ArticlePDF Available

Machine translation in the news: A framing analysis of the written press


Abstract and Figures

Machine translation (MT) is now firmly in the public eye. The media can reflect and influence the public perception of MT and, by extension, of translation itself, but the news coverage of MT has to date remained largely unexplored. This study draws on the news framing literature to present an analysis of how MT is described in the written press. Based on a sample of 284 MT-focused newspaper articles, the news reporting on MT was found to be significantly more positive than negative. This positive framing was unrelated to the launch of neural MT. Furthermore, the portrayal of MT in the press tended to lack nuance, with few instances that raised awareness of the technology’s use implications. The study calls for higher standards in the public discussion and promotion of MT and for more research on non-professional conceptualisations of translation technologies and their role in communication.
Content may be subject to copyright.
Vieira, L.N. (2020) Machine Translation in the News: A Framing Analysis of the
Written Press. Translation Spaces 9(1): 98122. doi:
Machine Translation in the News: A Framing Analysis of the Written Press
Lucas Nunes Vieira
Machine translation (MT) is now firmly in the public eye. The media can reflect and
influence the public perception of MT and, by extension, of translation itself, but the
news coverage of MT has to date remained largely unexplored. This study draws on
the news framing literature to present an analysis of how MT is described in the
written press. Based on a sample of 284 MT-focused newspaper articles, the news
reporting on MT was found to be significantly more positive than negative. This
positive framing was unrelated to the launch of neural MT. Furthermore, the portrayal
of MT in the press tended to lack nuance, with few instances that raised awareness of
the technology’s use implications. The study calls for higher standards in the public
discussion and promotion of MT and for more research on non-professional
conceptualisations of translation technologies and their role in communication.
Machine translation, news coverage, news framing, content analysis, translation
1. Introduction
Freely available MT systems such as Google Translate, Microsoft Translator and
DeepL allow users to obtain translations easily, fast and at no ostensible cost.
Translation research has placed great emphasis on how this technology can affect the
work of professional translators and how it can change human translation processes
and products (see Vieira, Alonso and Bywood 2019). To date, there has been less
research, however, on how the ease of access and widespread availability of MT
might influence societal perceptions of language, translation and multilingual
communication. MT has great potential to facilitate and promote multilingualism, but
its speed and usefulness may also prompt end-users to underestimate the complexities
of translation while overestimating the capabilities of the technology, which in turn
may lead to its misuse.
Importantly, the public perception of MT’s capabilities is likely to be influenced by
how the technology is marketed and publicly discussed. The history of MT is marked
by over-promising of what the technology can do and of how fast it is expected to
advance. In a landmark case from the 1960s, undelivered promises in MT advances
led to a major U-turn in government funding in the United States (see Hutchins 2010).
More recently, MT systems have gone through major improvements in their
architecture thanks to the use of artificial neural networks as a machine learning
methodology (Bahdanau et al. 2015). While superior results are often reported for
neural MT (NMT), a Microsoft research group recently took the bold, if not
conceptually problematic, step of declaring that their system was on a par with human
translation (Hassan et al. 2018). A methodological rebuttal showed that the claim of
human parity was unsound (Toral et al. 2018). Irrespective of subsequent academic
scrutiny, however, this claim was quickly reproduced in the news at the time with
much commotion and little nuance (e.g. Tan 2018; Perez 2018). At the other end of
the spectrum, MT errors have also hit the news recently in situations where
organisations and individuals use the technology in high-stakes contexts where
translation errors pose serious reputational risks (see Kassam 2015; BBC News 2019).
MT is therefore now firmly in the public eye and the news reporting on the
implications of its use is likely to reflect and influence the public perception of
translation technology and, by extension, of translation itself.
Previous research from the social sciences has shown that the way different subjects
are ‘framed’ in the news has great power, including the capacity to shift political
views and public opinion (de Vreese and Boomgaarden 2003; Schuck and de
Vreese 2009). To the author’s knowledge, however, the press coverage of MT has not
been systematically examined to date. This article draws on the news framing
literature to examine how MT is portrayed in newspaper articles where MT-related
keywords appear in the headline as well as in the body of the text. Frames are
generally defined as the angle from which an issue is described or as a specific facet
of the issue that is given more emphasis (de Vreese 2005, 53). In previous research,
news frames have corresponded to strategies used to contextualise eventse.g. a
conflict or attribution of responsibility frame (Semetko and Valkenburg 2000)
or simply to the story’s positive or negative inclination to the issue being recounted
(e.g. de Vreese and Boomgaarden 2003; Schuck and de Vreese 2009). In applying this
concept to MT, the present study has two objectives. First, it investigates if the news
coverage of MT is predominantly positive or negative and whether this is linked to the
articles’ date of publication. Second, the study qualitatively explores frequent themes
used to contextualise the reporting on MT in news items deemed to be positive,
negative and neutral. The discussion also raises broader societal issues linked to the
implications of MT use and of how this technology is publicly portrayed.
Since the study concerns use and perceptions of MT by the general public, rather than
by translators, it should be noted that most settings described here involve the use of
raw MT outputs as an end-product. Post-editing or human translation practices are
therefore not a prominent factor in the analysis. The study does not make a priori
distinctions between different use cases or types of MT system, however. Details of
this nature were largely allowed to emerge organically from the data based on the
news items themselves (see Section 4).
In the remainder of the paper, Section 2 provides a brief review of previous research
on perceptions of MT and on news framing. The article’s methodology is then
described in Section 3. The results are presented and discussed in Section 4 and,
finally, the article concludes in Section 5 with a summary of the findings and
suggestions for future research.
2. Review of Literature
Most research on perceptions of MT to date has concentrated on the views of
translation professionals (e.g. Cadwell et al. 2017; Guerberof 2013; Vieira 2020;
Rossi and Chevrot 2019). With a few exceptions (see e.g. Guerberof 2013), translators
are largely found to have negative attitudes to the use of MT in the translation process
(e.g. Moorkens and O’Brien 2015; Vieira and Alonso 2020; Läubli and Orrego-
Carmona 2017). This can be associated with MT’s unsuitability for certain tasks
(Cadwell et al. 2017) as well as with broader issues linked to the economy and the
structure of the translation industry (Vieira 2020).
Another strand of research on MT use has concentrated on the importance of
promoting awareness of MT’s limitations with a view to improving ‘literacy’ in the
technology (Williams 2006; Bowker and Buitrago Ciro 2019). There has also been
research on how MT is used in ‘everyday’ communicative acts and on what can be
learned from the public’s experience with MT in these contexts (Nurminen 2018).
These studies discuss important aspects of how MT is used and how this use can be
improved. Unlike the present analysis, however, they are not concerned with how
opinion-shaping, mass-distribution outlets contextualise the technology. The news
framing literature offers a useful framework for analysing this type of phenomenon.
To identify frames in the news, previous studies have usually adopted one of two
approaches. The frames can emerge inductively from the content itself (e.g. Neuman
et al. 1992) or they can be deductively assigned to the material based on a pre-
conceived definition of the frames of interest (e.g. Semetko and Valkenburg 2000). In
more recent research, scholars have largely favoured the deductive approach. This
approach is considered more reliable because it requires the frames to be
operationalised prior to the analysis, which avoids inconsistency in the frames’
identification (de Vreese 2005, 53). To identify the frames deductively, researchers
often draw on specific elements or ‘focal points’ of the content, such as headlines or
final statements in a news article, to establish if the material contains characteristics of
the frame(s) (ibid., 54). Studies based on this methodology usually examine the extent
to which certain frames are present in the news. In terms of the nature of findings
reported by these studies, of particular relevance to the present analysis are
investigations that classify frames based on their valence, i.e. whether they are
positive or negative (e.g. de Vreese and Boomgaarden 2003; Schuck and de
Vreese 2009). Some of this work presents empirical evidence showing that the extent
to which news stories are positively or negatively framed can influence the valence of
opinions held by those who consume the stories (de Vreese and Boomgaarden 2003,
373). These studies also provide a methodological framework that can be applied to
the present analysis, where the objectives also involve identifying valenced frames in
the press. With previous news framing research as a backdrop, the present study for
the first time applies this framework to an investigation of MT use and its
implications. The procedure for sampling and analysing the content is discussed in
detail below.
