126 127
Slovenščina 2.0, 2019 (1)
LANGUAGE IN THE AGE OF DATAISM
Š p e l a  V I N T A R
University of Ljubljana, Faculty of Arts, Ljubljana
Vintar, Š. (2019). Language in the Age of Dataism. Slovenščina 2.0, 7 (1): 126–143.
DOI: http://dx.doi.org/10.4312/slo2.0.2019.1.126-143.
The digital age brings dramatic changes to language and communication; its 
effects can be seen in the ways we use language, the channels we use to commu-
nicate and the manners in which ideas are spread. From the other end of the 
spectrum, our linguistic behaviour, communications and knowledge are trans-
formed into data which can be used or bought to feed intelligent technologies. 
The article presents a bird‘s eye view of these dynamics of change, first by focus-
ing on the impact of digitisation on language itself, further by analysing current 
trends in the language industry where traditional services are being replaced 
by technology- and data-driven solutions, and finally by exploring the impact 
of these technologies on people and society at large. We make a case for digital 
linguistics as an interdisciplinary field of study which adopts a human-centred 
approach to the sociolinguistic, technological, economic, infrastructural and 
ethical issues emerging with regard to language in the digital age.
Key words: digitisation, language change, language industry, digital linguistics
1  I N T R O D U C T I O N 
For some time now, the effects of digitisation on humanity no longer inspire 
just awe in the face of technological advances but increasingly raise concerns. 
In less than thirty years of its existence the internet has evolved from a medium 
charged with tremendous potential for freedom of communication, thought 
and global cooperation into its shadowy reverse – an environment which has 
become indispensable but obscured by infringements of privacy, security, dig-
nity, intellectual property rights and competition laws. As Pasquale observes 
in The Black Box Society, the “democratization” promised by Web 2.0 had “a 
127
Š. VINTAR: Language in the Age of Dataism
different – even an opposite effect. The very power that brought clarity and 
cooperation to the chaotic online world also spawns murketing, unfair compe-
tition, and kaleidoscopic distortions of reality” (2015, p. 98).
With rapid advances in Artificial Intelligence, similar concerns are arising in 
view of the many scenarios where machine learning algorithms are already re-
placing human decision-making. The fundamental questions are not whether 
certain jobs will disappear, in which work environments will humans be re-
placed with robots and when this is likely to happen for most fields of human 
endeavour. A more complex set of questions refers to issues such as AI bias 
(“Is the machine fair?”) and the moral status of AI (“Is it good or evil?”). In the 
Cambridge Handbook of Artificial Intelligence Bostrom and Yudkowsky lay 
the foundations for an ethics of AI, acknowledging that “[t]he term ‘Artificial 
Intelligence’ refers to a vast design space, presumably much larger than the 
space of human minds (since all humans share a common brain architecture)” 
(2014, p. 332) and that certain criteria which apply to humans performing 
social functions must also be considered in an algorithm intended to replace 
human judgement: responsibility, transparency, auditability, incorruptibility, 
predictability (2014, p. 318). 
The beginnings of the age of Big Data celebrated a technological milestone: a 
point in time when the computational and storage capacity on the one hand 
and the availability of digital data on the other would no longer present a bot-
tleneck for development. But consider the difference between data collected 
as sample of human activity in order to build better models, and collective 
data gathered through recording all human activity in order to be used, sold 
and resold by techno-giants and governments alike – this transition marks the 
beginnings of dataism, which, by Harari’s definition (2016, p. 351), declares 
that “the universe consists of data flows, and the value of any phenomenon or 
entity is determined by its contribution to data processing.”
It is against this background that we reflect on language in the digital age, 
whereby our focus shifts from language as a communicative device, language 
as an economic or business activity to language used as data. It will become 
clear that from all of these three aspects language has undergone profound 
changes under the influence of technology, and some of these changes may 
clearly be regarded as positive. In fact, while popular media will have us be-
128 129
Slovenščina 2.0, 2019 (1)
lieve that the future of everything is rather bleak, language in the digital age 
is, in many respects, thriving.
