Chap e 4
Teaching ansla ion s uden s abou da a
in he age o gene a i e AI
Lynne Bowke
Uni e si é La al, Canada
AI ansla ion ools a e now a key pa o ansla ion educa ion, bu many edu-
ca o s a e sea ching o e ec i e ways o each he essen ials o hese ools o
s uden s wi h no backg ound in compu e science. This chap e explains why co -
po a make a good en y poin o lea ning abou AI ansla ion ools, and i explo es
how science communica ion echniques such as aming, analogies and isualiza-
ion can be used o help ansla ion educa o s and s uden s come o g ips wi h da a
and machine lea ning.
1 In oduc ion
Though ansla o educa ion o en akes place in an a s o humani ies acul y,
echnology has been a i al pa o he ansla ion p o ession, and hence o ans-
la o educa ion p og ammes, o a leas hi y yea s. O e his pe iod, ansla o s
and ansla ion s uden s ha e shown ema kable esilience as hey adap o new
ools and new echnology-based ways o wo king. Howe e , his does no mean
ha eaching echnologies o ansla ion s uden s is easy, and educa ing he ed-
uca o s can be a pa icula challenge (Bowke 2023, Kenny 2020). As he pace o
new ool eleases ge s as e , i can be di icul o ansla o educa o s o know
whe e o begin. This chap e p oposes ha eaching ansla ion s uden s abou
da a is one o he key building blocks in p epa ing hem o use ansla ion ech-
nology e ec i ely. In he case o ansla ion, da a equen ly akes he o m o
ex s o ganised in o co po a. This chap e he e o e begins wi h a b ie e iew o
co po a and co pus-based ools, no ing how AI ansla ion ools ha e in luenced
Lynne Bowke . 2026. Teaching ansla ion s uden s abou da a in he age o gene a-
i e AI. in JC Pene , Joss Moo kens & Masa u Yamada (eds.), Teaching ansla ion in
he age o gene a i e AI: New pa adigm, new lea ning?, 67–85. Be lin: Language Sci-
ence P ess. DOI: 10.5281/zenodo.17641070
Lynne Bowke
he na u e o co po a. Nex , he chap e ou lines why a science communica ion
app oach, a he han adi ional scien i ic communica ion, could be use ul in he
con ex o ansla ion echnology educa ion. Th ee science communica ion ech-
niques – aming, analogies and isualisa ion – a e combined wi h examples o
demons a e how science communica ion could be adap ed o each ansla ion
s uden s abou da a- ela ed opics in he age o AI.
2 Co po a and co pus-based ansla ion ools
2.1 The changing cha ac e is ics o co po a in he age o AI
O e he pas ew decades, ansla o s ha e seen he in oduc ion and in eg a ion
o a ange o di e en ools, including conco dance s (Zane in 2023), e m ex-
ac o s (Ko kon zelos & Ananiadou 2022), ansla ion memo y sys ems (Melby
& W igh 2023), machine ansla ion sys ems (Way 2020), and mos ecen ly,
gene a i e AI (GenAI) sys ems (Siu 2024). The a ious ools ha e become in-
c easingly sophis ica ed wi h ega d o hei capabili ies and hei unde lying
a chi ec u e, bu a common ea u e o all o he ools named abo e is ha hei
co e unc ionali y e ol es a ound p ocessing da a in he o m o ex s. Collec-
ions o ex s a e usually e e ed o as co po a, and hese can ake di e en
o ms depending on he na u e o he ex s and he way ha hey a e o ganised
(McEne y 2022). T ansla ion ools ha p ocess co po a a e equen ly desc ibed
as co pus-based o da a-d i en ools (Wang e al. 2022).
As poin ed ou by Isabelle e al. (1993: 205), “exis ing ansla ions con ain mo e
solu ions o mo e ansla ion p oblems han any o he a ailable esou ce”. The e-
o e, one ype o co pus ha has been used e y o en by ansla ion ools is
he bilingual pa allel co pus (Sima d 2020). In his ype o co pus, a collec ion
o sou ce ex s a e aligned – usually a sen ence le el – wi h hei coun e pa
a ge ex s. In o he wo ds, each sen ence in he sou ce ex is linked o i s co -
esponding ansla ion in he a ge ex . Some ansla ion ools migh also use
monolingual co po a o o iginal ex s in he sou ce and/o a ge language o ac
as a linguis ic model o ha language.
