraints that make top quality not always possible or desirable
(Wright, 2006). Accepting these constraints has been crucial in this shift from perceiving
quality as a relatively unachievable abstraction to a practical construct operationalized
through a continuum of levels defined by situational criteria (i.e. Lommel et al., 2014).
It is also marked by a number of quality standards that apply to different aspects of
business provision, such as the ISO 900 quality, or translation specific ones such as the
European Union EN 15038. The acceptance of different levels of quality has also
started to gain traction in TS, mostly within professional-oriented approaches such as
Gouadec (2007, 2010), O’Brien (2012) or Drugan (2013).

Dynamicity in models of translation quality: towards adaptable models of

quality
While in the past TS devoted most of its efforts to establish evaluation criteria perceiving
quality as an absolute notion, industry and professional approaches claim that it is best to
separate between different levels of quality depending on a range of factors or constraints.
Current quality evaluation proposals embrace the dynamicity and highly constrained
nature of quality evaluation. For example, dynamic and highly customizable evaluation
metrics have been proposed, such as the TAUS Dynamic Quality Framework (DQF)
(Göroj, 2014b; O’Brien, 2012), the Multidimensional Quality Metrics (MQM) framework
developed by the EU-funded QTLaunchPad project (Lommel et al., 2014) or the recently
abandoned International Organization for Standardization (ISO) proposal for a quality
standard. They embrace the diverse nature of quality and the potential different pro-
cedures for its evaluation. They attempt to provide a platform to evaluate all types of trans-
lations, including MT output and non-professional ones. The logic behind these two last
approaches in industry metrics is clear: since they have been included in the mainstream
language industry, it is necessary to efficiently contrast and compare these lower ranked
options with professional ones. As García (2015, p. 31) indicates:
The translation industry has both room and need for a spectrum spanning professionals,
semi-professionals, casual aficionados and even untrained volunteers. Quality is not
always critical, and there is nothing inherently wrong with enterprises, institutions and
NGOs using these approaches.

This does not mean at all that the industry is attempting to justify the provision of low
quality levels. For example, the DQF model does not depart from the premise that
models should be a justification for low quality, ‘but that a dynamic QE model would
take into account the varying tolerance thresholds for quality that already exist in the pro-
fessional sphere’ (O’Brien, 2012, p. 68). It is also due to the fact that in certain situations
the industry has accepted that ‘it is best to provide some translations with available
resources than none, even if quality is lower’ (Drugan, 2013, p. 180). This means that
quality is de facto already understood in the industry as a gradable scale depending on
a range of factors, and it underlies the premise that the massive amount of content that
 PERSPECTIVES 483

needs to or could be translated is driven by user demand. In this respect, Göroj (2014b,
p. 389) from an industry perspective rightly indicated that:
The only way to offer large amounts of information and goods in multiple languages fast
while staying within reasonable budgets is by making a compromise and provide content
with different levels of quality using new translation channels and translation technology.

This also means that different quality levels might not require existing static or error-based
quality approaches to evaluation, and ‘in turn, require new ways of evaluating the quality of
translated content’ (2014b, p. 389). For example, in the overview of professional practices by
O’Brien (2012) for TAUS, it was found that content was prioritized based on parameters such
as utility, time and sentiment. Utility is related to the relative importance of the functionality
of the translated content. Time obviously refers to how quickly the translation is needed and
sentiment is related to the significance of the text for brand image etc. This shows that the
industry already has moved in some cases beyond the reliance on error-based approaches
and their inherent shortcomings (see Jiménez-Crespo, 2013a, pp. 116–120). Other proposals
such as Multidimensional Quality Metrics (MQM), incorporate several issues such as accu-
racy, fluency, design, locale convention, terminology, style, and verity, with an open custo-
mizable list of subtypes that includes 114 issue types (Lommel et al., 2014), with a core list of
19 basic ones. The models also open up the possibility of customizing the assessment through
error-based or holistic evaluation procedures. Basically, flexibility becomes the norm and the
issue then which parties have agency to customize or set up these procedures. However, the
potential for endless customization in these new proposals also opens the door to issues of
time and resource constraints. This was an issue with previous proposals with endless
error types (Gouadec, 1981) or complex contrastive procedures (House, 2014). In that
regard, these new proposals can be understood as an effort to explain and facilitate customi-
zation, rather than a complete new evaluation proposal. No matter how customizable a fra-
mework might be, customization models will probably tend to be replicated, and one or
several customization types might eventually become the norm.