2 D I G I T I S A T I O N A N D L A N G U A G E C H A N G E
Languages change over time, and the factors involved in this process range 
from social, political, technological and economic influences to interventions 
by normative bodies. It is therefore only to be expected that digitisation and 
the emergence of numerous new channels of communication would have an 
impact on language use, and this is often reflected in news articles with titles 
such as “Is the Innanet RUINING teh English Language??? ¯\(°_o)/¯”1 or 
“L3t‘s t@lk internet”2. Linguists have been alert to this topic since the ear-
ly days of texting (Crystal, 2008), and the expansion and diversification of 
digital media has given rise to numerous studies exploring their effects on 
language as a whole or on the use of written language by youth (Baron, 2008; 
Lenhart, 2008; Thurlow, 2007; Crystal, 2011). In his comprehensive and de-
tailed review of the field of computer-mediated communication (CMC), An-
droutsopoulos (2011) provides a sociolinguistic set of conditions which shape 
“digital networked writing”, defining it as “vernacular”, “interpersonal and 
relationship-focused”, “unplanned and spontaneous” and “dialogical and in-
teraction-oriented”. In a critical synthesis of research studies spanning over 
three decades, Androutsopoulos demonstrates that much of the language 
change ascribed to digital media is restricted to lexis, with notorious lists of 
CMC-typical acronyms and other lexical innovations from the field of technol-
ogy. The effects of the internet on spoken language seem to be negligible, but 
the productivity of neologisms derived from social media seems boundless 
across (written) genres and in languages other than English. 
As for netspeak ruining school writing and negatively influencing literacy, 
evidence is less conclusive, and it is clear that such studies are methodolog-
ically difficult to conduct. Lenhart (2008) reports on a large scale study of 
the attitudes and habits of US teens comparing their out-of-school written 
communication and school writing, and the prevailing opinion of teens was 
1 https://gizmodo.com/is-the-innanet-ruining-teh-english-language-_o-1680686542, 22. 1. 2015
2 https://www.deccanherald.com/sunday-herald/sh-top-stories/l3ts-tlk-internet-668377.
html, 6. 5. 2018
129
Š. VINTAR: Language in the Age of Dataism
that texting and communicating via digital media was not writing, and that 
electronic communication had little or no impact on their written production 
at school. Similarly, Androutsopoulos (2011) mentions an empirical study by 
Dürscheid and Wagner (2010) carried out in German-speaking Swiss schools, 
where results suggest that out-of-school digital writing does not visibly influ-
ence institutional language production. 
This is not to say that the entire landscape of language use has not dramat-
ically changed, mainly through the emergence of new digital genres, and an 
“unprecedented scale of publicness” that tweets, blogs, posts, news comments 
and user reviews can achieve. The internet is a mixture of editorial, profes-
sionally-crafted content intertwined with vernacular, spontaneous, informal 
texts; a “manifestation of the intermingling of the private and the public that 
characterises late modernity” (Androutsopoulos, 2011, p. 10). We might add 
that the private/public is only one of the dimensions along which internet dis-
course is intermingled, other candidate variables being standard/non-stand-
ard, true/fake, predominantly textual/predominantly visual, monolingual/
multilingual, human-written/machine-written, and many more.
In recent years, a number of language resources, tools and methods have been 
developed which allow researchers to ask not just whether internet language 
is different, but how different it is. Such studies attempt to quantify the degree 
to which a certain language variety deviates from standard language, where-
by basic corpus pre-processing steps such as lemmatization and PoS-tagging 
need to be fundamentally adapted or even developed anew to accommodate 
the transformations and innovations found across genres of the web. In an 
interesting study of tweets in three closely related languages of former Yu-
goslavia, Serbian, Croatian and Slovene, Miličević et al. (2017) performed a 
thorough investigation of spelling transformations and report on a number of 
similarities and differences. In all three languages, frequent transformations 
include the omission of diacritics, repetition of certain vowels for emphasis 
and omission or transformation of word-final vowels or suffixes. In general, 
the transformation frequency is highest in Slovene (17%) and lowest in Ser-
bian (10%), with Croatian in the middle (13%), and if the omissions of dia-
critics are not counted Slovene drops to 15% and Serbian to just over 3% of 
transformed tokens. This difference is significant – it means that in an average 
130 131
Slovenščina 2.0, 2019 (1)
Slovene tweet between 4 and 5 words will be spelled in a non-standard way, 
while in Serbian only one or none. It would appear, at least for these three 
languages, that the tendency of a language towards the use of non-standard 
forms correlates with the level of digital maturity of its country,3 which is an 
unexpected finding. 
On the other hand, the authors of the study observe that transformations in 
Serbian, while lower in frequency, occur at more varied positions and indi-
cate a more playful and creative use of language than Slovene or Croatian. 