Fo s uden s who ha e al eady lea ned abou conco dance s, e m ex ac o s
o ansla ion memo y sys ems, he no ion o a co pus is al eady amilia because
he co pus is a e y isible esou ce in such ools. When using hese ools, he
ansla o o ansla ion s uden o en has a hand in c ea ing he co pus o may
need o upload he co pus ha has been p o ided by he clien o educa o . Owing
o hei amilia i y, co po a make a good s a ing poin o lea ning abou neu al
machine ansla ion o GenAI ools.
68
4 Teaching ansla ion s uden s abou da a in he age o gene a i e AI
Acco ding o Bowke & Pea son (2002: 9),
A co pus can be desc ibed as a la ge collec ion o au hen ic ex s ha
ha e been ga he ed in elec onic o m acco ding o a speci ic se o c i-
e ia. The e a e ou impo an cha ac e is ics o no e he e: ‘au hen ic’,
‘elec onic’, ‘la ge’ and ‘speci ic c i e ia’.
While he gene al no ion o a co pus emained ela i ely s able in he pe iod
be o e AI ansla ion ools appea ed, he in oduc ion o hese ools has ushe ed
in some changes wi h ega d o he ea u es o co po a. The e o e, i is impo an
o ansla ion s uden s o unde s and how AI ools ha e in luenced and al e ed
he na u e o co po a.
2.2 Machine- eadable o m
The need o he co pus o be in elec onic o machine- eadable o m has no
changed. Indeed, we could say i is mo e impo an han e e since AI ans-
la ion ools ake on an e en g ea e deg ee o co pus p ocessing han do ools
such as conco dance s o ansla ion memo y sys ems. While conco dance s and
ansla ion memo y sys ems conduc pa e n ma ching and hen so and display
in o ma ion o he ool use s o in e p e , AI ansla ion ools go u he by a -
emp ing o in e p e he esul s and p esen ully o med ansla ion solu ions.
2.3 Size
Co po a a e used o e eal linguis ic pa e ns, which only become appa en when
he e a e mul iple examples o a gi en linguis ic phenomenon. The e o e, an-
o he ea u e o co po a is ha hey a e usually e y la ge collec ions o ex .
Howe e , ou unde s anding o wha cons i u es “la ge” has e ol ed o e ime.
The i s gene a ion o co po a c ea ed in he 1960s con ained hund eds o hou-
sands o wo ds and we e mainly consul ed by linguis s (McEne y 2022). These
linguis s used co pus analysis ools (e.g. conco dance s) o help hem so and
display he ex da a, bu he linguis s we e s ill esponsible o in e p e ing i .
Today, he co po a used o powe AI ools con ain hund eds o billions o wo ds
(Hughes 2023). This is in la ge pa because hese AI ools do no unde s and ex
in he way ha people do, and hey need a much la ge numbe o examples in
o de o p edic pa e ns wi h con idence. Howe e , his need o ex emely la ge
co po a is in luencing o he cha ac e is ics o co po a, such as he ex s selec ed
o inclusion (see Sec ion 2.4) and some imes e en he au hen ici y o he ex s
(see Sec ion 2.5).
69
Lynne Bowke
The a ailabili y o machine- eadable ex s can di e d ama ically om one lan-
guage o ano he , c ea ing dispa i ies wi h ega d o an AI ool’s pe o mance
in di e en languages. A high- esou ce si ua ion occu s when i is ela i ely
s aigh o wa d o ga he a la ge numbe o high-quali y esou ces o a gi en
language o language pai . Fo ins ance, bo h English and F ench a e widely used
languages, and he e is a lo o ansla ion ac i i y be ween hese wo languages.