A slightly different conceptualisation of ‘framing’ has been applied by studies based on narrative
theory (Baker 2006; Harding 2012). Unlike the present article, these studies are usually interested in
translations themselves and how they can be used to legitimise a stance or advance a specific agenda
(Harding 2012, 287). These studies are therefore not reviewed here in detail.
3. Methodology
3.1 Sampling
The Newspapers section of the Westlaw database (Thomson Reuters 2020) was used
to identify and retrieve newspaper articles with a focus on MT. Westlaw is an online
resource that includes a large database of print and online news from multiple
countries that can be searched based on detailed criteria. The search relied on a series
of keywords that included different ways of referring to MT technology, including the
names of popular MT systems, such as Google Translate and Microsoft Translator.
Variations of the term ‘babel fish’, for instance, were also included because of
previous MT systems with this name and because of the fictional alien fish with
multilingual powers from The Hitchhikers Guide to the Galaxy (Adams 1979), which
is often used as a symbol for MT technology. The full search equation is presented
and explained below. The search was case insensitive.
advanced: (machine translation OR automated translation OR
automatic translation OR machine translator OR automatic
translator OR automated translator OR google translate OR babel
fish OR babelfish OR microsoft translator OR bing translator OR
systran) & TI(machine translation OR online translation OR
automated translation OR automatic translation OR machine
translator OR online translator OR automatic translator OR
automated translator OR google translate OR babel fish OR
babelfish OR microsoft translator OR bing translator OR systran) %
The search was designed to ensure a sample with wide-ranging coverage, but which
reflected the topic as precisely as possible. The operator ‘OR’ searched for any of the
terms between brackets.
The search was set to return only records where the
keywords appeared both in the text and in the headlines indexed as the articles’ title,
which helped to concentrate the analysis on news stories that had MT as a central
subject. This was achieved by using the operator ‘&’ (‘and’) coupled with ‘TI’ (’title).
The sequence ‘% SO(patent)’, in turn, excluded publications with ‘patent’ in the
Due to restrictions affecting the length of the search equation, the keywords searched in the text did
not include ‘online translation’ and ‘online translator’, which were among the keywords searched in the
title. This asymmetry was not found to affect the search coverage.
name. This term matched exclusively the periodical Plus Patent News to
counterbalance a disproportionate number of records from this source (40% of the
items initially retrieved). This is a high-frequency periodical, indexed under ‘Pakistan
Newspapers’, which publishes news snippets about patents. Excluding this source
helped to balance the sample and focus the analysis on MT use at large rather than on
patent-specific issues. Lastly, an English-language filter was applied to the results to
exclude records in other languages that had been matched to some of the keywords.
Since it would be difficult to exhaustively represent all languages in which this
subject appeared in the news, the language of publication is an inevitable constraining
factor for analyses of this nature. Focusing the present analysis on English-language
records in any case provided consistency and helped to diversify the spread of sources
by including newspapers that are available in English even if this is not an official
language in the country of publication (see Appendix).
The above procedure returned 334 records. A series of manual filters were then
applied. Thirty-seven records were excluded because of a poor fit to the topic (e.g.
where ‘Babelfish’ was the name of a restaurant or where the article was about human
translation only). Nine records were excluded because access to the full article was
restricted by the database. Three records were excluded because they were duplicates,
and one additional record was excluded because it was a non-English record missed
by the language filter. The resulting sample contained 284 records in total. This was
the sample retained for analysis and the basis for all information provided in the
remainder of the paper.
The sample includes records from thirty countries. The most represented countries
were the UK (N=108), the USA (N=67) and India (N=31). English-language articles
from countries where English is not an official language include records from Japan
(N=8), China (N=7) and South Korea (N=5). The articles were published between
1986 and 26 June 2019, when the search was carried out. No date filter was applied,
so this date range represents the range in the database itself at the point the search was
conducted. A full list of the newspapers is provided in the Appendix.
3.2 Content Analysis
The content analysis can be divided into two stages. First, valenced frames were
deductively identified to assign the articles with scores reflecting their degree of
positive or negative framing. An inductive qualitative examination was then carried
out to explore the thematic aspects that underpinned the articles’ framing. All the
analysis was conducted by the present author with the involvement of an external
coder to check inter-annotator agreement in the identification of valenced frames (see
As per previous news framing research (Schuck and de Vreese 2006; 2009), the
procedure for estimating the extent to which the articles were framed positively or
negatively was based on a series of binary ‘yes/no’ questions that reflect the presence
(value of 1) or absence (value of 0) of positive and negative elements in the
newspaper articles. Six pairs of questions were originally devised for this purpose.
Each question was regarded as a potential item to be included in two multi-item scales
reflecting the articles’ positive and negative framing, respectively. The full articles
were the unit of analysis in all cases. Mokken scale analysis (MSA) (Mokken 1971)
was applied to check if the negative and positive questions formed strong and reliable
scales that could be used as a proxy for the articles’ framing. MSA is a checking and
validating procedure that is fit for binary-coded items. Specifically, it ensures that the
different items (in the present study, the questions) devised to reflect a latent concept
(in the present study, the negative or positive framing of the news articles) are
reliable, consistent with each other and can have their scores added up or averaged to
form a multi-item scale. By using MSA as a diagnostic device, two of the questions
originally used were merged and two others were discarded,
which resulted in three
question pairs used to measure the articles’ negative and positive framing,
respectively. The questions are presented below.
Negative Q1: Does the article contain negative phrases about MT in the
Negative Q2: Does the article place more emphasis on MT’s drawbacks
than on its benefits, portray MT technology as something
predominantly negative/ineffective or discourage its use?
Negative Q3: Does the article use negative phrases to describe MT
results or mention negative consequences of using MT?
Positive Q1: Does the article contain positive phrases about MT in the
The discarded questions were ‘Does the article include negative [or positive] phrases about MT or its
use consequences in a quote?’ and ‘Does the article include negative [or positive] phrases about MT or
its use consequences in its final statement?’.
Positive Q2: Does the article place more emphasis on MT’s benefits
than on its drawbacks, portray MT technology as something
predominantly positive/effective or encourage its use?
Positive Q3: Does the article use positive phrases to describe MT
results or mention positive consequences of using MT?
The two question groups formed strong and reliable scales that passed all
recommended MSA tests (see Stochl et al. 2012) (negative frame: scale coefficient H
= 0.88, reliability Rho = 0.84, scale Z = 16.83; positive frame: scale coefficient H =
0.88, reliability Rho = 0.80, scale Z = 14.18; N = 284).
An external researcher was
asked to provide binary answers to these same questions for a random sample of
approximately 10% of the newspaper articles (N = 28). The average level of
agreement with the external coding was 80% (88% for the negative framing questions
and 73% for the positive framing questions).
To form positive and negative frame scales, the average of the binary codes was taken
for each question group. For an article where the negative questions had been
answered with the codes 0-1-1, for instance, the negative score for that article was
0.667 (i.e. the average of the codes). This means that each newspaper article in the
sample was assigned with a negative frame score and a positive frame score, which
ranged between 0 (frame not present) and 1 (frame strongly present) (for a similar
frame scoring procedure, see Schuck and de Vreese 2009, 47).
4. Results
4.1 Valenced frames: quantitative analysis
The frame scores allowed the articles to be ranked and compared to each other in
relation to their valenced framing. The scales were analysed directly, as numeric
scores, and indirectly as the parameter for a nominal classification. An article was
classed as positive if its positive frame score was higher than its negative frame score.
Similarly, articles were classed as negative if the negative score was higher than the
positive score. Neutral articles were those where negative and positive frame scores
The values of the scale coefficient H and reliability Rho have a maximum of 1 and the closer to 1 the
stronger and more reliable the scale is deemed to be. The high Z scores, in turn, reflect the significance
of the H coefficients (for more details on Mokken scale analyses, see Stochl et al. 2012).
were the same (i.e. ties). Table 1 presents the proportion of positive, negative and
neutral articles in the sample.