On the whole, twitterese and other types of internet discourse mirror layers 
and layers of social, cultural, political, economic and historical circumstances, 
and therefore any study of computer-mediated communication limiting itself 
to just linguistic features necessarily remains incomplete. More importantly, 
in the same way that virtual communities are communities with their own 
sociological features, cyberlanguage is a language form in its own right whose 
properties cannot be described in terms of deviation or transformation from 
its standard or spoken relatives.
Digitisation affects language beyond the scope of netspeak and genres pre-
dominantly residing on the internet. Today, texts are created with the aid of 
AI technologies and although these are trained on large samples of human 
language, neural networks may have given rise to a new set of dialects. We are 
referring mostly to machine translation and the various levels of post-editing 
applied before such texts are made public. As shown by recent surveys of the 
language industry which we present in more detail in the next section, the 
use of MT is growing in all strands of professional translation, but few studies 
have systematically analysed the properties of post-edited texts. A recent pa-
per by Toral (2019) fills this gap by addressing the question whether human 
translation and post-edited machine translation differ significantly in terms of 
several quantifyable features: lexical variety, lexical density, length ratio and 
part-of-speech sequences. 
The underlying intuition is that translations produced by humans from ma-
chine-translated drafts must be somehow different from translations pro-
duced by humans from scratch, and Toral performs a number of experiments 
3 For Slovenia and Croatia, see the Digital Economy and Society Index 2019, for Serbia 
the I-DESI 2018.
131
Š. VINTAR: Language in the Age of Dataism
across six language pairs verifying the existence of post-editese. As his results 
show, post-edited texts have lower lexical variety and density than human 
translations, and their sentence length and PoS sequences are closer to the 
source than the target language. This is in line with the so-called “transla-
tion universals”, the properties of translations which appear across language 
pairs and include phenomena such as normalization, shining-through and 
source language interference (Baker, 1993; Mauranen & Kujamäki, 2004). 
Toral‘s experiment thus proves two important things: firstly, that MT has a 
lower percentage of content words than HT and is therefore lexically simpler, 
and secondly that humans striving to improve on MT and create a human-like 
translation fail to do so, at least as far as lexical variety and PoS sequences are 
concerned. The author concludes with a cautionary note that “the extensive 
use of PE rather than HT may have serious implications for the target lan-
guage in the long term, for example that it becomes impoverished” (2019, p. 
279). It remains interesting though that – as Toral himself and several other 
authors point out (Green, 2013; Bowker & Buitrago Ciro, 2015) – humans do 
not necessarily perceive HT as better or more acceptable than PE. A recent 
study by Screen (2019) compares the quality of human and post-edited trans-
lations from the end-user perspective. The experiment uses both eye-tracking 
and end-user assessments of readability and comprehensibility, and the re-
sults show no statistically significant difference or inferiority of post-edited 
texts. 
3 A I  A N D T H E L A N G U A G E I N D U S T R Y
In the previous section we briefly discussed some instances where digitisation 
has an impact on language itself, both within and beyond the scope of internet 
communication. We now turn our attention to the economic sector of lan-
guage-related services generally referred to as the language industry, which 
traditionally revolved around translation and interpreting but is increasing-
ly diversified and, as we shall see, datafied. The importance of aggregating 
translation data became apparent with the emergence of Translation Memory 
tools, commonly known as Computer-Aided Translation or simply CAT tools 
from the early 1990s. With growing needs for fast translation and localization 
in the globalized world the idea that past translation projects should be stored 
in bilingual segments and recycled in order to boost productivity seemed per-
132 133
Slovenščina 2.0, 2019 (1)
fectly logical. However, the reactions of translators to CAT tools were reserved 
at best, with much opposition to the notion that translation work could be 
conceived as being repetitive and recyclable. 
As with most novelties, the technology gradually became mainstream and is 
considered indispensable today – according to the latest 2018 Language In-
dustry Survey (LIS 2019, p. 17) less than 1% of language services companies 
report that they are not using CAT tools. An interesting historical trivium is 
that as early as 1997 Trados Translator’s Workbench, the predecessor of to-
day’s market-leading SDL Trados Studio product suite, boasted the use of 
neural networks for their fuzzy matching algorithms, thus anticipating the AI 
era in translation technologies.
The development of statistical MT engines and their growing accessibility 
brought about another shift, namely that of MT becoming a pre-processing step 
in professional translation, thereby generating the demand for post-editing. De-
spite the fact that numerous studies have demonstrated significant productivity 
gains even with early SMTs (O’Brian, 2007; Guerberof, 2008) the sentiments of 
practicing translators towards PE remain mixed to this day, as a recent survey 
by the American Translators Association shows (Zetzsche, 2019). The sentiment 
however is not shared by language service providers. According to the results of 
the Language Industry Survey for 2016, 2017 and 2018, the use of MT is grow-
ing steadily both by companies and individuals. The latest survey, which is con-
sidered representative for Europe but not the rest of the world, states that the 
number of companies and individuals who are not using MT at all has dropped 
to 31% and 38%, respectively (LIS, 2016, 2017, 2018). 