As a esul , i is no oo di icul o compile monolingual and bilingual pa al-
lel co po a o hese languages, and hey a e hus e e ed o as high- esou ce
languages. In con as , a low- esou ce si ua ion can occu when languages (o
language a ie ies) a e less widely used, such as some o he Indigenous lan-
guages o he Ame icas o he a ie y o F ench used in Canada. Fo languages
o language a ie ies o limi ed di usion, i can be mo e challenging o build a
la ge co pus. Mo eo e , e en i wo languages ha e a la ge numbe o speake s
(e.g. Russian and Hindi), he e may no be a lo o ansla ion ac i i y be ween
hem, making i ha d o c ea e bilingual pa allel co po a o his language pai .
The e o e, languages, language a ie ies o language pai s o which he e a e
ew co po a a ailable a e desc ibed as being low esou ce.
2.4 Speci ic c i e ia
As emphasised by McEne y (2022), in o de o be mos use ul, a co pus canno
consis o ex s ha ha e been ga he ed a andom o in a pu ely oppo unis ic
way. Ra he , he ex s in a co pus a e selec ed because hey co espond o speci ic
c i e ia and a e ep esen a i e o a la ge se o ex s wi h hose cha ac e is ics.
One clea c i e ion in he con ex o ansla ion is ha he ex s should be o
high quali y. Beyond his, he e a e many di e en op ions o designing a co pus
depending on i s in ended pu pose, bu he key poin he e is ha he choice o
which ex s o include is mo i a ed. Fo example, in he con ex o a co pus o
be used o ansla ion, i could be impo an o selec ex s ha a e on a gi en
opic, o a ce ain ex ype o egis e , o om a pa icula ime pe iod.
E idence o he impo ance o co pus design can be seen in he way ha
ansla o s cons uc ansla ion memo y da abases, such as by c ea ing di e -
en da abases (o adding ele an me ada a) o di e en domains o o di e en
clien s (e.g. o espec hei p e e ed e minology o house s yle). In his way,
hey can es ic a sea ch o ex s ha ha e speci ic ea u es (Melby & W igh
2023). Likewise, neu al machine ansla ion ools a e known o achie e be e
quali y when he co pus is adap ed o a speci ic domain (Chu & Wang 2018).
Howe e , as no ed in sec ion Sec ion 2.3, AI ansla ion ools need o ha e an
eno mous numbe o ex s in he co pus. As a esul , i can be challenging o
70
4 Teaching ansla ion s uden s abou da a in he age o gene a i e AI
achie e he necessa y size while ying o be selec i e abou he con en . The
low- esou ce si ua ion desc ibed in sec ion Sec ion 2.3 can be u he compli-
ca ed when i comes o inding ce ain ex ypes o ex s on e y specialised
opics in less widely used o ansla ed languages and language a ie ies. A con-
sequence o no ha ing a la ge enough high-quali y co pus is ha he AI ool
does no ha e enough good examples o d aw on, and so he ool’s pe o mance
may be poo e in low- esou ce si ua ions (Way 2025). I lowe quali y ex s a e
included in he co pus in o de o inc ease he size, hen he ool may gene a e
low-quali y ansla ions. The implica ions o insu icien quan i y and quali y o
ex s in aining co po a used by AI ansla ion ools is discussed in mo e de ail
in sec ions Sec ion 4.2 and Sec ion 4.3.
2.5 Au hen ici y
Finally, he need o ha e au hen ic ex s in he co pus used o be sac osanc .
Fo ins ance, McEne y (2022) desc ibes a co pus as “a la ge body o linguis ic
e idence composed o a es ed language use” (494) and “a collec ion o na u ally
occu ing language da a” (495). In he case o ansla ion, he e is a desi e o ha e
high-quali y da a, which means using ex s ha ha e been ansla ed by language
p o essionals. As obse ed by Kenny (2011: 2), he eason ha he de elope s
o ansla ion ools use co po a o human ansla ions o ain hei sys ems
is because such co po a a e assumed o con ain good answe s o ansla ion
p oblems; and hey a e assumed o con ain good answe s p ecisely because hey
con ain ansla ions pe o med by human beings.
