ISSN: 1579-9794
Hikma 19 (2) (2020), 163 - 182
How does machine translation and post-editing affect
project management? An interdisciplinary approach
¿Cómo afecta la traducción automática y la posedición a la
gestión de proyectos? Un enfoque interdisciplinar
CRISTINA PLAZA-LARA
cplaza@uma.es
Universidad de Málaga
Fecha de recepción: 23 de enero de 2020
Fecha de aceptación: 20 de octubre de 2020
Abstract: Machine translation (MT) and post-editing (PE) are two services
that are increasingly in demand in the translation industry. In a context in which
large-scale projects are required within tight deadlines, the deployment of this
technology to increase productivity is a reality. However, this increase in
productivity can only be achieved with appropriate management of the project:
MT must not be understood as a tool, but as a process, and project managers,
who are usually responsible for the project from start to finish, have to cope
with new MT and PE workflows that pose different challenges. Although much
has been written about the use of MT and PE in professional practice
(resulting in different lines of research in this field), little attention has been
paid to the role of the project manager in MT and PE projects. For this reason,
the main objective of this paper is to analyse how MT and PE affect the factors
that project managers must keep in mind when managing projects, taking as
a starting point the most important reference frameworks in project
management. The main objective is to offer an interdisciplinary perspective
that explains the new challenges the industry is facing and how these
challenges affect the different stakeholders involved in the project.
Keywords: machine translation, post-editing, project management,
translation industry challenges, interdisciplinary approach
Resumen: La traducción automática (TA) y la posedición (PE) son dos
servicios cada vez más demandados en el sector de la traducción. El uso de
esta tecnología para conseguir un aumento de la productividad se ha
convertido en una realidad en un contexto en el que se precisan proyectos a
gran escala con unos plazos muy ajustados. Sin embargo, este aumento de
la productividad solo es posible con una correcta gestión del proyecto: la TA
no se debe entender como una herramienta, sino como un proceso, y los
gestores de proyectos, que son los encargados del proyecto de principio a fin,
164 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
tienen que hacer frente a nuevos flujos de trabajo de TA y PE que presentan
diferentes desafíos. Aunque se ha investigado mucho sobre el uso de la TA
y la PE en la práctica profesional (lo que ha dado lugar a diferentes líneas de
investigación en este campo), se ha prestado poca atención al papel del
gestor de proyectos en los proyectos de TA y PE. Por este motivo, el objetivo
de este trabajo es analizar cómo la TA y la PE afectan a los factores que los
gestores de proyectos deben tener en cuenta, tomando como punto de
partida los marcos de referencia en gestión de proyectos más importantes. El
objetivo principal es ofrecer una perspectiva interdisciplinar que explique los
retos a los que hace frente la industria y cómo dichos retos afectan a los
diferentes agentes implicados en el proyecto.
Palabras clave: traducción automática, posedición, gestión de proyectos,
desafíos del sector de la traducción, enfoque interdisciplinar
INTRODUCTION
In the last decade, the research and technological advances in the field
of machine translation (MT) and the changes in the translation industry are
posing new challenges to the industry players. In the new global scenario,
given the complexity and size of translation projects, a team of experts is
usually needed to fulfil project requirements. It is no longer a matter of
translators and reviewers simply offering services to a client that needs to
translate texts.
Occasionally MT is understood as a tool that can reduce translation
effort, schedule and costs. However, as Thicke points out, MT is not a tool,
but a process that only “if correctly managed, is capable of lowering translation
costs, increasing productivity and even improving quality and consistency”
(2013, p. 9). To achieve this, the team involved in the project must be aware
of all the factors and stages that will allow a successful implementation of MT.
This misinterpretation of MT as a tool and not as a process may be one of the
causes of the negative attitude that translators have towards this technology.
As Nunes Vieira (2018, p. 16) mentions, most criticism of MT does not concern
a fear of being outperformed by MT systems, but rather concern about MT’s
limitations and some of the business practices that surround its use.
The complexity of MT and post-editing (PE) projects, known as MTPE
or PEMT in the industry, has given project managers (PMs) a very important
role. A definition of this profile can be found in the ISO 18587: 2017 standard,
which is focused on post-editing of MT output: “[the PM is a] person who
manages specified aspects of a translation or post-editing project and is
responsible for the process” (ISO, 2017, p. 4). This definition highlights the
Cristina Plaza-Lara 165
Hikma 19 (2) (2020), 163 - 182
importance of the PM as the person who guarantees the quality of the process.
Although the ISO 18587: 2017 standard is restricted to PE as a human
process and leaves aside MT, this research will consider both processes as
part of the project workflow. Depending on the company, PMs will be
responsible for both MT and PE, or only one of these steps.
Therefore, the aim of this paper is to analyse how MTPE affects project
management. To that end, in the first place, the importance of MTPE and PMs
in the translation industry will be briefly outlined. Secondly, the different project
management processes and knowledge areas according to the most
important reference frameworks will be presented. Finally, in the last section,
the way all these knowledge areas and processes are influenced by MTPE
will be analysed, in order to provide a global overview of the challenges that
PMs face when managing this kind of project.