Quality tiers in MT: towards a model for crowdsourcing and collaborative

models.
The introduction of the different levels of quality in TS and industry discourses can be
traced back to the resurgence of MT in the 1990s. The widespread adoption of MT
started to cast a shadow when statistical architectures started to produce acceptable
quality in certain areas and for certain purposes, such as customized engines in technical
domains (Hutchins, 2014). MT was thus a sort of Trojan horse that started to undermine
the top quality mantra that pervaded in academia and the industry for years. Users and
industry agents started to accept and use for certain purposes translation that could be
mired with errors and inadequacies, but still, they fulfilled a purpose or need at that
precise time. No longer quality, or better say, the absence of high quality, was an impedi-
ment to translations circulating and fulfilling some specific communicative purposes. The
negative discourse surrounding MT raw or gist translation in the discipline started to
extend to non-professional translations. This type of translation was defined in pro-
fessional circles as a rough translation ‘to get some essential information about what is
in the text and for a user to define whether to translate it in full or not to serve some
484 M. A. JIMÉNEZ-CRESPO

specific purposes’ (Chan, 2014, p. 42). FOMT translation has a relative long history since it
was first offered by CompuServe in collaboration with Systrans in 1994, and it has evolved
and expanded with Google Translate translating over 100 billion words per day in its 90
language combinations (Google, 2015).
The emergence of MT post-editing led to the proposal of different quality tiers depend-
ing on the extent of human involvement (Allen, 2003; Quah, 2006). The basic premise here
is that evaluation is not an infinite unlimited process but rather it follows the necessary and
sufficient rule (Wright, 2006). It assumes that a limited number of errors is unavoidable
due to the economic context in which this process is performed. Thus, MT approaches
were also instrumental in operationalizing quality in terms of equilibrium between a
wide range of constraints and the potential needs of users. For example, the Translation
Automation Society (TAUS) (TAUS 2010, pp. 3–4) guidelines for post editing MT intro-
duce two quality levels: good enough and publishable quality. The first one is defined as a
translation that is comprehensible and accurate so that it conveys the meaning of the
source text, but not necessarily grammatically or stylistically perfect. Another approach
based on how much post-editing is needed is found in Allen’s (2003) study, where differ-
ent quality tiers of machine translation output emerge: no post editing, minor post editing
intended for gisting purposes, and full post editing.
This range of quality seen in MT post editing also started to be found in TS literature on
professional translation. Gouadec, for example, several years later proposed a flexible para-
digm in professional translation and defended establishing degrees for different domains
or situations upon which quality evaluation and quality judgments can be built. Departing
mainly from the budgetary constraints for translation purposes in the Canadian govern-
ment, he proposed different quality tiers depending on how much a translation is fit for
delivery or broadcast (Gouadec, 2010, p. 273). The different customizable degrees of
quality would be: rough cut, fit for delivery (but still requiring minor improvements or
still not fit for broadcast medium) and fit for broadcast (accurate, efficient and ergonomic)
(2010, p. 273). He also included fit for revision grade to ‘describe translation that can be
revised within a reasonable time at a reasonable cost’ (2010, p. 273). Obviously, the indus-
try realization of these levels imply that they all would have different price points and/or be
related to different production models or cycles.
Other scholars such as Quah (2006, p. 154) resorted to MT approaches, identifying the
need for quality level depending on users’ needs. He also established a cline based on
whether machine translation or human translation would be needed (no post editing is
mentioned). The author placed highly creative and critical translation modalities as
requiring full quality, and therefore, human-based translation, while translation for
gisting purposes such as web articles is identified as a low priority process that can be
achieved through machine translation. This cline was subsequently adopted by Jiménez
Crespo (2013a, p. 110), who also added crowdsourcing to the potential levels of quality
that could emerge in relation to the type of content.
Figure 1 adapts the traditional cline of MT quality and its relation to types of content or
text, extending it to crowdsourcing and collaborative approaches. It is mainly based on
industrial crowdsourcing initiatives and not in unsolicited self-organized environments.
This does not mean that quality does not have different grades in fan or non-profit environ-
ments, but the distinctions are blurred in those cases. For example, crowdsourcing has been
used in emergency situations that involve life or death scenarios, such as humanitarian
 PERSPECTIVES 485