Proportion of
articles (%)
CI lower
bound (%)
CI upper
bound (%)
Table 1. Proportions and 95% confidence intervals
(CI) for articles classed as
positive, negative and neutral. N = 284
As shown in Table 1, most newspaper articles in the sample were positively framed.
For 57% of the articles (N = 162), the positive framing was stronger than the negative
framing. For 29% of the articles (N = 83), stronger negative framing was observed.
Positive and negative frame scores were the same for 14% of the articles (N = 39).
When considering the numeric scores directly, results of a Wilcoxon paired test
confirmed that the articles’ positive and negative frame scores were significantly
different (p < 0.01; PSdep = 0.66)
. The size of this effect is expressed here as the
probability of superiority for dependent groups (PSdep), which expresses the
probability of newspaper articles on MT being predominantly positive (Grissom and
Kim 2011, 172174). This measure (0.66) is equivalent to the rebased proportion of
positive articles (the largest group) when ties (i.e. neutral articles) are excluded.
These results indicate that the coverage of MT in the written press is significantly
more positive than negative. A comparison of the median negative and positive frame
scores across all articles is presented in Figure 1.
Throughout the study, decimal sample proportions used in any calculations are presented as
percentages in the tables to make results easier to interpret.
Confidence intervals were computed with the MultinomCI function of the DescTools R package. The
lower and upper bounds reflect the range of values where true population proportions are likely to fall.
Conducted with the coin R package, using the Pratt method.
Confidence intervals (95%, based on the BinomCI function of the DescTools R package) of the
rebased proportion of positive articles (excluding ties): 0.60, 0.72. Bootstrap resampling was also used
to obtain the confidence intervals, but the results were largely the same.
Another method for handling ties in these cases is to add half the count of ties to the number of
positive cases (Grissom and Kim 2011, 172174). This corresponds to PSdep = 0.64.
Figure 1. Median score for negative (left) and positive (right) frames.
To check if the date of publication of the articles was a factor in their framing, the
data was also analysed for two sub-groups of articles: those published before 2016
and those published from 2016. The year of 2016 was the sample’s median year of
publication. This year also coincides with Google’s public release of its neural
systems (Wu et al. 2016). Google’s roll-out of NMT systems made NMT gain wide
public attention, which could have been a factor in the framing of this subject in the
news. The proportions of positive, negative and neutral articles corresponding to the
period before and from 2016 are presented in Table 2.
Before 2016
From 2016
of articles
CI lower
CI upper
of articles
CI lower
CI upper
Table 2. Proportions and 95% confidence intervals (CI) for positive, negative and
neutral newspaper articles published before 2016 (N = 125) and from 2016 (N = 159).
Table 2 shows that most articles in the two date ranges were predominantly positive
i.e. the overall pattern presented in Table 1 can also be observed when articles
published only before or only from 2016 are considered. It can be noted, however,
that the proportion of positive articles was slightly higher before 2016. Conversely,
the proportion of negative articles had a slight increase from 2016. Kendall’s rank
correlation tests showed a very small but significant negative association between the
year of publication (treated as a direct numeric variable) and the positive frame score
(tau = -0.13, p < 0.01). No significant effect was observed for the negative frame
score (tau = 0.06, p = 0.22). A non-parametric multivariate analysis checking the
impact of date range (i.e. before and from 2016) on both the negative and positive
frame scores showed no significant effects.
These results have at least two implications. First, while the negative association
between the year of publication and the positive frame score suggests that the overall
news coverage of MT became less positive with time, the association is too weak (-
0.13) to warrant any conclusive interpretations. This result is regarded here as
negligible until otherwise confirmed by a larger sample. Second, what the
insignificant results do suggest is that NMT is unlikely to have made the news
reporting on MT become more positive. In fact, these results show that the positive
news framing of MT is not a recent phenomenon. The coverage of MT was more
positive than negative irrespective of the date of publication.
To gain further insight into how the articles were framed, prominent features of
positive, negative and neutral articles are qualitatively discussed below in
sections 4.2, 4.3 and 4.4.
4.2 Positive articles
In broad terms, positive articles were often focused on how MT technology is
advancing or on how it can be useful. The process of reading and analysing the
articles suggested Google Translate was frequently discussed in articles with a
positive framing. Further inspection showed that the word ‘Google’ appeared in the
headlines of 65% of positive articles. Of articles with the maximum positive score
(i.e. 1), 64% mentioned ‘Google’ in the title. A substantial amount of news traffic on
MT is therefore dedicated to a positive account of how Google Translate is improving
and how it can be beneficial. Among these articles, some focused on the technological
aspects of Google Translate’s updates, with headlines such as Google Translate is
about to become really accurate (Turner 2016) or Google Translate’s mobile app
just got even more magical (Ma 2015). Of these two articles, the former covered
Google Translate’s launch of NMT while the latter covered the release of a camera
feature that allows users to obtain translations of signs and other print material by
This analysis was conducted with the nonparatest function of the npmv R package, which examines
the effect of a single explanatory variable (in this case, period of publication) on multiple response
variables (in this case, negative and positive frame scores) (ANOVA Type test = 2.1, p = 0.13).
using the camera on their phones. This same camera functionality was described in
another article as a new, very snazzy feature (Tsukayama 2015). Although Westlaw
does not automatically index details of the specific news sections where the articles
appeared, a manual inspection suggested that articles focused on improvements and
technological advances were often published under ‘Science’ or ‘Technology’. It was
also noted that this type of article sometimes relied heavily on information provided
by MT developers themselves (e.g. Beall 2016; Turner 2016).
Among articles where the positive framing of MT was strong, at times the underlying
assumption was in fact that MT had infallible, if not ‘magical’ (see above), powers.
This is implied in headlines such as Google Translate now lets you chat to anyone in
a foreign language (Titcomb 2016) and Google Translate live translation upgrade to
let anyone speak any language, any time (Griffin 2015) In some of these articles, the
lack of nuance in the description of what MT can do is objectively inaccurate. The
claim that Google Translate can translate ‘any language’ is simply incorrect, for
Another type of positive reporting on MT consisted of articles that tried to elicit
empathy by telling stories with a clear human appeal. These included titles such as
Franklin cop uses Google Translate to talk to Chinese theft victim
(Muscavage 2019), Couple who dont speak same language fall in lovethanks to
Google Translate (Jolly 2019) and Google Translate: the unlikely World Cup hero
breaking barriers for fans (Ames 2018). In this group of articles, it is worth noting
how the stories paint MT in a positive light while at the same time reflecting some
level of surprise at how MT proved to be useful in real, often high-stakes,
communicative contexts. This is clear in how MT is described as an ‘unlikely hero’.
The very fact that the stories were published also implies that MT’s success in these
cases is unusual or unexpected and therefore something newsworthy. In these cases,
the articles are not directly motivated by technological advances, but rather by MT’s
impact on communication.
Associations between humans and machines were also, directly or indirectly, a theme
explored in the articles. It is worth noting how MT is personified or regarded as a
human-like feature in some of the stories, for instance by being described as a ‘hero’.
Something similar can be noted in the headline Just me, you and Google Translate
(Irish Independent 2012), where Google Translate is alluded to as the third person in a
trio. Among other reasons, articles that personify MT or regard it as a human-like
technology are notable especially because of how they implicitly attribute agency to
MT systems. Conceptually, this is consistent with theoretical frameworks that
challenge a distinction between human and non-human factors in understanding the
effects of technology on society (Ballantyne 2015). These frameworks, in particular
Actor Network Theory, conceptualise agency as a property that can be distributed
between human and non-human elements in a social network (see Latour 1996). An
in-depth analysis of MT use from this prism is beyond the scope of the present article.
Nevertheless, it is worth pointing out how public portrayals of MT as a human-like
technology may, advertently or inadvertently, entail expectations that the technology
should behave or be as effective as a human. This is clear in a headline that
announced ‘human parity’ even before this was claimed by Microsoft: Google
Translate now as good as a human’—could it save you on holiday? (Parsons 2016).
Similarly, in an article that reports on the availability of Swahili in Microsoft
Translator, it is mentioned that Microsoft is veering away from the human
translator,” which is described as having cost benefits for local governments in Africa
that wish to make documents available in different languages (Kuria 2015). The
benefits of MT in these cases notwithstanding, the article shows no attempts to temper
its stance or draw attention to the fact that MT also has limitations and may lead to
unexpected problems when used without post-editing in governmental settings and
other formal circumstances (see Section 4.3).