With the arrival of neural Machine Translation (NMT), the language industry 
was transported into the age of AI. Even if several respondents of the afore-
mentioned ATA survey on “(Why) Do you use MT?” answered “To get a good 
laugh”, numerous studies have been performed to prove that NMT systems 
generally outperform SMT models by two or more BLEU points (Bentivogli 
et al., 2016; Way 2018), whereby several authors warn that BLEU may be 
under-reporting the difference in quality. According to error analyses, NMT 
produces fewer morphological errors (–19%), lexical errors (–17%), and sub-
stantially fewer word order errors (–50%) than its closest statistical competi-
tor, and on average requires about a quarter fewer edits compared to the best 
133
Š. VINTAR: Language in the Age of Dataism
phrase-based SMT (Way, 2018). 
It is thus not surprising that the report issued after the annual TAUS Global 
Content Conference (TAUS, 2019), an event which attracted 130 world’s larg-
est players in translation and localization, begins with a chapter titled The 
Quantum Leap and proclaims that “the NMT revolution of the last few years 
has pretty much wiped out all previous technologies. In addition to this, MT 
post-editing (PE) has become mainstream, currently the most widely used set-
up is MT in conjunction with some degree of human PE.” The size of the Ma-
chine Translation market was estimated at 433 million USD in 20164 and was 
expected to grow at an annual rate of 19%. Google Translate’s daily through-
put exceeds the volume that all translators in the world translate in a year. 
According to some estimates, MT is expected to reach the point of human 
parity by 2029, but on the other hand the language industry has voiced sev-
eral concerns regarding the use of NMT in business solutions. The first has to 
do with the robustness of NMT when dealing with different types of content 
and different domains. This clearly presents a challenge for language service 
providers, as varying levels of MT quality may have an impact on productivity, 
return-on-investment and the payment schemes used for PE. A second chal-
lenge is the sentence-based mode of processing for most NMT systems which 
may result in incoherent and inconsistent translations. Research is being con-
ducted on paragraph- or document-level NMT which would allow systems to 
translate content, not isolated sentences (TAUS 2019, p. 5–6). 
Comparing reports about the language industry from Europe, such as the LIS 
(2016, 2017 and 2018), and those from more globally oriented organisations 
such as TAUS (Massardo et al., 2016; Keynotes Summer, 2019) or GALA5, it 
appears that the global or US-based view of the language industry anticipates 
more dramatic changes driven by technology and envisages translation as a 
utility available to everyone, everywhere and on every device. All reviewed 
studies however agree in forecasting a rapidly growing demand for transla-
tions and other language services, in fact, these demands even today quite 
significantly surpass the capacities of human language service providers.  
4 Global Market Insights: Machine Translation Market Size, https://www.gminsights.
com/industry-analysis/machine-translation-market-size
5 http://www.gala-global.org
134 135
Slovenščina 2.0, 2019 (1)
One obvious consequence of this fact is that the majority of translations reach 
their audiences as raw MT, and that even in professional translation varying 
levels of quality are required. Both of these facts are hard to digest for a typical 
professional translator who was trained to strive towards a single and uni-
versal highest quality standard, and the position of most translator training 
institutes regarding quality remains unchanged.
4 T H E D A T A F I C A T I O N O F T R A N S L A T I O N
There is another important trend we can discern from the reports, and it con-
cerns data. Translation memories and bilingual corpora have been considered 
important assets for some time now, and issues related to ownership, data 
protection and intellectual property rights have been a hot topic of debate for 
over a decade (Smith, 2009). The Language Industry Survey for 2017 (LIS, 
2018) introduced for the first time a question about the transfer of user rights 
or ownership to the client, and responses indicated that approximately half of 
the respondents would never transfer those rights, while the other half would 
do so sometimes. The results for 2018 reveal a strong trend towards this 
transfer, and a breakdown of responses by company size shows that for larger 
companies the transfer of user rights or ownership is now almost mandatory. 
Large companies work for large clients, and these adhere to the dataism motto 
that data is the new fuel.