One consequence o he need o bigge and bigge bilingual pa allel co po a
o aining AI ansla ion ools has been ha such au hen ic high-quali y human
ansla ed ex s ha e become inc easingly aluable commodi ies. This in u n has
aised many e hical ques ions abou owne ship o ansla ion da a and pe mis-
sion o use i , p omp ing an explici need o discuss hese issues wi h ansla ion
s uden s. Moo kens (2022) con ains a de ailed examina ion o such e hical issues,
along wi h sugges ions o how hese can be in eg a ed in o ansla o educa ion
(see Sec ion 4.1).
Ano he esponse o he po en ial sho age o ex needed o cons uc ing
e y la ge co po a o use wi h AI ansla ion ools has been o se aside he long-
es ablished adi ion o using au hen ic da a and o explo e he use o syn he ic
da a. In he con ex o ansla ion, syn he ic da a is c ea ed by using a machine
ansla ion ool o ansla e addi ional ex s, and hen adding hese machine-
ansla ed ex s o he co pus (Senn ich e al. 2016). The quali y o syn he ic da a
can a y, and while he need o addi ional da a ends o be o low- esou ce
71
Lynne Bowke
languages, hese a e he languages o which machine ansla ion al eady unde -
pe o ms. The e o e, using a low-pe o mance ool o gene a e mo e ansla ions
in ha same language is likely o esul in poo quali y ex . I his poo quali y
ex is hen used o u he ain he ool, he e is a isk ha using syn he ic da a
could pe pe ua e a cycle o medioc i y.
This sec ion has elabo a ed why unde s anding co po a is impo an o ans-
la ion s uden s, and has ou lined some o he ways ha co po a a e changing in
he age o AI. The ollowing sec ion mo es beyond wha s uden s need o know
abou ansla ion da a (i.e., co po a) o conside how educa o s can communica e
his in o ma ion e ec i ely.
3 F om scien i ic communica ion o science
communica ion
3.1 Scien i ic communica ion
Scien i ic communica ion (some imes called in e nal science communica ion) is
a ype o expe - o-expe communica ion (Hanauska 2019). I occu s when one
subjec ield specialis add esses ano he , wi h bo h pa ies ha ing a deep knowl-
edge o he complex ma e in ques ion. Academics a e ypically e y com o able
wi h his ype o communica ion since hey spend much o hei ime discussing
hei esea ch wi h hei pee s. E en as pa o hei eaching, hey a e no mally
communica ing in o ma ion di ec ly om hei own ield, helping hei s uden s
o acqui e he necessa y expe ise o become specialis s in hei own igh . How-
e e , he a i al o AI ools – mo e han any o he echnology – has changed ha
si ua ion o many ansla ion educa o s.
As no ed in he in oduc ion, echnology is no new ei he o he ansla ion
p o ession o o ansla o educa ion p og ammes. Ye as poin ed ou by Kenny
(2020), when i comes o ansla ion echnology, iden i ying wha o each has
o en been easie han wo king ou how o each i . As he ield o ansla ion be-
came mo e echnologised, some educa o s admi ed o eeling challenged by he
demands o keeping up wi h echnology (e.g. Kenny 2007, Ma shman & Bowke
2012), bu his esponsibili y lay p ima ily wi h hose educa o s who had chosen
o specialise in echnology. Technology- ela ed eaching was o en es ic ed o
a co e cou se on ansla ion echnologies, whe e a la ge ocus was on lea ning
how o use he ools, which o e ed an a ay o sophis ica ed ea u es (Bowke
2023).
AI ansla ion ools a e di e en in se e al espec s. Fi s ly, he use in e ace
is compa a i ely simple – o en equi ing he use o do li le mo e han selec
72
4 Teaching ansla ion s uden s abou da a in he age o gene a i e AI
he sou ce and a ge languages and pas e o ype a ex . In some cases, he
AI ool may e en be called au oma ically by ano he ool. The e o e, he “how-
o” aspec o ansla ion echnology eaching is no longe a p incipal ocus. A
mo e impo an di e ence, howe e , is ha he echnology behind he in e ace
is a om simple. As Kenny (2018) poin s ou , olde gene a ions o echnology
we e ela i ely anspa en and comp ehensible in hei inne wo kings, while
AI ansla ion ools a e mo e opaque in ha he people using hem – and e en
he people de eloping hem – canno always unde s and how he ool a i es a
i s p oposals o ou pu . In he ield o AI mo e b oadly, his opaci y has led o
a push o mo e explainable AI (XAI) (Ridley 2025), while machine ansla ion
esea che s a e also beginning o wo k on explainable aspec s o his echnology
(e.g. Lank o d e al. 2023). In he mean ime, AI ansla ion ools a e now s a ing
o appea in cou ses ac oss he ansla ion educa ion cu iculum a he han
solely in co e echnology cou ses, and ansla o educa o s a e seeking ways
o help hemsel es and hei s uden s unde s and how hese ools can a ec
ansla ion p ocesses and p oduc s, and how o use he ools esponsibly.