1. THE IMPORTANCE OF MTPE AND THE ROLE OF PROJECT MANAGERS IN THE
TRANSLATION INDUSTRY TODAY
Although the origins of MT go back to the middle of the 20
th
century
(García, 2012, p. 293), the achievements of recent years have given MTPE a
place among the services offered by translation companies. This has even
resulted in the development of new standards for the industry, such as ISO
18587: 2017 Translation services — Post-editing for machine translation
output — Requirements (ISO, 2017), which focuses on the process of MT
post-editing and post-editors’ competences. This standard complements the
previous ISO 17100: 2015 Translation services — Requirements for
translation services (ISO, 2015), whose scope comprises the processes,
resources and other aspects needed to deliver a quality translation service.
As far as MT and PE are concerned, this standard only provides a definition,
but does not go into further details.
On the basis of the data provided by the Language Industry Surveys,
published annually thanks to the collaboration between different international
organisations (Elia, EMT, EUATC, GALA, FIT Europe and LINDweb), the use
of MT has risen from 43% of companies and 33% of individual language
professionals in 2017, to 69% and 62% respectively in 2018. These
respondents confirmed that they are using MT to some extent, so it is possible
to state that “it is a strong indication that the market has accepted that machine
translation is here to stay” (Elia et al., 2018, p. 15).
In addition to this upward trend in the use of MTPE in professional
practice, the amount of research carried out in this field has considerably
increased in the last decade. Given that the aim of this work is not to present
the state of the art in the field of MTPE and due to the large number of
166 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
publications in recent years, no references will be made to specific authors
1
.
However, different lines of research can be identified: types of MT systems,
with recent attention to neural machine translation; quality assessment of MT
output and post-edited texts; productivity and performance studies; PE
methodologies and techniques; PE guidelines and effort; users’ attitudes and
perspectives; MTPE tools; MTPE teaching, and the profile and skills of post-
editors. These are just some examples that reveal the breadth of this field;
nonetheless, many other areas, such as pre-editing, controlled languages or
even computational linguistics, can be added to this long list.
With regard to the stakeholders involved in the new procedures
introduced by technological advances, little attention has been paid to the role
of the project manager, not only with respect to MTPE, but in Translation
Studies generally speaking.
Although project management is arguably the foundation of the
language industry, it has been largely overlooked as an object of
scholarly inquiry and critical pedagogical reflection in the field of
translation studies. While several translation scholars have
acknowledged the importance of developing project management
competencies and others have noted the central role of project
management in translation and localization […], few works devote
more than a cursory treatment to the topic. (Dunne & Dunne, 2011,
p. 6)
As Rodríguez-Castro points out “[t]he PM has not only become the hub
of the translator’s work environment, but plays a critical role in the organization
in order to mediate between all the stakeholders” (2013, p. 40), so this
professional profile also merits the attention of researchers. In the field of
MTPE, the guide Post-editing of Machine Translation for Project Managers
(Muzii, 2016) collects some practical advice for managing MTPE projects.
Some papers also address issues that directly affect PMs, but usually from
the translator’s and not the PM’s perspective: for example, productivity
predictions (O’Brien, 2011; Candel-Mora & Borja-Tormo, 2017),
implementation of MT in companies (Rico Pérez & Díez Orzas, 2013; De la
Fuente, 2014; Córdoba Mondéjar et al., 2015; Porro Rodríguez, Vázquez
Morado & Bouillon, 2017) or integration with other tools (Moorkens, Doherty,
Kenny & O’Brien, 2014; Zaretskaya, Corpas Pastor & Seghiri, 2015). Only two
papers by Sakamoto (2018 and 2019) that explore project managers’
perceptions about how MT is affecting their business practices have been
1
Koponen (2016) presents a complete description of the current state of research in this field.
Several special issues on this topic have also been recently published (see Tradumàtica, issue
15; JoSTrans, issue 31, or the conference proceedings of the bi-annual MT Summit organised by
the European Association for Machine Translation).
Cristina Plaza-Lara 167
Hikma 19 (2) (2020), 163 - 182
found. But, as many researchers have noted, the introduction of MTPE implies
a paradigm shift in translation (Reid, 2013; Rico Pérez & Díez Orzas, 2013;
Córdoba Mondéjar et al., 2015; Candel-Mora & Borja-Tormo, 2017) and PMs
are directly concerned as key stakeholders in the supply chain.
2. PROJECT MANAGEMENT PROCESSES ACCORDING TO THE PMBOK® GUIDE
AND ISO 21500
Due to the internationalisation of projects as a method of organising
work in different industries, the need arose to harmonise standards and to
establish principles and procedures that could be applicable to any
organisation regardless of the sector. It was not until 2012 that the
International Organization for Standardization published ISO 21500: 2012
Guide on project management (ISO, 2012), in spite of the fact that different
good practice reference frameworks had already been published at national
and international level: A Guide to the Project Management Body of
Knowledge (PMBOK® Guide) published by the Project Management Institute
(PMI, 1996/2017), the Individual Competence Baseline (ICB), proposed by the
International Project Management Association (IPMA, 2006/2015), or the
methodology PRINCE (PRojects IN Controlled Environments) (Axelos,
1996/2017), which was established in 1989 by the current Office of
Government Commerce in the United Kingdom. Since all these previous
models already existed when the ISO standard was published, it is not
surprising that much of the content of the latter corresponds to the theoretical
frameworks established by the different international organisations.