Figure 1. Translation quality cline in terms of human to machine translation including crowdsourcing.
Adapted from Jiménez-Crespo (2013a, p. 110).

crises (Munro, 2013), one of the top quality tiers in industry standards such as the SJAE
J2450. In these highly critical situations, which involve high text volumes in very short
periods of time, crowdsourcing has been used since professional environments could not
possibly accommodate the volume and speed at which these translations are needed.
The proposed categorization in Figure 1 includes high, medium and low grades of
quality and associates them with different types of content that might in principle be
amenable to different processes. Crowdsourcing is placed in both the medium and the
low quality spectrum of the triangle, and the different approaches are incorporated,
from general crowdsourcing and MT post editing by volunteers all the way to paid crowd-
sourcing in the middle of the spectrum. The medium grade also includes professionally
managed crowdsourcing, such as social networking sites, since they do produce highly
usable translations in their processes (Jiménez-Crespo, 2013b). The incorporation of
paid crowdsourcing in the cline reflects current industry practices and business models
(García, 2015), and it also suggests that some contexts require professional translation.
For example, a paid crowdsourcing portal such as Gengo excludes highly critical, legal
or marketing materials from their services. Paid crowdsourcing also includes a wide
range of potential quality with companies offering services from translation produced
in collaboration, such as Translation Cloud, all the way to professionals and specialized
experts working in crowdsourcing inspired hive environments such as the specialized ser-
vices by OneHourTranslation or Speaklike. The wide range of quality tiers and price
ranges offered by paid crowdsourcing portals thus requires a closer analysis.

Paid crowdsourcing and the customization of translation quality

The different quality tiers offered by paid crowdsourcing portals depend on several par-
ameters, such as the type of content, the type of crowd (professional or non-pro-
fessional/generalist or specialist), or the type of process, such as translation by the crowd
plus professional editing, translation by a crowd of specialists and editing, open or close
community involvement, etc. These translation companies offer different quality tiers
486 M. A. JIMÉNEZ-CRESPO

that are often described in relation to best-suited content types. The interrelation of content
type and levels of quality is considered one of the main dynamic factors in recent quality
evaluation proposals (O’Brien, 2012). For example, Gengo.com and Speaklike both offer
three tiers of paid crowdsourcing translation services that are not explicitly attached to
quality levels, but rather to the type of content that they can handle. They are also attached
to a range of price points, depending on the level (see García, 2015). Gengo offers standard,
business and ultra services while Speaklike offers basic, marketing and specialized levels. In
the case of Gengo, the standard level is recommended for everyday content, business for
professional content and ultra for business levels, such as static content, marketing etc.
Gengo also implicitly includes two other levels that are not provided by the company,
MT and specialist. MT is displayed as having a zero in terms of quality while the
company does not provide specialist translation, defined as texts related to legal, medical
or safety critical content. They also do not provide creative texts or those related to
image branding. Speaklike does offer specialized translations as part of the third quality
tier, and they are related to medical, legal or transcription services. This service is provided
by a different specialized crowd that is established by creating specialized enterprise groups
with pre-selected specialized translators with domain-specific experience or qualifications.
Thus, the levels of quality are interrelated with different levels of specialization by the par-
ticipating translators, even when in principle Gengo has a team of ‘over 15,000 pre-tested
translators’ and Speaklike has a qualified translation team. Get Localization also offer paid
crowdsourcing but it offers different customized solutions to crowdsource the translation
including the option of building or designing the crowd targeted, including own employees
of the firm requesting the translation, or building a crowd only of professional translators.
Other paid crowdsourcing vendors such as OneHourTranslation offer not only differ-
ent tiers such as general, expert and exclusive, but also the possibility of including revision
services in the price or not. The types of translation recommended are also attached to the
content requested and this company, in terms of quality, also offers different rates for
translation than for translation plus proofreading. That is, not only does the customer
need to select the type of process to achieve the quality requested, he/she also has to
pay a higher rate if proofreading is requested. For example, expert translation is offered
for 0.139 per word while translation plus proofreading would be 0.239 cents per word.
Again, the company claims that they work ‘only professional translators with rich trans-
lation experience’ that are pre-screened to work for them. They also offer separate exper-
tise groups in order to produce expert translations.
Arguably, the dynamicity in these business models relies on the fact that users and
clients ultimately decide what a fit for purpose translation is. In this change of paradigm,
both business agents and users understand that translation is a time, resource and money
constrained process (Jiménez-Crespo, 2013a; Wright, 2006) and therefore, fit for purpose
models could be considered as a ‘conscious attempt at using translation and revision
resources intelligently’ (Drugan, 2013, p. 42). Nevertheless, in these business contexts,
quality decisions reside on relatively uninformed translation clients and/or users based
on the price and the value that they place on the translated text or the service provided,
even when clients might not possess any knowledge of the difference between a standard
and an ultra translation or a basic and a specialist one. Quality is conceptualized as a scal-
able commodity that can be requested on different degrees depending on the character-
istics of the crowd of participants, with a similar approach to how any other products
 PERSPECTIVES 487