In hese ci cums ances, scien i ic communica ion is no easible since ansla-
ion educa o s and s uden s a e no compu e scien is s. The echnology behind
neu al machine ansla ion and GenAI ools is highly sophis ica ed. The ools
may appea o be simple because he use in e ace is simple, bu behind he
scenes, he e a e e y complex a i icial neu al ne wo ks, wo d embeddings, ec-
o s, ans o me s, and mo e. I is no necessa y o mos ansla ion educa o s
o s uden s o unde s and all he de ails o how an a i icial neu al ne wo k ope -
a es in o de o use AI-based ansla ion ools.1Howe e , unde s anding he ole
o da a (i.e., co po a) in a da a-d i en AI ansla ion ool is aluable and can help
ansla ion s uden s o be mo e in o med and esponsible use s o his ype o
echnology. A he e y leas , unde s anding he ole o da a is a use ul i s s ep,
and he app oach p oposed in sec ion 4 is in ended as way o in oduce da a on
p og ammes aimed a educa ing s uden s who will wo k p ima ily as language
1This is cu en ly ue o many ansla ion-o ien ed jobs and ansla ion educa ion p o-
g ammes. Howe e , as B i a-Iglesias & O’B ien (2022) poin ou , new jobs a e eme ging ha
s addle he bounda y be ween ansla o and echnologis and ha equi e a deepe g ound-
ing in echnologies. To add ess his changing ma ke , new educa ion p og ammes a e also
eme ging ha place mo e emphasis on echnological skills, such as he MSc in T ansla ion
Technology a Dublin Ci y Uni e si y in I eland, and he MSc in Mul ilingual Digi al Commu-
nica ion a McGill Uni e si y in Canada. In hese mo e echnology-o ien ed p og ammes, a
deepe unde s anding o he unde lying a chi ec u es is bo h equi ed o he educa o s and
o e ed o he s uden s, and da a li e acy aining may also be deepe and mo e echnical, such
as he app oach desc ibed by Hackenbuchne & K üge (2023) as pa o he Da aLi MT p ojec .
73
Lynne Bowke
p o essionals a he han as echnology specialis s. Bu i scien i ic communica-
ion is no a easible app oach, can science communica ion help?
3.2 Science communica ion
When ansla ion s uden s hea people alking abou AI-based ansla ion ools
using e ms such as ans o me s o wo d embeddings, hey may eel in imida ed
and lack con idence in hei abili y o use hese ools e ec i ely. The e o e, when
in oducing hese ools, ansla ion educa o s may ind i help ul o d aw on
science communica ion echniques. As poin ed ou by Bu ns e al. (2003), science
communica ion is icky o de ine p ecisely, and he e a e my iad compe ing
and o e lapping desc ip ions o his concep in he li e a u e. Howe e , he e is
gene al ag eemen ha he essen ial goal is o make expe knowledge accessible
o non-expe s. This is why his ac i i y is some imes desc ibed mo e speci ically
as ex e nal science communica ion (Hanauska 2019). In he case o ansla ion
s uden s, while hey a e de eloping expe ise in ansla ion, hey a e no expe s
in AI o machine lea ning.