For this reason and given that ISO 21500: 2012 takes the same project
management processes and knowledge areas as the PMBOK® Guide (PMI,
2017), in this paper the knowledge areas defined by the PMI will be taken as
a starting point in order to analyse how each of them is affected by MTPE
practices. According to this publication, ten knowledge areas can be
distinguished in project management:
1. Project Integration Management
2. Project Scope Management
3. Project Schedule Management
4. Project Cost Management
5. Project Quality Management
6. Project Resource Management
7. Project Communication Management
8. Project Risk Management
9. Project Procurement Management
10. Project Stakeholder Management
168 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
All these areas intersect with the five project management process
groups (initiating, planning, executing, monitoring and controlling, and closing
a project) and the guide provides a description of tools and techniques used
within these processes. The next sections will examine how MTPE affects
these knowledge areas, when compared with traditional translation project
management.
3. HOW DOES MTPE AFFECT THE KNOWLEDGE AREAS INVOLVED IN PROJECT
MANAGEMENT?
Despite the theoretical framework defined by the different organisations
mentioned in the previous section, it cannot be denied that it is business
circumstances and organisational needs that will determine which processes,
knowledge areas and competences are required in each situation.
It must also be kept in mind that MT is not used in the same way by all
translation organisations in the language industry. For this reason, the
relationship between PMs and MT can take different forms:
• PMs working for a company that has its own MT engine and
can offer MTPE as a customer service.
• PMs working for a company that does not have its own MT
engine, but whose clients process texts with an MT engine and
want them to post-edit that text.
• PMs working for a company that does not have its own MT
engine, but that use MT features offered by different platforms.
• PMs working with translators that use MT engines.
Depending on the circumstances, the role of the PM may vary. In this
paper, special attention will be paid to PMs whose company has an MT
engine. Although it may not be the most frequent scenario, it provides a
complete overview of all the challenges that MTPE is posing to PMs
throughout the project. However, in certain cases, challenges for PMs not
having their own engine will also be mentioned.
Of the ten knowledge areas identified in the PMBOK® Guide (PMI,
2017), three of them take on paramount importance: costs, schedule and
quality. As demonstrated by the study carried out by Plaza-Lara (2018, p. 524-
525), according to the information provided by a corpus of job advertisements
for translation PMs, employers particularly value the ability of the PMs to
manage costs, quality and schedule. These factors are also mentioned at the
beginning of ISO 18587: 2017. The implementation of MTPE only makes
sense if the MT output is good enough to reduce costs and shorten schedules
without jeopardising the expected quality of the translated text. If one of these
three areas is compromised, the viability of the project as initially conceived
Cristina Plaza-Lara 169
Hikma 19 (2) (2020), 163 - 182
should be reconsidered. For this reason, in this paper, these three knowledge
areas will be analysed first.
3.1. Schedule
This knowledge area, initially called project time management, was
renamed in the 6
th
edition of the PMBOK® Guide (PMI, 2017) to clarify that it
is the schedule that is managed, not the time. In this regard, PMs are
supposed to sequence the activities to be performed in order to produce the
deliverables and guarantee the timely completion of the project.
Both ISO 17100: 2015 and ISO 18587: 2017 include pre-production
processes as part of the translation/post-editing project respectively. This
covers enquiry and feasibility study, quotation, negotiation with the client and
project preparation. However, with the addition of the machine factor, the
number of preparatory activities involved in initiating a post-editing project
increase. First of all, an initial analysis of the source language content, the
language combination and the domain should be carried out taking into
account the MT engine. These factors are essential to determine MTPE
efficiency (ISO, 2017, p. 5). In second place, it should be decided whether the
source text should be pre-edited in order to improve translation output and
reduce post-editing efforts. This decision depends on costs and the quality of
the customer's text. Finally, MT should be implemented to obtain a first version
of the translated text that is subsequently sent to the post-editor. During this
step, a quality evaluation of the MT output should be carried out, in order to
guarantee that the text to be post-edited meets minimum requirements. These
three basic steps may vary from one project to another.
The development of the MT translation engine also plays an important
role, as they must be trained in the initial stages and this requires some
investment of time before giving satisfactory results. As Muzii points out “SMT
[statistical machine translation] and NMT [neural machine translation] training
cycles might be very different: Quite brief for the former, a few hours tops, and
very, very long for the latter, days when not weeks” (2016, p. 26). This should
be taken into account by PMs, as MT output should also be evaluated before
proceeding with PE.
Once the text has been post-edited and delivered, it is essential to carry
out some post-production tasks. PMs should receive feedback on the MT
output from the PE team (for example, using templates or questionnaires or
collecting samples of PE issues) and should plan metrics, such as the edit
distance, that will be of great help to the MT development team, in order to
improve the engine performance and ensure maintenance.
170 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
In the case of companies that do not have their own MT engine, but
which are required to post-edit a text that the client has processed with MT,
PMs should check MT quality to set a schedule with the aid of post-editors
and even with the client, who should provide information on the MT output. If
quality is good enough, delivery times should be reduced; according to Muzii
“throughput rates usually range between 450 to 750 words per hour and
4,000-6,000 words per day” (2016, p. 49). Notwithstanding, this also depends
on other factors such as the source content, language combination and even
the post-editor’s expertise. In this respect, efforts are being made to develop
tools that help to estimate PE effort and time (Candel-Mora & Borja-Tormo,
2017).
As can be seen, MTPE involves some extra steps that may be
disregarded by PMs due to the short turn-around times in the industry. If this
occurs, other knowledge areas analysed here will be directly affected.
3.2. Cost
According to the PMBOK® Guide (PMI, 2017, p. 24), this knowledge
area refers to the processes to complete a project within the approved budget.