or services are purchased in the market, such as a car or a house. This process has been
argued to lead to a downward pressure on prices, rather than improving the overall con-
ditions in the industry (García, 2015). Nevertheless, almost a decade of crowdsourcing in
the industry has proven that the initial estimated economic impact has been much lower
than anticipated, if not negligible (DePalma & Kelly, 2011; Valli, 2015).
Normative standards throughout the years in the industry have been criticized for being
process-oriented instead of product-oriented (Wright, 2006, p. 256). The main argument
against them is that they normally indicate how language vendors establish procedures for
achieving quality, rather than providing normative statements about what constitutes
translation quality (Martínez Melis & Hurtado Albir, 2001, p. 274). New paid crowdsour-
cing approaches also fail to define what translation quality means but in this case, rather
than resorting to norms or standards, companies associate different quality levels to types
of content or content prioritization. In this shift, translation quality is not defined per se,
but rather replaced by a scale of potential value or worth of different types of content, from
the trivial to the highly important, defined in the way the quality of translation through a
preliminary analysis of the value source text rather than the actual translation quality.
Again, if the industry resorted to process-based rather than product based criteria
through their norms and standards, paid crowdsourcing models have also resorted to
extratextual parameters such as the estimated or sentimental value of the content.
The fact that translation quality can be secondary to other considerations can be ident-
ified in industry publications, speed, cost or even the motivation of volunteers to partici-
pate can be a higher priority than top quality. For example, Desilets and Van de Meer
indicate that:
Quality Control issues tend to resolve themselves, provided that enough of the right people
can be enticed to participate and that you provide them with lightweight tools and processes
by which they can spot and fix errors. (Desilets & van de Meer, 2011, p. 41)

In this case, the issue would be that even when quality might not be that significant in
some scenarios, it is not defined what the right people or what the minimum amount
of people might be. It shows a blind faith in a process-oriented or management-oriented
approach in which setting up the process will lead to adequate quality, even when the
notion of what quality might not be defined.

Crowdsourcing and different quality tiers and the consolidation of the

economic turn in TS
The diversification and widening of quality tiers with different price points in the trans-
lation industry is linked to research in what is starting to be known as the economic
turn (Gambier, 2014), a necessary link between translation and the economic factors
behind it that was anticipated by Pym, Shlesinger, and Jettmarová (2006). This turn or
research direction is due to the fact that ‘economic and financial dimensions can no
longer be ignored. There are [economic] factors that orient, and even determine, specific
choices and decisions’ (Gambier, 2014, p. 8). Nevertheless, if turns can only be defined or
perceived after they are completed, then the call for a research agenda into the economic
aspects of translation could not be considered a turn as such (Snell-Hornby, 2010, p. 368).
Therefore, simply making the call to engage in linking economic aspects in the study of
488 M. A. JIMÉNEZ-CRESPO