Academics in all disciplines a e inc easingly encou aged o engage in sci-
ence communica ion o sha e he esul s o hei esea ch wi h a wide audi-
ence, which could include policy make s, pa icipan s in esea ch s udies, und-
ing agencies and e en he gene al public (e.g. Na ional Academies o Sciences,
Enginee ing, and Medicine 2017). Ye while science communica ion is s ongly
encou aged, many academics ecei e li le aining in his a ea, e en hough i
is a complex and skilled ask (Bo owiec 2023). Fo una ely, ips, guides and ec-
ommenda ions a e eme ging o help ill his gap (e.g. Bo owiec 2023, Cooke e
al. 2017, Hen ille 2020). F om hese, we can glean se e al sugges ions ha can
be in eg a ed in o eaching o help explain da a- ela ed concep s o ansla ion
s uden s as a ounda ion o lea ning abou AI ansla ion ools.
F ame you message: F ame you message in e ms ha a e accessible, ela able
and meaning ul o you audience. F aming is no abou ma ke ing you
poin o iew bu abou inding a way o ac i ely engage he audience wi h
an issue by showing hem why hey should ca e abou i .
Make he abs ac conc e e and connec un amilia concep s o amilia ones: Use
me apho s o analogies ha connec a he le el o unde lying commonal-
i ies o help he audience unde s and an un amilia o abs ac concep in
ela ion o a amilia and conc e e one.
74
4 Teaching ansla ion s uden s abou da a in he age o gene a i e AI
Visualise he con en : Use isual aids (e.g. images, concep maps, g aphs) o sup-
po you message. Making he in o ma ion a ailable in di e en modes
o o ma s can help o c ys allise he ideas in he mind o you audience.
The ollowing sec ion conside s how hese h ee science communica ion ech-
niques could be implemen ed o help ansla ion educa o s each concep s ela ed
o da a.
4 Applying science communica ion echniques o each
ansla ion s uden s abou da a
4.1 F aming he message abou da a
I was no ed in sec ion 2 ha co po a can be a good en y poin o in oduc-
ing da a-d i en AI ools o ansla ion s uden s because hey ha e likely been
exposed o co po a when using o he ools (e.g. conco dance s, e m ex ac o s,
ansla ion memo y sys ems). Co po a a e less immedia ely isible in he mo e
au oma ed AI ansla ion ools, bu his makes i all he mo e impo an o ac-
i ely discuss hem. As Kenny (2011) emphasises, co pus-based ools canno unc-
ion wi hou co po a. I no co pus is p esen , he ool i sel has no hing o o e as
a ansla ion aid. This is a e y impo an poin o sha e wi h ansla ion s uden s
because i demons a es he essen ial con ibu ion o ansla o s. O cou se he
ool de elope s play a key ole in bo h p og amming he ools and de e mining
how co pus da a can be p ocessed e ec i ely, bu he da a i sel – he ansla ed
ex – is indispensable, and his ac is o en glossed o e when he ools a e de-
sc ibed o p omo ed. Ye as pa o ansla o educa ion, i is impo an o make
s uden s awa e o he alue o co po a because his helps hem o unde s and
hei own wo h as language p o essionals in addi ion o unde s anding how
co pus-based o da a-d i en ools wo k, and whe e he po en ial pi alls migh
lie.
As men ioned in sec ion Sec ion 2.5, ansla ion co po a ha e become aluable
esou ces, leading o some ques ionable p ac ices. F aming he discussion o AI
ansla ion ools as pa o a b oade issue o p o essional p ac ice, e hics, and
digi al ci izenship, a he han simply as an ins umen al means o making hei
job easie , can help ansla ion s uden s o ecognise no only he alue o hei
wo k, bu also he ac ha echnology does no exis in a acuum. Discussions
abou he ole o da a in da a-d i en ools can encou age ansla ion s uden s o
e lec on how AI ansla ion echnology can impac ool use s and end use s o
ansla ed p oduc s in a ious ways. I also empowe s language p o essionals o
75
Lynne Bowke
Bu ns, T. W., D. J. O’Conno & S. M. S ocklmaye . 2003. Science communica ion:
A con empo a y de ini ion. Public Unde s anding o Science 12(2). 183–202. DOI:
10.1177/09636625030122004.
Chu, Chenhui & Rui Wang. 2018. A su ey o domain adap a ion o neu al ma-
chine ansla ion. In P oceedings o he 27 h in e na ional con e ence on compu-
a ional linguis ics, 1304–1319. San a Fe, USA: Associa ion o Compu a ional
Linguis ics.