Another TAUS publication titled The Translation Industry in 2022 (Massardo 
& van der Meer, 2017) identifies Data as one of the six drivers of change and 
contains a valuable explanation of the difference between language data and 
translation data. While the former consists of translation memories, corpora, 
lexicographical and terminological collections, the latter is essentially meta-
data (2017, p. 18):
Translation data is typically metadata: data about translation that can 
be harvested downstream the closure of a translation project/job/task, 
such as content type, language pair(s), domain, subject, number of char-
acters/words/lines, quote/price, scheduled time, time spent, technol-
ogies used, translation stats (e.g. source - translation memory match, 
automatically propagated, machine translated - edited, approved) date 
and time of last saving, etc. 
135
Š. VINTAR: Language in the Age of Dataism
The analysis of translation data can provide a very valuable insight into 
the translation processes to find the best resource for a job, to decide 
what to translate and which technology to use for which content.
Eavesdropping on the debates amongst the tech giants such as Amazon, Apple, 
Google, Microsoft, Adobe, and the largest LSPs such as Lionbridge, SDL and 
TransPerfect, the power of data and the central role of AI remain recurring top-
ics. Language data markets have been established, but a lot of data collection 
goes on backstage using home-grown solutions. Machine translation is but the 
most obvious application fuelled by data; there is much demand for other in-
telligent services such as speech processing, user profiling, sentiment analysis, 
question answering, social network analytics, and there is a pronounced trend 
towards machine learning for a better management of multilingual workflows. 
In view of these developments it would appear that language as a business, 
not unlike other technology-driven businesses, is under threat of monopoli-
sation by the big players who simultaneously own the bulk of the data, devel-
op the smartest technologies and increasingly own research infrastructures 
significantly more powerful than those provided by the academia or public 
research funding.
5 A C A S E F O R D I G I T A L L I N G U I S T I C S
We have examined some of the challenges that language is facing in the digital 
age; it is now time to reflect on the possible measures to be taken by research-
ers, academia, practitioners and policy makers in order not to be reduced to 
mere instruments of change but assume an active role, and possibly direct 
the course of development into one which is fairer and more inclusive for all 
members of society. 
The advances that Artificial Intelligence is enabling in natural language pro-
cessing are truly impressive, and scientific progress is accelerated by the enor-
mous amount of private funding flowing into research and by e.g. Google’s 
policy6 to openly share some of its AI tools with the community, thus enhanc-
ing competition. Clearly though, it will be increasingly hard for researchers to 
keep up with the speed of discoveries produced by the techno-giants.
6 https://ai.google/tools/
136 137
Slovenščina 2.0, 2019 (1)
It is important to remember that the role of science in these – or any other – 
times is not to blindly compete in the race towards singularity, but to provide 
critical insights, analyse impact, advocate responsibility, and safeguard the 
ethical principles fundamental to our society. With regard to the ethics of AI, 
strong initiatives are underway within leading research institutes, such as the 
Future of Humanity Institute7, the IEEE8 or the Foundation for Responsible 
Robotics9, and the European Commission has recently passed a communi-
cation titled Building Trust in Human Centric Artificial Intelligence, which 
defines AI “not as an end in itself, but a tool that has to serve people with the 
ultimate aim of increasing human well-being” (EC COM, 2019, p. 168). 
Returning to language and AI, ethical concerns regarding the use of human 
translations to train MT systems have been raised by Kenny (2011), especially 
because “the role of translators in creating vital data has been mostly down-
played or ignored” by MT developers. She also touches upon another interest-
ing ethical question, namely the (im)possibility of computers communicating 
like humans. According to Melby and Warner (1995) and Kenny (ibid.), “in 
order to communicate with others, we must have agency, which involves the 
capacity to make real choices for which we take responsibility, and we must 
also regard our interlocutors as having agency. […] Without agency, we are 
reduced to the status of machines and there is no dynamic general language.” 
It is needless to point out that from today’s perspective, with chatbots and 
automated dialogue systems lurking around every corner of the internet, the 
ethics of communication seems a considerably more complex issue. 
A more recent contribution to the discussion about language resources and 
the ethics of their reuse was made by Moorkens et al. (2016) who systemati-
cally describe the practices prevalent in the language industry regarding data 
ownership, the “disempowered” translator in precarious working positions 
and the legal situation “in which laws of copyright are effectively bypassed in 
content collection, curation, and exploitation, [and which] permits resource 
holders to retain data at a cost to disempowered human writers and transla-
tors”. The authors’ recommendations for translators include collective bar-
7 https://www.fhi.ox.ac.uk/
8 https://ethicsinaction.ieee.org/
9 https://responsiblerobotics.org/
137
Š. VINTAR: Language in the Age of Dataism
gaining, informing themselves about their legal rights and using TM metadata 
more effectively in order to explicitly assign usage rights to their assets.  