On this basis, cost saving can be considered one of the major drivers to putting
into practice certain procedures during a project and this is the case for MTPE.
As Sakamoto indicates, customers “know that [MTPE] is [a] way of getting
savings” (2018, p. 7) and “[t]his is very likely the source of resentment
expressed amongst translators and PMs” (2018, p. 8).
A concept closely related to the project cost management is that of
productivity. As productivity studies have shown (see review presented by
Koponen, 2016), PE productivity increases when compared to translation if
MT output quality is good. As a consequence, rates have been adjusted, as it
is supposed that PE is faster and less keyboard intensive than human
translation (Muzii, 2016, p. 27). This results in cost savings and it would not
be justified if the post-editors, who are normally at the end of the supply chain,
do not observe the increase in productivity that compensates for the actual
work effort and, as a result, for the discount. Therefore, defining rates or
compensation models can be one of the major challenges PMs have when
dealing with MTPE.
Industry practices show that it is quite common to establish a rate
before the completion of the project (around 61% of the full per-word rate
according to Vashee, 2013, p. 143). This approach, called the ex-ante
compensation model by Muzii (2016, p. 47), is even adopted when MT output
has not been properly evaluated and it may be the case that the client does
not want to compensate for the extra efforts caused by poor quality MT. In
contrast to this approach, Muzii presents two alternatives: (1) an ex-post
Cristina Plaza-Lara 171
Hikma 19 (2) (2020), 163 - 182
compensation model, that “requires an accurate measurement of the actual
work performed. This is usually made by calculating the edit distance and then
inferring the percentage on the hourly rate” (Muzii, 2016, p. 47); (2) a
compensation model based on productivity gain (also mentioned by Vashee,
2013, p. 143), for which he recommends carrying out a pilot project that could
help estimate the throughput rate and establish fair compensation (Muzii,
2016, p. 48). Maybe because these two approaches require more time
investment once the project is completed, the ex-ante compensation model is
preferred by most of the stakeholders in the industry. However, as indicated,
post-editors at the end of the supply chain may find this approach unfair and
it may be one of the causes of their negative attitude towards MTPE (Nunes
Vieira, 2018, p. 16; Sakamoto, 2018, p. 8).
Deciding whether or not to purchase and deploy an MT engine is
another issue related to cost management. Although this paper focuses on
the challenges of MTPE to PMs and the acquisition of an MT engine is not
normally their responsibility, their opinion can be of great help, as they should
be familiar with the projects they manage. In this respect, two financial
concepts come into play: the return on investment (ROI) and the total cost of
ownership (TCO). The ROI measures the profits generated by an investment
and the TCO refers to the costs of an asset and its operation costs. The latter
is examined thoroughly by Vashee (2013, p. 140-141), and he mentions
human resources, infrastructure, data, skills development, TMS/workflow
integration, customisation costs, management costs, time to quality, cost of
post-editing and prevailing market rates for translation and editing. All these
factors should also be taken into account by PMs, as they have an indirect
impact on their work productivity. In essence, any company considering
investing in an MT system should bear in mind the following:
MT requires simultaneous and ongoing investments in technology,
process and training to deliver long-term benefits and competitive
advantage. MT has increasing value with long-term volume and
repeated use; the greater the volume and usage in a specific domain
or subject area, the greater the economic benefits and value to an
enterprise. It is rarely if ever possible to obtain a turnkey ‘solution’
by simply paying money to a vendor or to an internal team that will
develop the solution. Any MT initiative is an evolutionary and
iterative process along multiple dimensions. (Vashee, 2013, p. 130)
3.3. Quality
The concept of project quality management in MTPE can have different
meanings:
172 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
1. Quality of the project management process to guarantee that
the project requirements are met. For this purpose, ISO
21500: 2012 Guide on project management (ISO, 2012) was
created, but the good practice frameworks mentioned in section
2 are also a reference.
2. Quality of the deliverables of the project, that will establish the
post-editing level (light or full post-editing).
3. Quality of the MT output, which has a direct impact on the rest
of the knowledge areas and determines the feasibility of the
MTPE project.
Regarding the quality of the process, there is a concept that is also
closely related to that of costs, but that is included here because the aim is to
maintain quality. It is the cost of quality (COQ) and according to the PMBOK®
Guide:
The cost of quality (COQ) includes all costs incurred over the life of
the product by investment in preventing nonconformance to
requirements, appraising the product or service for conformance to
requirements, and failing to meet requirements (rework). (PMI,
2017, p. 274)
According to the PMI (2017, p. 282), the COQ can include prevention
costs (to prevent poor quality in the deliverables or services), appraisal costs
(to audit quality in the deliverables or services) and failure costs (caused by
the non-conformance of the deliverables or services). Whereas in translation
projects (even if aided by computer-assisted translation tools), the translation
is under the translator’s control, in MTPE projects the machine component is
added. This results in the inclusion of certain factors that are not easily
controlled and which can have a positive or negative impact on the process
and the product. The PM should try to identify all the factors (source quality,
purpose of the text, etc.) that can affect quality in order to ensure that the COQ
does not exceed the cost savings of MTPE.