Generally, articles with a positive framing tended to focus on technical aspects of MT
technology or on MT’s usefulness rather than on textual characteristics of MT
outputs, which were more salient among articles with a negative framing (see
Section 4.3). While some articles made the point that MT also had shortcomings, the
articles with a predominantly positive framing largely lacked nuance and often put
forward conceptualisations of MT as a technology with at times human-like, at times
‘magical’ powers. Some articles just disseminated information provided by MT
developers. This also raises important questions about the role of the press in
technological news and about how it can act as a conduit for usually non-peer-
reviewed research published through MT companies’ blogs.
4.3 Negative articles
Unlike articles with a predominantly positive framing, negative articles tended to
focus on linguistic errors and, sometimes, on their consequences. These articles were
often published by tabloids. Some of them appealed to humour or attempted to shock
readers by calling attention to the unusual nature of certain machine translations.
Among these, there were articles with headlines such as Egg-streme sports:
Norways Olympic squad accidentally orders 15,000 eggs after asking Google
Translate how to say 1,500 in Korean (Sear 2018) or Google Translate users
TERRIFIED as site reveals disturbing DOOMSDAY prophecies (capitals in original,
Blair 2018). There were also articles that were in direct opposition to positive articles
that stressed how MT had been useful in high-stakes settings. These included articles
with headlines such as Patients at risk in care home where staff who cant
understand English use Google translate to speak to residents (Hudson 2019) and
Stop using Google Translate! Wrexham Council staff warned after complaints over
Welsh language errors
(Randall 2018). These articles highlighted the consequences
of MT errors in serious or formal contexts.
Some negative articles implied that the general level of awareness of how MT
systems work is often low, which has implications for what users might expect of the
technology. In relation to a case where a translation error ended up on the website of a
town in Galicia, it is mentioned that the town was considering legal action against
Google for the mistranslation (Saxena 2015). Here it is worth noting how
mistranslations can be regarded as unexpected issues rather than occurrences that are
in the nature of MT. A different article covering this same story provided explanations
such as Google Translate is an automatic translator, meaning it works without the
intervention of human translators (Eleftheriou-Smith 2015). A Google expert is
quoted in this article saying, “[s]ince the translations are generated by machines, not
all translation will be perfect and sometimes there will be mistakes or mis-
translations (Eleftheriou-Smith 2015).
While in professional translation settings statements saying that MT is not perfect
may come across as obvious, it was clear from articles with a negative framing that,
when users encounter MT use issues, they can feel misled. This raises questions
concerning the attribution of responsibility for MT errors. On the one hand, as seen
above, MT is often portrayed as a ‘magical’ technology with infallible powers that
allow users to translate ‘any language’. As previously mentioned, this can follow
These headlines appeared in banners at the top of the page which were provided in addition to the
articles’ titles.
claims announced by MT developers themselves. On the other hand, however, when
problems arise, quotes by major commercial developers state that MT is by nature
imperfect and that errors should be expected. One article covering the case involving
the town in Galicia indirectly suggests that MT use is a matter of choice and that users
should know when to hire professional translators instead: the gaffe is proof that the
Internet is no substitute for a professional translator (Saxena 2015). This pushes the
burden of responsibility closer to users and further away from developers. From the
user’s perspective, however, the overall message is conflicting, so it is not surprising
that they might consider legal action when MT errors have serious consequences.
These cases expose a need for greater awareness of MT’s limitations and clearer
guidelines on the situations where the technology should and should not be used, and
on who should be held accountable for cases where poor MT has serious
Generally, articles that framed MT negatively tended to exploit humour or situations
of distress to provoke a reaction in the readership. While some articles alluded to
broader questions about how MT can affect language learning (Stapleton 2019) and
how its problems may be down to ‘misuses’ rather than issues that are intrinsic to the
technology (Wooten 2011), these articles were the exception rather than the rule. A
lack of nuance is therefore also apparent among negative articles.
4.4 Neutral articles
As previously mentioned, articles classed as ‘neutral’ were those that had equal
positive and negative frame scores. As shown in Section 4.1, neutral articles
represented 14% of the sample. Of these, the majority (67%) were articles where
positive and negative frames were both absent (i.e. where both frames had a score of
zero). These articles often described MT-related developments without advancing a
particularly negative or positive argument. Some of them were published by an
institutional source rather than by a specific reporter (e.g. Korea Times 2016;
LETA 2016). Others were slightly different from the rest of the sample and deserve
special attention. These were articles that mentioned MT use in relation to crimes.
These articles were published either in tabloids or in regional newspapers. They
included, for instance, Sex attacker used Google Translate to search I am interested
in you before attacking schoolgirl’” (Armstrong 2016) and Sham wedding gang
who made £500,000 marrying bogus couples who had to communicate via Google
Translate are jailed for 20 years (Boyle 2018).
Based on the headlines, it is plausible to expect that these articles provide more details
of how exactly MT was used or why its use is important in relation to the reported
events. Surprisingly, however, MT is mentioned in the body of the text in these
articles usually once or twice more, often just to restate the information already
provided in the headlinei.e. that individuals linked to the crime used MT to
communicate. While MT use may be construed negatively by these articles’ readers,
based on the deductive frame identification method adopted here, there was no basis
to classify these articles as either positive or negative because the articles stance on
MT itself or its usefulness is not explicit. In one similar case, MT had been used by
the victim (to report the crime) rather than by the perpetrator (Morrison 2016).
readers may construe MT positively, but again there was no basis to classify the
article’s framing as either positive or negative since the article does not go into detail
about its stance on the technology itself or its implications. Nine neutral articles
overall followed this format and associated MT with those who had been involved in
a crime as victim, perpetrator or some connected third party. This was not therefore a
common phenomenon, but these articles raise relevant questions about the different
ways in which MT, and communication technologies in general, can be construed and
associated with different social stereotypes. In the examples mentioned above,
especially where MT had been used in criminal activity, the mention of MT in the
headline could be seen as an attempt to portray those committing the crime as having
a ‘foreign’ or ‘outsider’ status.
In addition to articles with no negative or positive framing, there were also articles
where the positive and negative frames were present in equal measure. As mentioned
above, these articles were also treated as neutral. There were only thirteen of them in
the sample. These articles can be regarded as having a balanced stance on MT by
giving comparable emphasis to its positive and negative points. Among these, there
were articles where the headline was a questione.g. Can machine translation stand
alone?(Jun-ho 2019); Are we using Google Translate responsibly? (Stachova
This article was indexed by Westlaw without a named author. Some inconsistencies of this nature
were noted for online articles. They were most likely due to updates published after a copy of the
article had been retained by the database and are not deemed to influence the analysis in any way.
References provided here are up to date at the time of writing.
2019); Is Google Translate improving or ruining the way we travel?
(Dickinson 2018)as well as articles where negative and positive points were both
included in the headline itself: e.g. blunders and successes (Wooten 2010); some
success and serious limitations (Nikkei Weekly 1993). These articles tended to
consider the strengths and the limitations of MT in more depth.
Due to their small number, it is not easy to discern patterns in the group of neutral
articles with equal positive and negative scores. It was noted, however, that these
articles spread across a relatively wide temporal range (from 1993 to 2019). It is also
worth noting that one article in this group was published by a Translation Studies
academic (Kockaert 2019), another was an opinion piece by a professional interpreter
(Jun-ho 2019) and another was an article by the director of translation services at a
company (Wooten 2010). These three articles raised awareness of the complex and
context-dependent nature of the concept of translation quality. They also highlighted
how human and machine translation can complement each other and how the issue of
MT’s usefulness is not necessarily to do with a dichotomy between humans and
machines. The fact that these authors were translation professionals most likely
explains the more carefully considered stance of these articles. While the present
paper does not expect the news coverage of this topic to have a widespread
educational agenda, the fact that arguments of this nature were rare in the sample
suggests that the public discussion of MT in the news often fails to take into account
important factors that are usually considered by experts in language and translation.