Establishing fair practices for data sharing and a transparent regulative system 
for its collection and processing is just one of the challenges we need to face 
up to, and the present situation gives little grounds for optimism. As Pasquale 
writes, “top legal scholars have already analogized the power relationships in 
virtual worlds and cloud computing to medieval feudalism” (2015, p. 218). 
Considering all the other profound changes that language and communication 
are undergoing in the digital society, some of which we have discussed above, 
it becomes clear that to understand and adequately describe these phenome-
na an interdisciplinary approach is required, and that linguistics alone, even 
with all its applied subfields, lacks the methodological inventory to approach 
this task. Analysing large communication networks, proposing new workflows 
of content creation, developing intelligent knowledge solutions or modelling 
emotions, to name but a few non-futuristic scenarios, all require a combina-
tion and integration of knowledge from different domains. 
If solutions for the processing of natural language were traditionally developed by 
computational linguists, we are now entering an era in which AI technologies are 
becoming mainstream in many areas of everyday life, and we may well imagine 
the not-so-distant future when these now separate intelligences begin interacting 
to solve complex problems, much like intelligent humans do. As we have demon-
strated before, any intelligent technology imposed on the human society has a 
social dimension in that it modifies the social practices that were in place before, 
and it may also have ethical, legal, psychological and other dimensions.
We thus propose the term digital linguistics to designate a human-centred 
approach to digitally-driven language and communication as well as the study 
thereof, utilizing methodologies and theoretical backgrounds from a range of 
“feeder” disciplines: linguistics, including computational, corpus, cognitive, 
socio- and psycholinguistics; computer and information science, including 
machine learning, data mining, knowledge modelling and AI; social sciences, 
including law, journalism, communication and media studies; and the rele-
vant humanities, in particular ethics, psychology and philosophy. The list is 
not exhaustive and serves primarily to emphasize the interdisciplinary nature 
of digital linguistics. 
138 139
Slovenščina 2.0, 2019 (1)
We further believe it is paramount that universities and other higher education 
institutions respond not only to the skills gap reported by employers, but more 
importantly to the expectations and concerns of the civil society which already 
feels insecure in the “feudalism” of digital communication channels. One attempt 
to bridge this education gap is the joint master‘s degree in Digital Linguistics in 
preparation by a consortium of three universities, Ljubljana, Zagreb and Brno, ex-
pected to launch in 2021/2022. The model curriculum was developed within the 
recently concluded DigiLing10 project and is based on the findings of a trans-Eu-
ropean survey of language-related needs amongst employers.11 
6 C O N C L U S I O N S
It seems that digitisation affects language in ways different from what the aver-
age person or even linguist might expect. The examples selected for discussion 
above show that the language of internet communications develops under its 
own rules, not dissimilar to other language varieties known from pre-internet 
times. Contrary to urban myth, teenagers do know how to draw the bounda-
ry between formal and informal writing, while adults or even language pro-
fessionals have a hard time distinguishing between human and post-edited 
translations and do not have a clear preference for either. Machine-translated 
and post-edited texts are found increasingly acceptable by end-users despite 
the fact that they exhibit pronounced features of the source language.
Word embeddings and neural networks allow us to discern semantic change 
(Hamilton et al., 2016) or translate between languages for which no parallel 
data exist (Johnson et al., 2017), but at the same time language professionals 
feel disempowered as their intellectual property rights are being ignored in 
the global data collection frenzy. In this article we attempted to present a se-
lection of recent trends involving language and communication in the digital 
age, and their implications may range from fantastic to catastrophic, depend-
ing on one’s point of view. A concluding thought might be that as academics 
and researchers we should strive towards objectivity and realism in the face 
of the complex challenges, but also towards a responsible stance and a keen 
interest in the dynamics of change, the only constant of our times.
10 https://www.digiling.eu
11 https://www.digiling.eu/deliverables
139
Š. VINTAR: Language in the Age of Dataism
A C K N O W L E D G M E N T
This article was presented as keynote lecture at the INFuture 2019: Knowl-
edge in the Digital Age conference and is republished here with the permission 
of the INFuture programme board. 
R E F E R E N C E S
Androutsopoulos, J. (2011). Language change and digital media: a review of 
conceptions and evidence. In N. Coupland & T. Kristiansen (Eds.), Stand-
ard languages and language standards in a changing Europe (pp. 145–
161). Oslo: Novus Press.