With regard to the quality of the project deliverables, as extensively
described in the MTPE literature, two different levels of PE can be
distinguished: light PE, when lower standard quality is enough, and full PE,
when publishable quality is needed (TAUS & CNGL, 2010, p. 16). Muzii adds
gisting PE, which “consists in raw MT output with virtually no corrections but,
possibly, with automatic fixing of mechanical errors by using regular
expressions” (2016, p. 32). In this respect, the PM should analyse clients’
needs to advise them when deciding on the level of PE to be applied to each
project, as well as to provide the project team with the correct guidelines to
achieve that level of quality. Furthermore, especially when publishable quality
Cristina Plaza-Lara 173
Hikma 19 (2) (2020), 163 - 182
is required, the PM should plan quality assurance tasks (for example, review
by a second linguist).
Both the COQ and the PE level are subordinated to the quality of the
MT raw output. The PM, as the person responsible for the project from start
to finish, should at least understand the operating principles of the engine to
provide the team with the necessary instructions to satisfy clients’ demands.
PMs must inform the MT team about the type of texts and the content to be
translated, so that the training data can be cleaned and properly labelled.
Knowing the data with which the MT engine has been trained can also help
PMs to decide if a text can be translated using MTPE or even to select a
sample to assess MT raw output (either automatically or using human
evaluation) and estimate PE effort.
3.4. Resources and stakeholders
These two knowledge areas will be addressed together, because they
are intimately related. Project resource management
2
includes the human
team and the equipment, materials or supplies necessary to accomplish a
project (PMI, 2017, p. 307), whereas project stakeholder management refers
not only to the project team, but in general to the “people, groups, or
organizations that could impact or be impacted by the project” (PMI, 2017,
p. 503). Therefore, stakeholders also include clients or end users, to mention
two examples.
Before going into the human factor in detail, attention will be paid to the
equipment needed to complete an MTPE project. The main difference with
respect to a translation project is the implementation of an MT engine. As
previously stated, although this decision is not the sole responsibility of PMs,
their knowledge about the project features can be helpful to decide what type
of engine is most appropriate. Statistical Machine Translation (SMT) and
Neural Machine Translation (NMT) seem to be the two most used systems
today: in the early 2000s, SMT became commonplace, but recently NMT has
gained a foothold in academia and industry (Moorkens, 2018, p. 375-376). As
Muzii mentions, SMT is becoming cheaper and the great amount of
documentation available nowadays makes configuration easier, whereas
NMT systems are usually more expensive and challenging as far as
development is concerned (2016, p. 17-18). In both cases, the quality of data
is vital to obtaining reasonable results. Nitzke, Hansen-Schirra and Canfora
(2019, p. 244) advise training an MT system when the company has a lot of
reliable multilingual text. On the other hand, external corpora should be used
if there is not much in-house data.
2
In the 5
th
edition of the PMBOK® Guide, it only included human resources.
174 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
With regard to human resources, a distinction will be made between the
team that actively takes part in the project to obtain the final product and other
stakeholders, that are not directly involved in the project execution, but that
impact/are impacted by its results. Among the latter, PMs should pay special
attention to customers and end-users, as they may influence the decision as
to whether to apply MT or not depending on the number of words to be
translated, the schedule, the budget and the end purpose of the translated
text. In this case, vendors, business partners and external companies are
considered part of the human resource team, because their services are
usually hired to achieve the project goals.
As highlighted in other knowledge areas, the particularities of the MTPE
workflow also affect the skills of human resources. On the one hand, PMs
should manage a pool of post-editors; on the other, they should count on an
interdisciplinary team of experts in MT that put into practice the necessary
measures for the successful implementation of the MT. Communication plays
an essential role when coordinating this team. As Vashee mentions (2013,
p. 140), the team involved in MT implementation should have linguistic
expertise (especially in natural language processing), technical expertise and
programming expertise to tailor the engine to meet the organisation’s needs.
With regard to the PE team, the competences described in ISO
18587: 2017 do not differ much from those of the translator. However, this
standard mentions “the knowledge and ability to establish whether editing MT
output makes sense, in terms of time and effort estimation” (ISO, 2017, p. 8).
This ability is very important from the PM’s perspective, as post-editors may
complaint about MT quality. If the steps prior to sending the text to the post-
editor have been correctly carried out and MT output has been positively
assessed, these complaints may be due, for example, to a lack of experience
on the post-editor side. In this regard, in Annex A, ISO 18587: 2017 details
the knowledge and skills that post-editor training should take into account.
3.5. Risks
According to the PMBOK® Guide, a risk can be defined as “an
uncertain event or condition that, if it occurs, has a positive or negative effect
on one or more project objectives” (PMI, 2017, p. 397). In the field of
translation, Dunne’s doctoral dissertation (2013) addresses project risk
management for translation projects and recently the article published by
Nitzke, Hansen-Schirra & Canfora (2019) focuses on risk management and
post-editing, but mainly from the post-editor perspective.
As mentioned in section 3.3, in MTPE projects the addition of the
machine component puts into play certain factors that may not always be
under control. As Nitzke, Hansen-Schirra & Canfora point out “the PE process
Cristina Plaza-Lara 175
Hikma 19 (2) (2020), 163 - 182
should be subject to risk management right from the beginning” (2019, p. 252)
and this means that poor risk management could compromise other areas
such as quality, schedule or costs. These risks could include, for example,
incorrect labelling of the data, lack of skills of the post-editing team or breach
of confidentiality. PMs must be aware of the limits and possibilities of MT in
order to minimise risks, and understand it as a process, not a tool (Thicke,
2013, p. 9) that can be easily implemented only to obtain certain benefits.