5. Conclusion
This study presented a framing analysis of the newspaper coverage of MT. Based on a
sample of 284 newspaper articles where different MT-related keywords are mentioned
in the text and in the headlines, the results showed that the coverage of MT in the
written press tends to emphasise the positive aspects of the technology. NMT was not
found to be associated with a significant increase in the positive tone of the news
reporting on MT. In fact, MT’s positive framing was observed irrespective of the date
of publication.
Positive articles emphasised MT’s benefits especially by describing technological
advances or positive consequences of MT use. Some of these articles inflated the
capabilities of the technology by comparing it to human translators or by implying
that MT had the power to make users speak and understand any language without the
prospect of encountering any issues. Negative articles were more narrowly focused on
the MT output and errors that could elicit humour or shock. These articles largely
missed opportunities to discuss ways in which MT-related issues could be minimised
or avoided. Given the overstating tendencies of positive articles and the often
sensationalist tone of negative articles, this paper does not deem the news reporting on
MT technology to be a fair account of MT’s strengths and limitations or of the
implications of its use.
While measuring the specific effects of the news frames on the readership is beyond
the scope of this paper, some of the positive articles analysed put forward
oversimplified notions of language that risk exacerbating misuse of MT. MT errors
can have reputational and professional consequences and, in covering cases where
these arose, the newspapers did little more than exploit the public appeal of the story.
Some positive articles acted as conduits for information disseminated by MT
companies themselves, which raises further questions about a potential lack of public
scrutiny of MT developers’ claims. As MT becomes increasingly available across
platforms and devices, more attention should be drawn to questions of agency and
attribution of responsibility for MT errors or problems resulting from its misuse.
Given MT’s ease of access, low awareness of its benefits and limitations is an issue of
potentially great magnitude.
The results above also point to a divide between expert and non-expert perceptions of
MT technology and of its potential. In recent years, translation technology research,
especially research on human factors in MT use, has been predominantly focused on
professional issues or issues related to translation as a process or practice. Meanwhile,
MT has become a mainstream communication tool. While previous research indicates
that professional translators’ attitudes to MT are usually negative (see Section 2), the
results presented here show that the public coverage of this technology veers towards
the other end of the spectrum. While news articles are just one data source on public
conceptualisations of MT, the coverage of MT in the news is an important indicator of
how this technology is portrayed to non-linguists and the wider public. The results
reported here therefore call for research and other initiatives that can promote fruitful
and informed uses of MT. There is a need for the MT and professional translation
community, including industry and academia, to engage with the wider population to
raise awareness of how MT works and of what it can do. This has implications for
several other subjects where MT-mediated communication may be a factor, including
migration, social media use, political discourse and language policy. It is hoped that
future research on these topics will shed further light on how translation technologies
might influence attitudes to translation and, more broadly, to language.
Adams, Douglas. 1979. The Hitchhiker's Guide to the Galaxy: Hitchhiker's Guide to
the Galaxy. Basingstoke: Pan Macmillan.
Ames, Nick. 2018. Google Translate: The Unlikely World Cup Hero Breaking
Barriers for Fans. The Guardian, 11 July 2018.
Armstrong, Jeremy. 2016. Sex Attacker ‘Used Google Translate to Search I Am
Interested in You Before Attacking Schoolgirl. Mirror Online, 29 September 2016.
Bahdanau, Dzmitry, Kyunghyun Cho, and Yoshua Bengio. 2015. Neural Machine
Translation by Jointly Learning to Align and Translate. Accessed 3 March 2020,
Baker, Mona. 2006. Translation and Conflict: A Narrative Account. London:
Ballantyne, Neil. 2015. Human Service Technology and the Theory of the Actor
Network. Journal of Technology in Human Services 33 (1):104117.
DOI 10.1080/15228835.2014.998567.
BBC News. 2019. Translation Errors Force Osaka Metro Websites Offline. BBC
News, March 19, 2019. Accessed 28 October 2019,
Beall, Abigail. 2016. Google Translate Just Got Better: AI-Powered Software Is
Now Almost as Good at Translating Chinese as Humans. Daily Mail Online, 28
September 2016.
Blair, Anthony. 2018. Google Translate Users TERRIFIED as Site Reveals
Disturbing DOOMSDAY Prophecies. Daily Star Online, 23 July 2018.
Bowker, Lynne, and Jairo Buitrago Ciro. 2019. Machine Translation and Global
Research: Towards Improved Machine Translation Literacy in the Scholarly
Community. Bingley: Emerald Publishing Limited.
Boyle, Darren. 2018. Sham Wedding Gang Who Made £500,000 Marrying Bogus
Couples Who Had to Communicate via Google Translate Are Jailed for 20 Years.
Daily Mail Online, 4 April 2018.
Cadwell, Patrick, Sharon O’Brien, and Carlos S. C. Teixeira. 2017. “Resistance and
Accommodation: Factors for the (Non-) Adoption of Machine Translation Among
Professional Translators. Perspectives 26 (3): 121.
DOI 10.1080/0907676X.2017.1337210
de Vreese, Claes H. 2005. News Framing: Theory and Typology. Information
Design Journal + Document Design 13 (1): 5162.
de Vreese, Claes H., and Hajo Boomgaarden. 2003. Valenced News Frames and
Public Support for the EU. Communications 28 (4): 361381.
DOI 10.1515/comm.2003.024
Dickinson, Greg. 2018. Is Google Translate Improving or Ruining the Way We
Travel? Telegraph Online, 12 April 2018.
Eleftheriou-Smith, Loulla-Mae. 2015. Google Translate Error Sees Galician Town of
As Pontes Accidentally Celebrating ‘Clitoris Festival. Independent Online, 3
November 2015.
Griffin, Andrew. 2015. Google Translate Live Translation Upgrade to Let Anyone
Speak Any Language, Any Time. Independent Online, 12 January 2015.
Grissom, Robert J., and John J. Kim. 2011. Effect Sizes for Research: Univariate and
Multivariate Applications, 2nd edition. London: Routledge.
Guerberof, Ana. 2013. "What Do Professional Translators Think About Post-
Editing? JosTrans. The Journal of Specialised Translation 19:7595.
Harding, Sue-Ann. 2012. “‘How Do I Apply Narrative Theory?’ Socio-Narrative
Theory in Translation Studies. Target 24 (2): 286309.
Hassan, Hany, Anthony Aue, Chang Chen, Vishal Chowdhary, Jonathan Clark,
Christian Federmann, Xuedong Huang, Marcin Junczys-Dowmunt, William Lewis,
Mu Li, Shujie Liu, Tie-Yan Liu, Renqian Luo, Arul Menezes, Tao Qin, Frank Seide,
Xu Tan, Fei Tian, Lijun Wu, Shuangzhi Wu, Yingce Xia, Dongdong Zhang, Zhirui
Zhang, and Ming Zhou. 2018. Achieving Human Parity on Automatic Chinese to
English News Translation. Accessed 3 March 2020,
Hudson, Rebecca. 2019. Glenside Care Home Pembroke Lodge Rated 'Inadequate'
by Care Quality Commission. Salisbury Journal, 17 January 2019.
Hutchins, John. 2010. Machine Translation: A Concise History. Journal of
Translation Studies. Special issue: The Teaching of Computer-Aided Translation 13
(1 and 2) :2970.
Jolly, Bradley. 2019. Couple Who Don't Speak Same Language Fall in Love
Thanks to Google Translate. Mirror Online, 9 January 2019.
Jun-ho, Lee. 2019. Can Machine Translation Stand Alone? The Korea Times, 2
March 2019.
Kassam, Ashifa. 2015. Google Translate Error Sees Spanish Town Advertise Clitoris
Festival. The Guardian, 3 November 2015. Accessed 28 October 2019,
Kockaert, Hendrik J. 2019. Machine Translation Still Needs Human Aid. China
Daily Global, 14 June 2019.
Korea Times. 2016. Automated translation app available for PyeongChang
Olympics. The Korea Times, 18 July 2016.
Kuria, Winfred. 2015. Swahili the First African Language on Microsoft Translator.
Nation, 21 October 2015.
Latour, Bruno. 1996. On Actor-Network Theory: A Few Clarifications.Soziale
Welt 47 (4):369381.