Baker, M. (1993). Corpus linguistics and translation studies: Implications and 
applications. In M. Baker, G. Francis & E. Tognini-Bonelli (Eds.), Text and 
technology: In honour of John Sinclair (pp. 233–250). Amsterdam: John 
Benjamins. 
Baron, Naomi S. (2008). Always on: Language in an Online and Mobile 
World. Oxford: Oxford University Press.
Building Trust in Human Centric Artificial Intelligence. EC COM(2019) 168. 
European Commission, 8. 4. 2019. 
Bentivogli, L., Bisazza, A., Cettolo, M., & Federico, M. (2016). Neural versus 
phrase-based machine translation quality: a case study. In Proceedings of 
EMNLP 2016.
Bostrom, N. & Yudkowsky, E. (2014). The ethics of artificial intelligence. In 
W. Ramsey & K. Frankish (Eds.), The Cambridge handbook of artificial 
intelligence (pp. 316–334). Cambridge: Cambridge University Press.
Bowker, L. & Buitrago Ciro, J. (2015). Investigating the usefulness of machine 
translation for newcomers at the public library. Translation and Inter-
preting Studies, 10(2), 165–186.
Hamilton, W.L., Leskovec, J. & Jurafsky, D. (2016). Diachronic word embed-
dings reveal statistical laws of semantic change. In Proceedings of the As-
sociation for Computational Linguistics (ACL), Berlin.
Crystal, D. (2008). Txtng: The gr8 db8. Oxford: Oxford University Press.
Crystal, D. (2011). Internet Linguistics. London: Routledge.
140 141
Slovenščina 2.0, 2019 (1)
Dürscheid, C., Wagner, F. & Brommer, S. (2010). Wie Jugendliche schreiben: 
Schreibkompetenz und neue Medien. Berlin & New York: de Gruyter.
Green, S., Heer, J. & Manning, C. D. (2013). The efficacy of human post-edit-
ing for language translation. In Proceedings of the SIGCHI Conference on 
Human Factors in Computing Systems (pp. 439–448).
Goel, R., Soni, S., Goyal, N., Paparrizos, J., Wallach, H., Diaz, F. & Eisenstein, J. 
(2016). The social dynamics of language change in online networks. In In-
ternational Conference on Social Informatics (pp. 41–57). Cham: Springer.
Guerberof, A. (2009). Productivity and quality in MT post-editing. In MT 
Summit XII-Workshop: Beyond Translation Memories: New Tools for 
Translators MT. August 29, 2009. Ottawa, ON, 8ff.
Harari, Y. N. (2015). Homo Deus: a Brief History of Tomorrow. London: 
Harvill Secker. 
Johnson, M., Schuster, M., Quoc V. Le, Krikun, M., Wu, Y., Chen, Z. et al. 
(2017). Google’s multilingual neural machine translation system: Ena-
bling zero-shot translation. Transactions of the Association for Compu-
tational Linguistics 5, 339–351.
Kenny, D. (2011). The ethics of machine translation. In Proceedings of the 
New Zealand Society of Translators and Interpreters Annual Conference 
2011. Auckland, New Zealand. 
Lenhart, A. (2008). Writing, technology, and teens. Washington, DC: Pew Internet 
and American Life Project.  Retrieved from http://pewresearch.org/pubs/808/writ-
ing-technology-and-teens
[LIS 2016] 2016 Language Industry Survey – Expectations and Concerns of 
the European Language Industry. Retrieved from https://www.euatc.org/in-
dustry-surveys/item/download/5_57a02b9c45602ea9f7daf4440a7b2979
[LIS2017] 2017 Language Industry Survey – Expectations and Concerns of the 
European Language Industry. Retrieved from https://ec.europa.eu/info/sites/
info/files/2017_language_industry_survey_report_en.pdf
[LIS2018] 2018 Language Industry Survey – Expectations and Concerns of 
the European Language Industry. Retrieved from https://ec.europa.eu/info/
sites/info/files/ 2018_language_industry_survey_report_en.pdf 
Massardo, I.; van der Meer, J. (2017). The translation industry in 2022. TAUS 
BV, De Rijp, The Netherlands.
141
Š. VINTAR: Language in the Age of Dataism
Massardo, I., van der Meer, J., Khalilov, M. (2016). TAUS Translation Tech-
nology Landscape Report. September 2016, TAUS BV, De Rijp, The Neth-
erlands.