Those companies that decide to invest in an MT engine are taking their
first business risk that could involve others. This investment should be
preceded by a thorough analysis of the ROI and TCO, as explained by Vashee
(2013, p. 140). It is also of great importance that PMs receive the proper
training in MT processes, so that they can monitor the whole workflow and
mitigate risks with the help of the team. For example, if no quality feedback is
gathered to maintain and improve the engine, the ROI will be significantly
reduced.
But MTPE is not only a risk for those having their own MT engine. PMs
can receive projects in which MT has been applied and the scope is to post-
edit that text. If no further information is provided on MT quality, they may
consider it a risky project, as they cannot control the circumstances under
which it was decided that MTPE was the best solution for that project.
Finally, it must be remarked that MT can also be a risk in translation
projects. Sometimes, translators apply MT engines without prior authorisation
and, as it will be commented in section 3.7, this can lead to confidentiality
issues.
3.6. Integration and scope
Although these two knowledge areas are treated separately in the
PMBOK® Guide (PMI, 2017), they will be addressed jointly in this paper, as
they both refer to the processes needed to complete the project satisfactorily
and to obtain the desired product. Project integration management is defined
as “the processes and activities to identify, define, combine, unify, and
coordinate the various processes and project management activities”
3
(PMI,
2017, p. 23), whereas project scope management refers to the “processes
required to ensure that the project includes all the work required, and only the
work required, to complete the project successfully” (PMI, 2017, p. 129). If
these knowledge areas are not clear from the beginning of the project, PMs
3
As previously mentioned, they include initiating, planning, executing, monitoring and controlling,
and closing a project.
176 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
may incur costs and experience schedule and quality issues that could put the
project at risk.
In order to establish the different project management activities, the
scope must be correctly defined. This requires active stakeholder involvement
to collect and analyse the requirements of the final product (PMI, 2017,
p. 140). The creation of a work breakdown structure (WBS) can help to divide
the project work into more manageable components (PMI, 2017, p. 158) that
will be part of the different stages that comprise project integration processes.
In these two knowledge areas, one of the main challenges for PMs
whose companies have its own MT engine may be the lack of knowledge of
MTPE projects processes. Experience in translation project management
allows them to define the features and functions of the final product (PMI,
2017, p. 131). Nonetheless, that product could not be delivered if the work and
processes needed to achieve it are unknown to the PM. As stated in previous
sections, with the inclusion of MTPE, new procedures enter the project
workflow and PMs may not have the proper knowledge to cope with them.
This does not mean that PMs must master all the skills to accomplish an
MTPE project, but they have to be aware of all the factors and activities that
must be performed to manage them properly and avoid trial-and-error
learning.
3.7. Procurement
According to the PMBOK® Guide, procurement includes “the
processes necessary to purchase or acquire products, services, or results
needed from outside the project team” and “the management and control
processes required to develop and administer agreements such as contracts,
purchase orders, memoranda of agreements (MOAs), or internal service level
agreements (SLAs)” (PMI, 2017, p. 459).
In the translation industry, outsourcing services is a common practice
involving the establishment of agreements between the vendor and the
service requestor with regard to general terms and conditions, pricing,
statements of work or acceptance criteria, among others (PMI, 2017, p. 489).
These agreements may be similar to those reached for translation projects,
but for MTPE the ISO 18587: 2017 standard gathers in Annex D a list of
elements that should be included in a client-translation service provider
agreement: confidentiality clauses and non-disclosure agreements (NDA);
restrictions on use of by-products such as translation memories; liability, etc.
Confidentiality has become a controversial issue in this kind of project.
For PMs whose company owns and trains their in-house engine using the
client’s own data (Muzii, 2016, p.19), confidentiality should not be considered
Cristina Plaza-Lara 177
Hikma 19 (2) (2020), 163 - 182
a matter of major concern, as the client’s data are always under control. The
problem arises when web-based translation applications, such as Google
Translate, Bing Translator or Yandex Translate, are used (Lagarda, Ortiz-
Martínez, Alabau & Casacuberta, 2015, p. 116). These systems may breach
data protection law and, as Sakamoto’s research (2018, p. 6) shows, although
some companies explicitly ban their translators from using MT, they cannot
control them. For this reason, the project procurement management
knowledge area gains great importance, especially since free online MT
engines are increasingly improving their output. As Muzii states:
Since the protection of data integrity, confidentiality, and intellectual
property is a legitimate expectation that must be fulfilled when
represented, it must be made explicit in the specifications of
requirements, in agreements and contracts, and in the statements
of work or checklists (if any). (Muzii, 2016, p. 28)
3.8. Communications
Although the project communications management may seem a quite
obvious knowledge area in every project, the particularities of the industry give
it an important role:
Within the subcontracting model, the PM’s role in managing the
translation process and its associated communication workflow has
become crucial to success. As the distributed workforce moves
offsite (and often, offshore) managing process and workflow takes a
central place in the organization; the PM becomes the main mediator
between upper management, language professionals and the end
client. The PM has become a key, if not the key component of the
translator’s work environment. (Rodríguez-Castro, 2013, p. 44)
As can be inferred from this quotation and as mentioned in the resource
and stakeholder section, the complexity of the team involved in MTPE projects
obliges PMs to have a well-defined communications management and control
plan. On the one hand, the client’s needs must be communicated to the MTPE
team that will handle the information to satisfy the project requirements. On
the other hand, communications between the MT team and post-editors are
essential and PMs should mediate in order to make the most of MTPE (for
example, through the feedback provided by the post-editors). As Sakamoto
(2017 and 2018) shows, according to the PMs included in her study, if
translators do not like MT, they often report the problem. However, if they are
happy with MT quality, they tend not to inform PMs. This fact shows the
reluctant attitude of certain language professionals towards the use of this
technology.