Läubli, Samuel, and David Orrego-Carmona. 2017. When Google Translate is Better
than Some Human Colleagues, those People Are no Longer Colleagues. In
Proceedings of the 39th Conference Translating and the Computer, London, UK, 16
17 November, 5969.
LETA. 2016. Riga to Host Annual Conference of European Association for Machine
Translation at End-May. Latvia National News Agency, 6 April 2016.
Ma, Alexandra. 2015. Google Translate’s Mobile App Just Got Even More
Magical. Huffington Post, 29 July 2015.
Mokken, R.J. 1971. A Theory and Procedure of Scale Analysis: With Applications in
Political Research. The Hague: Mouton.
Moorkens, Joss, and Sharon O’Brien. 2015. “Post-Editing Evaluations: Trade-Offs
Between Novice and Professional Participants. In Proceedings of European
Association for Machine Translation (EAMT) 2015, Antalya, edited by İlknur Durgar
El-Kahlout, Mehmed Özkan, Felipe Sánchez-Martínez, Gema Ramírez-Sánchez, Fred
Hollowood, and Andy Way, 7581.
Morrison, Sean. 2016. Teenager ‘Dragged From Pizza Restaurant, Raped and
Stabbed and Then Used Google Translate to Report Crime. Mirror Online, 22
November 2016.
Muscavage, Nick. 2019. Franklin Cop Uses Google Translate to Talk to Chinese
Theft Victim. Courier News, 18 February 2019.
Neuman, Russell, Marion R. Just, and Ann N. Crigler. 1992. Common Knowledge:
News and the Construction of Political Meaning. Chicago and London: The
University of Chicago Press.
Nikkei Weekly. 1993. Automatic Translation Approaching Reality: Some Success
Achieved, but Systems Still Have Serious Limitations. Nikkei Weekly, 22 February
Nurminen, Mary. 2018. Machine Translation in Everyday Life: What Makes FAUT
MT Workable? TAUS eLearning Blogs, 29 March 2019. Accessed 3 March 2020,
Parsons, Jeff. 2016. Google Translate Now ‘As Good As a Human Could It Save
You on Holiday? Mirror, 30 September 2016.
Perez, Sarah. 2018. Microsoft Announces Breakthrough in Chinese-to-English
Machine Translation. Tech Crunch. Accessed 5 September 2019,
Randall, Liam. 2018. Wrexham Council Staff Told to Stop Using Google Translate
because of Complaints over Welsh Language Errors. The Leader, 20 December
Rossi, Caroline, and Jean-Pierre Chevrot. 2019. Uses and Perceptions of Machine
Translation at the European Commission.JosTrans. The Journal of Specialised
Translation 31: 177200.
Saxena, Jaya. 2015. Google Translate Hilariously Messed up the Invite for This
Regional Festival. The Daily Dot, 3 November 2015.
Schuck, Andreas R.T., and Claes H. de Vreese. 2006. Between Risk and
Opportunity: News Framing and its Effects on Public Support for EU Enlargement.
21 (1): 532. DOI 10.1177/0267323106060987
Schuck, Andreas R. T., and Claes H. de Vreese. 2009. Reversed Mobilization in
Referendum Campaigns: How Positive News Framing Can Mobilize the Skeptics. 14
(1): 4066. DOI 10.1177/1940161208326926
Sear, Laura. 2018. Egg-Streme Sports: Norway's Olympic Squad Accidentally
Orders 15,000 Eggs After Asking Google Translate How to Say 1,500 in Korean.
Daily Mail Online, 8 February 2018.
Semetko, HA, and PM Valkenburg. 2000. Framing European Politics: A Content
Analysis of Press and Television News. 50 (2): 93109. DOI 10.1111/j.1460-
Stachova, Olga. 2019. Are We Using Google Translate Responsibly? Hamilton
Spectator, 1 January 2019.
Stapleton, Paul. 2019. How Google Translate May Disrupt English Classes, and
Every Other Subject, in Hong Kong Schools. South China Morning Post, 16 January
Stochl, Jan, Peter Jones, and Tim Croudace. 2012. Mokken Scale Analysis of Mental
Health and Well-Being Questionnaire Item Responses: A Non-Parametric IRT
Method in Empirical Research for Applied Health Researchers.BMC Medical
Research Methodology 12: 74. DOI 10.1186/1471-2288-12-74
Tan, Aaron. 2018. Microsoft's Chinese-English Translation System Achieves Human
Parity. Computer Weekly, Accessed 5 September 2019,
Thomson Reuters. 2020. Westlaw. Thomson Reuters, Accessed 26 June 2019.
Titcomb, James. 2016. Google Translate Now Lets You Chat to Anyone in a Foreign
Language. Telegraph Online, 12 May 2016.
Toral, Antonio, Sheila Castilho, Ke Hu, and Andy Way. 2018. Attaining the
Unattainable? Reassessing Claims of Human Parity in Neural Machine Translation.
In Proceedings of the Third Conference on Machine Translation (WMT), 113123.
Association for Computational Linguistics.
Tsukayama, Hayley. 2015. Google Translate Will Let You Read Signs, in Real
Time, With Your Phone Camera., 14 January 2015.
Turner, Karen. 2016. Google Translate Is About to Become Really Accurate. Daily
Herald, 4 October 2016.
Vieira, Lucas N. 2020. Automation Anxiety and Translators.Translation Studies
13(1):121. DOI 10.1080/14781700.2018.1543613
Vieira, Lucas N., and Elisa Alonso.2020. Translating Perceptions and Managing
Expectations: An Analysis of Management and Production Perspectives on Machine
Translation. Perspectives 28(2): 163184. DOI 10.1080/0907676X.2019.1646776
Vieira, Lucas N., Elisa Alonso, and Lindsay Bywood. 2019. Introduction: Post-
editing in Practice Process, Product and Networks.” JosTrans. The Journal of
Specialised Translation 31: 213.
Williams, Lawrence. 2006. “Web‐Based Machine Translation as A Tool for
Promoting Electronic Literacy and Language Awareness. Foreign Language Annals
39 (4):565578.
Wooten, Adam. 2010. International Business: Machine Translation Can Bring
Blunders, Successes in International Business. Deseret Morning News, 19 November
Wooten, Adam. 2011. International Business: Google Translate Has Great Uses,
Disastrous Misuses. Deseret Morning News, 22 April 2011.