translation is insufficient to perceive it. Whether the turn will occur or not, this approach
is very much needed in analysis of the potential disruption to the market and notions of
quality brought by collaborative approaches. The interest of this trend resides in the fact
that, if research into translation has often been perceived or theorized within a pro-
fessional economic process in which agents act mainly according to economic motives,
taking the business aspects out of the picture gives ground to the possibility of different
conceptualizations of translation and translation quality and the forces that shape them.
Similarly, in solicited crowdsourcing cases in which companies decide to engage commu-
nities either to empower them and give them a voice, such as social media platforms, or in
cases in which translation cannot proceed in terms of return on investment principles
(ROI), such as software localization into minority languages, volunteer initiatives fill a
gap between translation as an economic activity and translation as a facilitator for cross
cultural communication based on different interests, such as cultural, linguistic, moral,
ethical, ideological, political, etc. Economic forces are responsible for the evolution of
crowdsourcing initiatives such as offering different tiers of translation quality based on
the selected community of participants, with different price points for communities
with different levels of skills. Not only are skilled translators perceived in these initiatives
as different from bilinguals, skilled translators are also divided between general or special-
ized translators for distinct translation types, offering different economic conditions for
both the participants and the clients. In tune with the findings of empirical studies into
motivations to volunteer that see intrinsic motivations as the main drive, Pym indicates
from a Bourdieuian perspective, that
The discussion of commerce is still very relevant, but we now have to recognize that the kind
of value for which effort is exchanged it not just economic: translators also work, legitimately,
for value of a social, symbolic and cultural kind. (Pym, 2012, p. 4)

The emergence of a research trend marked by economic issues makes true that turns can
overlap or that one turn can motivate another. This can be perceived in the fact that the
influence of economic and market forces were part of the research agenda of the sociologi-
cal turn (Inghilleri, 2009). Thus, the interdisciplinary nature of TS means that importing
theories and principles from Economics and Business studies could potentially bring about
new insights into the role of these forces in shaping translation practices, quality levels,
socio-professional considerations, etc. One example of this interdisciplinary process
could be the incorporation of the notion of asymmetric signaling from the economist
Specter (1973) in the study of how volunteer practices can disrupt the professional
market of translation (Pym, Orrego-Carmona, & Torres-Simon, 2016).

Conclusions
Quality has often been one of the most controversial issues in both the industry and the
discipline, and the interest in economic aspects related to translation highlights that both
are highly interrelated and deserve to be jointly studied. If TS has often proposed theor-
etical models that often cannot be fully implemented mainly due to economic restrictions
(time or resources), the industry has often quickly adapted to new realities and offered real
world applications within the traditional two-sided triangle. Either enough time and
quality is offered, but at a high cost, or short time deadlines and low cost is offered
 PERSPECTIVES 489

without quality. The industry has also responded to the inherent constraints in providing
translation quality at reasonable costs offering new dynamic models of quality provision
that depart from recognizing constrains and prioritizing certain aspects of translation
quality (Göroj, 2014a; Lommel et al., 2014). This paper has focused on how new crowd-
sourcing models have opened up the quality debate expanding the dynamic approaches to
quality and, at the same time, exploring processes that can also tackle the traditional
quality triangle limitations. Economic issues and the ingenuity of technology based
language service providers have explored how a large volunteer or professional crowd
can provide high quality translation at reasonable speed and at reasonable cost. These
industry explorers have started to produce a shift in the symbolic value of translation
quality. This radical shift involved a progression from a model in which language provi-
ders and professionals fought for recognition in the industry through the adoption of
international standards with clients being mostly unknowledgeable of what exactly consti-
tutes translation quality, to a model in which clients are responsible not only by checking
price differences, but also decide the price to pay for the quality or adequacy defined on the
basis of how the translation is carried out. From virtual zero cost of FOMT systems to
basic crowdsourcing carried out by bilingual students all the way to crowdsourcing to
crowds of specialized translators. Translation quality, as such, cedes to be perceived as a
monolithic and single construct to be a multidimensional approach with different price
points related to how the translation is produced.
This new shift involves setting up workflows and mechanisms depending on the initiat-
ives to guarantee the highest possible level of quality under the specific circumstances and
the types of crowds. The article has reviewed different mechanisms explored by new free
initiatives and those paid at below market rates in order to identify the workings behind
the creative drive of workflow managers and developers to provide different levels of
quality that might be attractive to customers and users. This creativity also extends to
free initiatives in non-profits that have benefited from the development of collaborative
platforms and different organizational possibilities.
To conclude, it should be acknowledged that quality has always been secondary to
economic considerations and that language service providers have constantly attempted
to develop new mechanisms to produce the desirable level of quality within reasonable
costs. The initial globalization and outsourcing of translation services has been expanded
exponentially through new crowdsourcing workflows with different specifications,
crowds, technology support etc. This has allowed opening up the provision of quality
with different quality tiers related to different production mechanisms rather than internal
features of the final translated product, shifting the ultimate responsibility of the quality of
the translation to customers who select the level of quality through a wide range of con-
siderations, such as the available budget, permanence of the translation, potential risks
involved, receiving audience, etc. As two of the main researchers into crowdsourcing indi-
cated, this is a phenomenon that ‘is still in its infancy’ (Estellés & González, 2012, p. 198).
The translation market should be prepared for new and exciting developments that might
continue reframing the provision of translation as we know it.