178 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
In the case of companies or freelancers that do not have their own MT
engine, but that do use MT features offered by different platforms, Sakamoto’s
studies reveal that silence occurs when MT is used without the customer’s
consent and in spite of the implications this use could have, no specific
measures are implemented (Sakamoto, 2018, p. 6). These measures should
include, for example, the signing NDAs and delivery of checklists. Although
this does not prevent translators from using these MT engines, it creates a
confidential relationship between the parties.
CONCLUSION
The breakthrough of MT engines and the need to reduce time and cost,
increasing productivity and quality, especially in the case of large-volume
projects, have placed the focus of many companies on MTPE projects. This
relatively new market trend affects the way projects are managed and the aim
of this research has been to analyse which new concepts, factors and
processes have come to modify project management. Nonetheless, the
interdisciplinary perspective here presented should be complemented with
further studies about the current practices in the industry in order to confirm
whether project managers have adapted to the challenges posed by MTPE
projects.
One of the main conclusions drawn from the interdisciplinary approach
presented in this paper highlights the convergence of the knowledge areas
involved in project management. PMs should be able to take a global
perspective that allows them to monitor the different stages and factors that
form part of the project in order to reduce possible problems. This does not
differ from translation project management. Notwithstanding, it has become
clear that one of the challenges that PMs face when managing MTPE projects
are the new processes and factors that the introduction of MT entails and that
differ from those in translation projects: training of the MT engine; new pricing
and compensation models; different interpretations of the concept of quality;
confidentiality issues, etc. The PM needs a general overview of all of these to
be able to establish connections, especially when planning, executing and
monitoring the project.
In this context, it can be confirmed that PM’s skills must be adapted to
the new market requirements. In the case of companies having their own MT
system, it is crucial for PMs to have a well-trained team capable of carrying
out the project from start to finish while minimising possible risks. Although the
execution of the project is the responsibility of these experts, the PM should
be able to plan, schedule and control the different tasks taking into account all
the knowledge areas here analysed and the new factors and processes that
Cristina Plaza-Lara 179
Hikma 19 (2) (2020), 163 - 182
MTPE projects involve. The control and communication role of this
professional profile is thus essential.
MTPE is also a challenge for those PMs who do not have to manage
the MT stage, but who are required to post-edit a project that the client has
previously translated using MT. The uncertainty created when managing a
project that depends on the results offered by a machine that is not under
control causes some reticence (Nunes Vieira, 2018, p. 16; Sakamoto, 2018,
p. 8). In this case, special efforts should be made to bring these positions
closer together, for example, by providing edit distance data or involving post-
editors in the MT training.
In conclusion, there is a need to re-educate all the stakeholders
involved in this cycle: on the one hand, in order to understand the complexity
of MTPE and see it not only as cost-saving tool; on the other, to promote
greater transparency in the MTPE process to diminish fears and negative
attitudes, not only among post-editors but also among PMs.
REFERENCES
Axelos. (2017). Managing Successful Projects with PRINCE2® 2017 Edition
(sixth edition). Norwich, United Kingdom: Axelos
Candel-Mora, M. Á., & Borja-Tormo, C. (2017). Desarrollo de la aplicación
Post-editing Calculeffort para la estimación del esfuerzo en posedición.
Tradumàtica. Tecnologies de la Traducció, 15, 1-9. DOI:
10.5565/rev/tradumatica.190
Córdoba Mondéjar, I., Rico Pérez, C., Ortiz Jiménez, M., Doquin de Saint
Preux, A., Arevalillo Doval, J. J., Arcedillo Jiménez, M., & Cabero
Zumalacárregui, J. (2015). Estudio de viabilidad para la implantación
de la traducción automática en la empresa. VITAE. Madrid, España:
Gobierno de España, AEI y FEDER. Retrieved from
https://abacus.universidadeuropea.es/handle/11268/4362?show=full
De la Fuente, R. (2014). Sampling for Machine Translation Evaluation.
Tradumàtica. Tecnologies de la Traducció, 12, 476-183. DOI:
10.5565/rev/tradumatica.61
Dunne, E. (2013). Project Risk Management: Developing a Risk Framework
for Translation Projects (Doctoral dissertation). Kent State University.
Retrieved from https://etd.ohiolink.edu/!etd.send_file?accession=kent1
368700402&disposition=attachment
180 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
Dunne, K., & Dunne, E. (2011). Translation and Localization Project
Management. The Art of Possible. Amsterdam, The Netherlands;
Philadelphia, United States: John Benjamins
Elia, EMT, EUATC, GALA, FIT Europe, & LINDweb. (2018). 2018 Language
Industry Survey — Expectations and Concerns of the European
Language Industry. Retrieved from
https://ec.europa.eu/info/sites/info/files/2017_language_industry_surv
ey_report_en.pdf
García, I. (2012). A Brief History of Postediting and of Research on
Postediting. Revista Anglo Saxonica, 3(2), 293-308
IPMA. (2015). Individual Competence Baseline for Project, Programme &
Portfolio Management (fourth edition). Zurich, Switzerland:
International Project Management Association
ISO. (2012). ISO 21500: 2012 Guide on Project Management. Geneva: ISO
ISO. (2015). ISO 17100: 2015 Translation Services — Requirements for
Translation Services. Geneva: ISO
ISO. (2017). ISO 18587: 2017 Translation Services — Post-editing of Machine
Translation Output — Requirements. Geneva: ISO
Koponen, M. (2016). Is Machine Translation Post-editing Worth the Effort? A
Survey of Research into Post-editing and Effort. JosTrans. The Journal
of Specialised Translation, 25, 131-148. Retrieved from
https://www.jostrans.org/issue25/art_koponen.pdf
Lagarda, A. L., Ortiz-Martínez, M., Alabau, V., & Casacuberta, F. (2015).