Wu, Yonghui, Mike Schuster, Zhifeng Chen, Quoc V. Le, Mohammad Norouzi,
Wolfgang Macherey, Maxim Krikun, Yuan Cao, Qin Gao, Klaus Macherey, Jeff
Klingner, Apurva Shah, Melvin Johnson, Xiaobing Liu, Łukasz Kaiser, Stephan
Gouws, Yoshikiyo Kato, Taku Kudo, Hideto Kazawa, Keith Stevens, George Kurian,
Nishant Patil, Wei Wang, Cliff Young, Jason Smith, Jason Riesa, Alex Rudnick, Oriol
Vinyals, Greg Corrado, Macduff Hughes, and Jeffrey Dean. 2016. Google's Neural
Machine Translation System: Bridging the Gap between Human and Machine
Translation. Accessed 3 March 2020,
Articles (count)
Advertiser (Australia)
Afternoon Voice (India)
Arizona Republic, The (Phoenix, AZ, USA)
Arkansas Democrat Gazette (Little Rock, AR, USA)
Australian Financial Review (Australia)
Baltimore Sun (MD, USA)
Bangkok Post (Thailand)
Belfast Telegraph Online (UK)
Boston Business Journal (USA)
Boston Globe (MA, USA)
Bradford West Gwillimbury Topic (Canada)
Business and Financial Times (Ghana)
Cambodia News Gazette (Cambodia)
Cape Argus Weekend (South Africa)
Cape Community Newspapers (South Africa)
Chicago Sun Times (IL, USA)
China Daily (China)
China Knowledge Press (China)
Chosun Ilbo (South Korea)
Christian Science Monitor (USA)
City AM (UK)
Columbus Telegram (NE, USA)
Corkman (Ireland)
Courier News (Bridgewater, NJ, USA)
Daily American, The (Somerset, PA, USA)
Daily Dot (USA)
Daily Herald (Arlington Heights, IL, USA)
Daily Home, The (Talladega, AL, USA)
Daily Mail (UK)
Daily Mail Online (UK)
Daily News Egypt (Egypt)
Daily Pak Banker (Pakistan)
Daily Pakistan Today (Pakistan)
Daily Post (North Wales Edition) (UK)
Daily Press (USA)
Daily Record (Rochester, NY, USA)
Daily Star Online (UK)
Daily Telegraph (UK)
Dallas Morning News, The (USA)
Derby Evening Telegraph (UK)
Deseret Morning News (USA)
Economic Journal Insight (Hong Kong)
Economic Times (India)
European Voice (Belgium)
Evening Standard Online (London, UK)
Evening Times (Glasgow, Scotland, UK)
Express (UK)
Express Online (UK)
Express Tribune, The (Pakistan)
Financial Express (India)
Forward, The (New York, NY, USA)
Globe and Mail (Canada)
Globes Online (Israel)
Guardian (UK)
11 (UK)
Hamilton Spectator (Canada)
Hindu (India)
Hindustan Times (India)
Honolulu Star-Advertiser (Honolulu, HI, USA)
Houston Chronicle (USA)
Huffington Post (USA)
i Newspaper (UK)
Idaho Press-Tribune (Nampa, ID, USA)
Independent (UK)
Independent Online (UK)
Indian Express (India)
Industrial Worker (USA)
International Herald Tribune (USA)
International New York Times (USA)
Iran Daily (Iran)
Iraq News Gazette (Iraq)
Irish Independent (Ireland)
Irish Times (Ireland)
Israel National News (Israel)
Japan News, The (Japan)
Jarrow and Hebburn Gazette (UK)
Journal Gazette (USA)
Kitchener Record (Canada)
Korea Times (South Korea)
Lanka Puvath (Sri Lanka)
Latvia National News Agency (LETA) (Latvia)
Leader, The (UK)
Los Angeles Times (USA)
Mail on Sunday (UK)
Maryborough Herald (Australia)
Metro - New York (New York City, NY) (USA)
Mint (India)
Mirror (UK)
Mirror Online (UK)
Mizzima (Burma)
Montreal Gazette (Canada)
Nation (Kenya)
National Post (Canada)
National, The (Scotland, UK)
New Zealand Herald (New Zealand)
News Point (India)
Nikkei Weekly (Japan)
Oregonian (Portland, OR) (USA)
Ottawa Citizen (Canada)
Peninsula, The (Qatar)
Phnom Penh Post (Cambodia)
Richmond Times Dispatch (VA) (USA)
Salisbury Journal (UK)
San Francisco Chronicle (CA) (USA)
Saudi Gazette, The (Saudi Arabia)
Scotland on Sunday (Scotland, UK)
Scotsman (Scotland, UK)
Shenzhen Daily (China)
Silicon Valley Business Journal (USA)
South China Morning Post (China)
South China Morning Post (Contributor Content)
South China Morning Post Online (China)
Star-Ledger, The (USA)
Star Tribune: Newspaper of the Twin Cities (USA)
Straits Times (Singapore)
Sun Sentinel, Fort Lauderdale, FL (USA)
Sunday Business Post (Ireland)
Sydney Morning Herald, The (Australia)
Telegraph Online (UK)
Telegraph, The (India)
Telegraph, The (Nashua, NH) (USA)
Times of Central Asia, The (Kyrgyzstan)
Times of India (India)
Times of India (Online Edition) (India)
UB Post, The (Mongolia)
USA Today (USA)
Vancouver Province (Canada)
Vancouver Sun (Canada)
Victoria Times Colonist (Canada)
Washington Business Journal (USA)
Washington Post, The (USA)
1 (USA)
Western Mail (UK)
Western Star (Australia)
WND (WorldNetDaily) (USA)
York Daily Record (PA) (KRT) (USA)
Author’s Address
Lucas Nunes Vieira
17 Woodland Road
Bristol, BS8 1TE
United Kingdom
... Machine translation and language technologies based on the development of artificial intelligence have altered societal perceptions and attitudes about multilingual communication. Such technologies are now daily used in personal circumstances and in high-risk settings in order to immediately overcome linguistic gaps at basically no cost [1]. However, today, little is known about end users' attitudes towards machine translation despite its enormous impact on cross-cultural communication, where members of society find themselves in various situations with machine translation as the main means of communication [2]. ...
... However, today, little is known about end users' attitudes towards machine translation despite its enormous impact on cross-cultural communication, where members of society find themselves in various situations with machine translation as the main means of communication [2]. In general, machine translation technology is expected to closer approximate human capabilities as artificial intelligence and machine learning advance [1,3]. The accuracy, quality, and efficiency of machine-translated texts of specific sorts can already be fairly good in particular language combinations. ...
... However, not every user is aware of the errors and risks related to this technology when it comes to everyday usage; according to research, the quality of today's machine translation systems is insufficient for supporting intercultural conversation [10]. Many authors pay attention to the serious risks that machine translation errors pose in the field of health care, legal services, culture and media as well as many areas of social life [1,11,12]. The misuse of machine translation can have particularly serious consequences in high-stakes settings [1]. ...
Full-text available
Artificial intelligence-grounded machine translation has fundamentally changed public awareness and attitudes towards multilingual communication. In some language pairs, the accuracy, quality and efficiency of machine-translated texts of certain types can be quite high. Hence, the end-user acceptability and reliance on machine-translated content could be justified. However, machine translation in small and/or low-resource languages might yield significantly lower quality, which in turn may lead to potentially negative consequences and risks if machine translation is used in high-risk contexts without awareness of the drawbacks, critical assessment and modifications to the raw output. The current study, which is part of a more extensive project focusing on the societal impact of machine translation, is aimed at revealing the attitudes towards usability and quality as perceived from the end-user perspective. The research questions addressed revolve around the machine translation types used, purposes of using machine translation, perceived quality of the generated output, and actions taken to improve the quality by users with various backgrounds. The research findings rely on a survey of the population (N = 402) conducted in 2021 in Lithuania. The study reveals the frequent use of machine translation for a diversity of purposes. The most common uses include work, research and studies, and household environments. A higher level of education correlates with user dissatisfaction with the generated quality and actions taken to improve it. The findings also reveal that age correlates with the use of machine translation. Sustainable measures to reduce machine translation related risks have to be established based on the perceptions of different social groups in different societies and cultures.
... In this era of scientific and technological innovation, language and translation technologies have become more mature with the advancement of artificial intelligence. Users can obtain translations and overcome language barriers easily and quickly at no cost using free online translation tools, such as Google Translate and DeepL, for daily communication purposes [1,2]. In the translation industry, technology has shown increasing importance [3]. ...
... Another issue is about MT translations in high-risk conditions, where meanings are highly crucial, such as corresponding in crisis situations. Errors of MT output can result in reputational and professional costs for MT users [1]. A previous study has found that the comprehensibility of a translation is closely associated with greater trust in crisis communications [34]. ...
Full-text available
In this era of globalisation, translation technologies have become more popular in daily communication, the education sector, and the translation industry. It is observed that there is a prevalent use of machine translation (MT) among translation learners. The proper use versus abuse of MT can be a critical issue regarding its role in and impact on translation teaching. This exploratory study aims at investigating learners’ and instructors’ knowledge of MT, experience in MT use, perceived MT quality, ethics of MT use, and the perceived relationship between MT and translation training, in order to figure out the usefulness of MT in translation competence acquisition and the necessity of MT training. To this end, we conducted surveys and semi-structured interviews and found that the influence of MT in translation competence acquisition is determined by the properties of MT and learners’ quality. MT is particularly helpful in gaining lexical knowledge and knowledge to ensure translation efficiency, but not in bicultural knowledge. However, such usefulness builds on learners’ language proficiency, analytic ability, and learning motivation. In light of the findings, issues including the sustainability of MT from ethical and linguistic perspectives, and the potential and proper use of MT to inform translator training, are discussed.