Mauranen, A., Kujamäki, P. (2004, Eds.). Translation universals: do they ex-
ist?  Amsterdam: John Benjamins.
Melby, A. K., Warner, C. T. (1995). The possibility of language: a discussion 
of the nature of language, with implications for human and machine 
translation. Amsterdam & Philadelphia: John Benjamins Publishing.
Miličević, M., Ljubešić, N. & Fišer, D. (2017). Birds of a feather don’t quite 
tweet together. In D. Fišer & M. Beißwenger (Eds.), Investigating Com-
puter-Mediated Communication: Corpus-Based Approaches to Lan-
guage in the Digital World (pp. 14–43). Ljubljana: Faculty of Arts.
Moorkens, J., Lewis, D., Reijers, W., Vanmassenhove, E. & Way, A. (2016). 
Translation resources and translator disempowerment. In Tenth Interna-
tional Conference on Language Resources and Evaluation (LREC 2016), 
Portorož, Slovenia.
O’Brien, S. (2007). An Empirical Investigation of Temporal and Technical 
Post-Editing Effort. Translation and Interpreting Studies, 2(1), 83–136.
Pasquale, F. (2015) The black box society. Cambridge, MA: Harvard Univer-
sity Press.
Screen, B. (2019). What effect does post-editing have on the translation prod-
uct from an end-user’s perspective?. Journal of specialised translation, 
31, 133–157.
Smith, R. (2009). Copyright issues in translation memory ownership. ASLIB 
Translating and the Computer 31.
TAUS Keynotes Summer 2019. A Review of the TAUS Global Content Confer-
ence in Salt Lake City, UT (USA). TAUS Signature Editions, Amsterdam. 
Retrieved from www.taus.net
Toral, A. (2019). Post-editese: an exacerbated translationese. Proceedings of 
the Machine Translation Summit XVII Volume 1: Research track (pp. 
273–281). Dublin, Ireland: EAMT. 
Thurlow, C. (2007). Fabricating youth: new-media discourse and the technol-
ogization of young people. S. Johnson & A. Ensslin (Eds.), Language in 
the Media (pp. 213–233). London: Continuum.
142 143
Slovenščina 2.0, 2019 (1)
Way A. (2018). Quality Expectations of Machine Translation. In J. Moorkens, 
S. Castilho, F. Gaspari, S. Doherty (Eds.), Translation Quality Assess-
ment: From Principles to Practice (pp. 159–178). Cham: Springer. doi: 
https://doi.org/10.1007/978-3-319-91241-7_8
Zetzschke, J. (2019). (How) Do You Use MT? The Tool Box Journal, 19-11-
306(2019). Retrieved from http://www.internationalwriters.com/toolkit/
current.html#LETTER.BLOCK5
143
Š. VINTAR: Language in the Age of Dataism
JEZIK V ČASU DATAIZMA
Digitalizacija vnaša korenite spremembe v jezik in komunikacijo, saj vpliva 
na načine izražanja, sredstva komunikacije in poti, prek katerih se širijo nove 
ideje. Obenem živimo v času, ko se naše jezikovno vedenje, naša sporočila in 
znanje skrbno beležijo, ti podatki pa se uporabljajo in prodajajo za urjenje pa-
metnih tehnologij. V prispevku skušamo zaobjeti dinamiko teh sprememb s 
širše perspektive, in sicer se najprej osredotočimo na vpliv digitalizacije na sam 
jezik, nato analiziramo sodobne težnje v jezikovni industriji, kjer opažamo, da 
mnoge tradicionalne jezikovne storitve nadomeščajo tehnološko podprte in na 
podatkih temelječe rešitve, nazadnje pa obravnavamo vpliv tehnologij na člove-
ka in družbo kot celoto. Iz širšega okvirja razprave izpeljemo utemeljitev in 
opredelitev področja digitalnega jezikoslovja kot interdisciplinarne vede, ki se 
z jezikom v digitalni dobi ukvarja s humanistično-družboslovnega izhodišča in 
vanj vključuje jezikoslovne, tehnološke, družbenoekonomske, infrastrukturne, 
kognitivne, etične in pravne vidike. 
Ključne besede: digitalizacija, jezikovne spremembe, jezikovna industrija, digi-
talno jezikoslovje
To delo je ponujeno pod licenco Creative Commons: Priznanje avtorstva-Deljenje pod enakimi 
pogoji 4.0 Mednarodna. / This work is licensed under the Creative Commons Attribution-Share-
Alike 4.0 International.
https://creativecommons.org/licenses/by-sa/4.0/