Translating without In-Domain Corpus: Machine Translation Post-
editing with Online Learning Techniques. Computer Speech and
Language, 32, 109-134. DOI: 10.1016/j.csl.2014.10.004
Moorkens, J. (2018). What to Expect from Neural Machine Translation: a
Practical In-Class Translation Evaluation Exercise. The Interpreter and
Translator Trainer, 12(4), 375-387. DOI:
10.1080/1750399X.2018.1501639
Moorkens, J., Doherty, S., Kenny, D., & O’Brien, S. (2014). A Virtuous:
Laundering Translation Memory Data Using Statistical Machine
Translation. Perspectives: Studies in Translatology, 22(3), 291-303.
DOI: 10.1080/0907676X.2013.811275
Muzii, L. (2016). Post-editing of Machine Translation for Project Managers.
The Netherlands: TAUS and sQuid
Cristina Plaza-Lara 181
Hikma 19 (2) (2020), 163 - 182
Nitzke, J., Hansen-Schirra, S., & Canfora, C. (2019). Risk Management and
Post-editing Competence. JosTrans. The Journal of Specialised
Translation, 31, 239-259. Retrieved from
http://jostrans.org/issue31/art_nitzke.php
Nunes Vieira, L. (2018). Automation Anxiety and Translators. Translation
Studies. DOI: 10.1080/14781700.2018.1543613
O’Brien, S. (2011). Towards Predicting Post-editing Productivity. Machine
Translation, 25(3), 197-215. DOI: 10.1007/s10590-011-9096-7
Plaza-Lara, C. (2018). Las competencias del gestor de proyectos de
traducción: análisis de un corpus de anuncios de trabajo. Meta : Journal
des Traducteurs, 63(2), 510-531. DOI: 10.7202/1055150ar
PMI. (2017). A Guide to the Project Management Body of Knowledge
(PMBOK® Guide) (sixth edition). Newtown Square, United Kingdom:
Author
Porro Rodríguez, V., Vázquez Morado, L., & Bouillon, P. (2017). Study on the
Use of Machine Translation and Post-editing in Swiss-Based Language
Service Providers. Parallèles, 29(2), 21-35. DOI:
10.17462/para.2017.02.02
Reid, A. (2013). Introduction to Machine Translation and the Post-Editing
Paradigm Shift. Retrieved from
http://www.scottykaye.com/development/sdl/wp-
content/uploads/2013/07/MT_ebook_EN_RH_V7_tcm10-31349.pdf
Rico Pérez, C., & Díez Orzas, P. L. (2013). EDI-TA: Post-editing Methodology
for Machine Translation. Multilingualweb-LT Deliverable 4.1.4. Annex I.
Retrieved from https://www.w3.org/International/multilingualweb/lt/wiki
/images/1/1f/D4.1.4.Annex_I_EDI-TA_Methology.pdf
Rodríguez-Castro, M. (2013). The Project Manager and Virtual Translation
Teams. Critical Factors. Translation Spaces, 2, 37-62. DOI:
10.1075/ts.2.03rod
Sakamoto, A. (2017). No News Is Good News? Translation Spaces, 6(2), 333-
352. DOI: 10.1075/ts.6.2.08sak
Sakamoto, A. (2018). Unintended Consequences of Translation
Technologies: from Project Managers’ Perspectives. Perspectives,
27(1), 58-73. DOI: 10.1080/0907676X.2018.1473452
Sakamoto, A. (2019). Why Do Many Translators Resist Post-editing? A
Sociological Analysis Using Bourdieu’s Concepts. JoSTrans. The
182 How does machine translation and post-editing affect project management? […]
Hikma 19 (2) (2020), 163 - 182
Journal of Specialised Translation, 31, 201-216. Retrieved from
https://www.jostrans.org/issue31/art_sakamoto.php
TAUS & CNGL. (2010). Machine Translation Post-editing Guidelines.
Retrieved from https://www.taus.net/academy/best-practices/postedit-
best-practices/machine-translation-post-editing-guidelines
Thicke, L. (2013). The Industrial Process for Quality Machine Translation.
JoSTrans. The Journal of Specialised Translation, 19, 8-18. Retrieved
from https://www.jostrans.org/issue19/art_thicke.pdf
Vashee, K. (2013). Understanding the Economics of Machine Translation.
Translation Spaces, 2, 125-149. DOI: 10.1075/ts.2.07vas
Zaretskaya, A., Corpas Pastor, G., & Seghiri, M. (2015). Integration of
Machine Translation in CAT Tools: State of the Art, Evaluation and User
Attitudes. SKASE Journal for Translation and Interpretation, 8(1), 76-
88