Disclosure statement
No potential conflict of interest was reported by the author.
490 M. A. JIMÉNEZ-CRESPO

Note on contributor
Miguel A. Jiménez-Crespo is an Associate Professor in the Department of Spanish and Portuguese
at Rutgers University, where he directs the MA program in Spanish Translation and Interpreting.
He holds a PhD in Translation and Interpreting Studies from the University of Granada, Spain. He
is the author of Translation and Web Localization published by Routledge in 2013, and the upcom-
ing monograph entitled Crowdsourcing and Online Collaborative Translations: Expanding the
Limits of Translation Studies to be published by John Benjamins in 2017. His research focuses
on the intersection of translation theory, translation technology, digital technologies, corpus-
based translation studies and translation training.

ORCID
Miguel A. Jiménez-Crespo http://orcid.org/0000-0002-4938-3095

References
Allen, J. (2003). Post-editing. In H. Sommers (Ed.), Computers and translation: A translators guide
(pp. 297–317). Amsterdam: Benjamins.
Angelelli, C. (2009). Using a rubric to assess translation ability: Defining the construct. In C.
Angelelli & H. Jacobson (Eds.), Testing and assessment in translation and interpreting studies.
ATA scholarly monographs series (pp. 13–47). Amsterdam: John Benjamins.
Bogucki, Ł. (2009). Amateur subtitling on the Internet. In J. Diaz-Cintas & G. Anderman (Eds.),
Audio-visual translation. Language transfer on screen (pp. 49–57). London: Palgrave Macmillan.
Brabham, D. (2013). Crowdsourcing. Cambridge: MIT Press.
Chan, S.2014). Routledge encyclopedia of translation technology. London: Routledge.
Colina, S. (2009). Further evidence for a functionalist approach to translation quality evaluation.
Target, 21(2), 235–264. doi: 10.1075/target.21.2.02col
DePalma, D., & Kelly, N. (2011). Project management for crowdsourced translation: How user-
translated content projects work in real life. In K. Dunne & E. Dunne (Eds.), Translation and
localization project management: The art of the possible (pp. 379–408). Amsterdam: Benjamins.
Desilets, A., & van de Meer, J. (2011). Co-creating a repository of best-practices for collaborative
translators. Linguistica Antverpiensia, 10, 11–27.
Drugan, J. (2013). Quality in professional translation. London: Bloomsbury.
Estellés, E., & González, F. (2012). Towards an integrated crowdsourcing definition. Journal of
Information Science, 38(2), 189–200. doi: 10.1177/0165551512437638
Gambier, Y. (2014). Changing landscape in translation. International Journal of Society, Culture &
Language, 22, 1–12.
García, I. (2015). Cloud marketplaces: Procurement of translators in the age of social media.
Jostrans, 23, 18–38. Retrieved from http://www.jostrans.org/issue23/art_garcía.pdf
Google. (2015). Google Translate keeps getting better with your help. Google Blogs. Retrieved from
http://googletranslate.blogspot.in/2015/06/google-translate-keeps-getting-better.html
Göroj, A. (2014a). Quantification and comparative evaluation of quality: The TAUS dynamic
quality framework. Tradumatica, 12, 443–454.
Göroj, A. (2014b). Translation and quality: Editorial. Tradumatica, 12, 388–391.
Gouadec, D. (1981). Paramètres de l’évaluation des traductions. Meta: Journal des traducteurs, 26
(2), 99–116. doi: 10.7202/002949ar
Gouadec, D. (2007). Translation as a profession. Amsterdam: Benjamins.
Gouadec, D. (2010). Quality in translation. In Y. Gambier & L. van Doorslaer (Eds.), Handbook of
Translation Studies (pp. 270–275). Amsterdam: Benjamins.
House, J. (1977). A model for translation quality assessment. Tübingen: TBL Verlag Gunter Narr.
House, J. (2013). Quality in Translation Studies. In C. Millan & F. Batrina (Eds.), Routledge hand-
book of Translation Studies (pp. 534–547). London: Routledge.
 PERSPECTIVES 491