This article intends to contribute to the current debate on the quality of neural machine translation (NMT) vs. (professional) human translation quality, where recently claims concerning (super)human performance of NMT systems have emerged. The article will critically analyse some current machine translation (MT) quality evaluation methodologies employed in studies claiming such performance of their MT systems. This analysis aims to identify areas where these methodologies are potentially biased in favour of MT and hence may overvalue MT performance while undervaluing human translation performance. Then, the article provides some Translation Studies informed suggestions for improving or debiasing these methodologies in order to arrive at a more balanced picture of MT vs. (professional) human translation quality.
Full-text available
The Routledge Handbook of Translation and Methodology provides a comprehensive overview of methodologies in translation studies, including both well-established and more recent approaches. The Handbook is organised into three sections, the first of which covers methodological issues in the two main paradigms to have emerged from within translation studies, namely skopos theory and descriptive translation studies. The second section covers multidisciplinary perspectives in research methodology and considers their application in translation research. The third section deals with practical and pragmatic methodological issues. Each chapter provides a summary of relevant research, a literature overview, critical issues and topics, recommendations for best practice, and some suggestions for further reading. Bringing together over 30 eminent international scholars from a wide range of disciplinary and geographical backgrounds, this Handbook is essential reading for all students and scholars involved in translation methodology and research.
Full-text available
Si vous avez déjà utilisé un logiciel de traduction automatique (en ligne), vous avez peut-être remarqué combien le système se perfectionne au fil du temps: de plus en plus rapide, précis, fluide... Un progrès technologique fulgurant et une facilité d’usage qui ne doivent toutefois pas masquer les – nombreuses – limites de ce recours à l’intelligence collective: textes en apparence fluides mais criblés d'erreurs grossières, entreprises à la réputation ruinée, appauvrissement de la langue, failles de confidentialité, etc. À travers cet état des lieux argumenté et documenté, je tente de montrer avec nuance et pédagogie au grand public et aux «consommateurs» de services de traduction pourquoi l’intelligence artificielle ne doit pas être considérée comme un substitut aux traducteurs et traductrices de chair et d’os et pourquoi tout recours «aveugle» à la machine doit être évité.
Full-text available
Despite the immense influence of machine translation (MT) on cross-cultural communication worldwide, little is known about end users’ predispositions toward MT. Our online experiment ( N = 284) compares people’s perceptions of MT and human translation in an ethically charged situation, in which the translation serves an immigrant worker in an interaction defined by power imbalance. Using hierarchical linear regression, we found that an otherwise identical translation was evaluated differently when it was attributed to MT or human translation. Results reveal that translators and non-translators alike exhibit a negative bias toward the MT product when asked to assess its accuracy and reliability, its ability to convey cultural and emotional otherness, and its potential effectiveness in helping the disadvantaged immigrant in need of the translation. We also demonstrate how lower evaluations of the MT product lead to a stronger wish to intervene in the translation by introducing changes to the original message. Our results suggest that predispositions toward MT must be taken into account in any consideration of MT-mediated communication, as these predispositions may shape the communicative act itself.
Full-text available
The use of machine translation (MT) in professional translation tasks can change not only how translators work, but also how projects are managed and the expectations they entail across translation supply chains. Previous research has looked extensively into translators’ attitudes to MT but has often ignored important aspects of how translators’ views interact with those of other language industry stakeholders. This article presents a contrastive analysis of attitudes to MT which covers management and production perspectives. The discussion draws on semi-structured interviews which were thematically coded and qualitatively examined. The study shows how MT adds uncertainty to translation production networks. It argues that the challenges posed by MT are exacerbated by how the current makeup of the language industry restricts translators’ field of influence to texts while possibly alienating them from wider aspects of business strategy. The article makes two suggestions. First, it calls for increased translator involvement in the management aspects of service provision. Second, it emphasises the need for a deeper discussion of MT which, rather than framing the technology itself as a potential ‘threat’, addresses broader societal issues involving misguided perceptions and mismatched expectations.
Full-text available
The potential benefits of integrating machine translation into human translation workflows are now widely recognised. In many sectors of the translation industry, translators’ throughput is improved with the use of machine translation as a tool in the translating process. Post-editing of machine translation is also a service in its own right, with specific guidelines and, more recently, an international standard. We introduce this special issue by providing a brief overview of post-editing as a practice, service and research topic. Contributions to the issue are then presented. The issue moves from a magnified perspective of translators’ work to important aspects of translation products, translators’ attitudes and translator training. It has four themes: the post-editing process; reception of post-edited products; attitudes and perceptions; and competence, training and education.
Full-text available
Current research on translation technology seeks to integrate physical, cognitive and organisational ergonomics, and uses insights from the situated cognition paradigm to bring together social and technical perspectives on fast-evolving human-computer interactions. Even though these trends imply that a wider variety of professional contexts should be considered, studies of institutional translation are still scarce. This paper reports on a three-week research stay in the French language department of the European Commission (DGT-Fr2), aimed at understanding current uses and perceptions of machine translation (MT) and post-editing within Europe's biggest translation institution. Based on ethnographic data, we established a survey that we tested among French translators before translating it into English and submitting it to all DGT translators. Our quantitative data include 89 respondents from 15 language departments. We perform multiple linear regressions to assess technology acceptance, before focusing on the variance that the model leaves unexplained. Our findings show that perceptions of control, subjective norm and image, as well as insecurity (fear of MT) have an impact on professional MT acceptance.
Full-text available
Translation is currently described as a profession under pressure from automation, falling prices and globalized competition. Translators' stance on machine translation (MT) is famously negative, but the economic dimension of this positioning is scarcely researched and often unclear. This article provides an analysis of translators' blog and forum postings contextualized within general trends in employment, the economy and work automation. The analysis concentrates on MT and pay. Two key findings are reported. First, MT was found to be a secondary issue in translators' comments on pay; most grievances were based on business practices themselves. Second, most criticisms of MT were rooted not in fears of being outperformed by MT systems, but rather in the technology's limitations and market consequences. This article calls for a broadening of translators' role across areas of specialization and argues that, in the debate on translation's future, MT cannot be decoupled from its economic effects.
Full-text available
Machine translation has made rapid advances in recent years. Millions of people are using it today in online translation systems and mobile applications in order to communicate across language barriers. The question naturally arises whether such systems can approach or achieve parity with human translations. In this paper, we first address the problem of how to define and accurately measure human parity in translation. We then describe Microsoft's machine translation system and measure the quality of its translations on the widely used WMT 2017 news translation task from Chinese to English. We find that our latest neural machine translation system has reached a new state-of-the-art, and that the translation quality is at human parity when compared to professional human translations. We also find that it significantly exceeds the quality of crowd-sourced non-professional translations.
Conference Paper
Full-text available
We analyse posts on social media (Facebook, LinkedIn, and Twitter) as a means to understand how translators feel about machine translation (MT). A quantitative analysis of more than 13,000 tweets shows that negative perceptions outweigh positive ones by a ratio of 3:1 overall, and 5:1 in tweets relating MT to human translation. Our study indicates a disconnect between translation and research communities, and we outline three suggestions to bridge this gap: (i) identifying and reporting patterns rather than isolated errors, (ii) participating in evaluation campaigns, and (iii) engaging in cross-disciplinary discourse. Rather than pointing out each other's deficiencies, we call for computer scientists, translation scholars, and professional translators to advance translation technology by acting in concert.
In the global research community, English has become the main language of scholarly publishing in many disciplines. At the same time, online machine translation systems have become increasingly easy to access and use. Is this a researcher’s match made in heaven, or the road to publication perdition? Here Lynne Bowker and Jairo Buitrago Ciro introduce the concept of machine translation literacy, a new kind of literacy for scholars and librarians in the digital age. For scholars, they explain how machine translation works, how it is (or could be) used for scholarly communication, and how both native and non-native English-speakers can write in a translation-friendly way in order to harness its potential. Native English speakers can continue to write in English, but expand the global reach of their research by making it easier for their peers around the world to access and understand their works, while non-native English speakers can write in their mother tongues, but leverage machine translation technology to help them produce draft publications in English. For academic librarians, the authors provide a framework for supporting researchers in all disciplines as they grapple with producing translation-friendly texts and using machine translation for scholarly communication—a form of support that will only become more important as campuses become increasingly international and as universities continue to strive to excel on the global stage.