House, J. (2014). Translation quality assessment: Past and present. London: Routledge.
Hutchins, W. J. (2014). Machine translation: History and applications. In S.W. Chan (Ed.),
Routledge encyclopedia of translation technology (pp. 120–135). London: Routledge.
Inghilleri, M. (2009). Sociological approaches. In M. Baker & G. Saldanha (Eds.), Encyclopedia of
Translation Studies (pp. 279–282). Amsterdam: Benjamins.
Jiménez-Crespo, M. A. (2013a). Translation and web localization. London: Routledge.
Jiménez-Crespo, M. A. (2013b). Crowdsourcing, corpus use, and the search for translation natur-
alness: A comparable corpus study of Facebook and non-translated social networking sites.
Translation and Interpreting Studies, 8(1), 23–49. doi: 10.1075/tis.8.1.02jim
Jiménez-Crespo, M. A. (2015). Volunteer and collaborative translation. In C. Angelelli & B. Baer
(Eds.), Researching translation and tnterpreting (pp. 58–70). London: Routledge.
Jiménez-Crespo, M. A. (2017). Crowdsourcing and online collaborative translations: Expanding the
limits of Translation Studies. Amsterdam: Benjamins.
Jiménez-Crespo, M. A. (2017). Mobile apps and translation crowdsourcing: The next frontier in the
evolution of translation. Tradumática, 14. Retrieved from http://www.raco.cat/index.php/
Tradumatica/article/view/316856/406952
Kelly, N. (2009). Freelance translators clash with LinkedIn over crowdsourced translators. Retrieved
from http://www.commonsenseadvisory.com/Default.aspx?Contenttype=ArticleDetAD&tabID
=63&Aid=591&moduleId=391
Lommel, A., et al. (2014). Multidimensional Quality Metrics (MQM): A framework for declaring
and describing translation quality metrics. Tradumatica, 12, 455–463.
Martínez Melis, N., & Hurtado Albir, A. (2001). Assessment in Translation Studies: Research needs.
Meta: Journal des traducteurs, 46(2), 272–287. doi: 10.7202/003624ar
Mitchell, L., O’Brien, S., & Roturier, J. (2014). Quality evaluation in community post-editing.
Machine Translation, 28, 237–262. doi: 10.1007/s10590-014-9160-1
Munro, R. (2013). Translation and crowdsourcing. Blog. Retrieved from http://idibon.com/
translation-and-crowdsourcing/
O’Brien, S. (2012). Towards a dynamic quality evaluation model for translation. Jostrans, 17, 55–77.
Retrieved from http://www.jostrans.org/issue17/art_obrien.pdf
Pym, A. (2011). Translation research terms: A tentative glossary for moments of perplexity and
dispute. In A. Pym (Ed.), Translation research projects 3 (pp. 75–110). Tarragona: Intercultural
Studies Group.
Pym, A. (2012). On translator ethics: Principles for mediation between cultures. Amsterdam:
Benjamins.
Pym, A., Orrego-Carmona, D., & Torres-Simon, O. (2016). Status and technology in the professio-
nalization of translators: Market disorder and the return of hierarchy. Jostrans: The Journal of
Specialized Translation, 25, 33–53. Retrieved from http://www.jostrans.org/issue25/art_pym.pdf
Pym, A., Shlesinger, M., & Jettmarová, Z. (Eds.). (2006). Sociological aspects of translating and inter-
preting. Amsterdam: Benjamins.
Quah, C. K. (2006). Translation and technology. London: Palgrave Macmillan.
Snell-Hornby, M. (2010). The turns in Translation Studies. In Y. Gambier & L. van Doorslaer (Eds.),
Handbook of Translation Studies. Volume I (pp. 366–370). Amsterdam: Benjamins.
TAUS. (2010). Post editing guidelines. Retrieved from https://www.taus.net/academy/best-
practices/postedit-best-practices/machine-translation-post-editing-guidelines
Valli, P. (2015). Disrupt me not. Keynotes 2015 A Review of the TAUS October Events, pp. 46–54.
Retrieved from https://www.taus.net/think-tank/reports/event-reports/keynotes-2015
Wright, S. E. (2006). Language industry standards. In K. Dunne (Ed.), Perspectives on localization
(pp. 241–278). Amsterdam: Benjamins.