Cooke, S e en J., Aus in J. Gallaghe , Na alie M. Sopinka, Vi ian M. Nguyen,
Rachel A. Skubel, Neil Hamme schlag, Sa ah Boon, Na han Young & Andy J.
Danylchuk. 2017. Conside a ions o e ec i e science communica ion. Face s
2. 233–248. DOI: 10.1139/ ace s-2016-0055.
Eh ensbe ge -Dow, Mau een, Alice Delo me Beni es & Ca oline Leh . 2023. A
new ole o ansla o s and aine s: MT li e acy consul an s. The In e p e e
and T ansla o T aine 17(3). 393–411. DOI: 10.1080/1750399X.2023.2237328.
Flach, Pe e . 2012. Machine lea ning. Camb idge, UK: Camb idge Uni e si y P ess.
Hackenbuchne , Janiça & Ralph K üge . 2023. Da aLi MT – eaching da a li e acy
in he con ex o machine ansla ion li e acy. In Ma y Nu minen, Judi h B en-
ne , Maa i Koponen, Si ku La omaa, Mikhail Mikhailo , F ede ike Schie l,
Tha indu Ranasinghe, E a Vanmassenho e, Se gi Al a ez Vidal, No a A an-
be i, Ma a Nunzia ini, Ca la Pa a Esca ín, Mikel Fo cada, Maja Popo ic,
Ca olina Sca on & Helena Moníz (eds.), P oceedings o he 24 h annual con-
e ence o he Eu opean Associa ion o Machine T ansla ion, 285–293. Tampe e,
Finland: Eu opean Associa ion o Machine T ansla ion.
Hanauska, Monika. 2019. His o ical aspec s o ex e nal science communica ion.
In Anne e Leßmöllmann, Ma celo Dascal & Thomas Gloning (eds.), Science
communica ion, 585–600. Be lin, Ge many: De G uy e Mou on. DOI: 10.1515/
9783110255522-027.
Hen ille, Le i ia. 2020. E ec i e science communica ion decons uc ed. Uni e -
si y A ai s. h ps://uni e si ya ai s.ca/ca ee -ad ice/ask-d -edi o /e ec i e-
science-communica ion-decons uc ed/.
Hook, Daniel. 2023. The lone banana p oblem. Digi al Science. h ps: // www.
digi al- science.com/ ld /a icle/ he- lone- banana- p oblem- o - he- new-
p og amming-speaking-ai/.
Hughes, Alex. 2023. Cha GPT: e e y hing you need o know abou OpenAI’s GPT-4
ool. h ps://www.science ocus.com/ u u e- echnology/gp -3.
Isabelle, Pie e, Ma c Dyme man, Geo ge Fos e , Jean-Ma c Ju as, Ellio
Macklo i ch, F ançois Pe aul , Xiaobo Ren & Michel Sima d. 1993. T ansla-
ion analysis and ansla ion au oma ion. In P oceedings o he i h in e na-
82
4 Teaching ansla ion s uden s abou da a in he age o gene a i e AI
ional con e ence on heo e ical and me hodological issues in machine ansla ion
o na u al languages: Associa ion o compu a ional linguis ics, 201–217.
Kenny, Do o hy. 2007. T ansla ion memo ies and pa allel co po a: Challenges
o he ansla ion aine . In Do o hy Kenny & Kyongjoo Ryou (eds.),
Ac oss bounda ies: In e na ional pe spec i es on ansla ion s udies, 192–208.
Newcas le-Upon-Tyne, UK: Camb idge Schola s Publishing.
Kenny, Do o hy. 2011. The e hics o machine ansla ion. In Sybille Fe ne (ed.),
P oceedings o he 20 h New Zealand Socie y o T ansla o s and In e p e e s na-
ional con e ence. Auckland, New Zealand: New Zealand Socie y o T ansla o s
& In e p e e s.
Kenny, Do o hy. 2018. Machine ansla ion. In Pie s Rawling & Philip Wilson
(eds.), The Rou ledge handbook o ansla ion and philosophy, 428–445. London,
UK: Rou ledge.
Kenny, Do o hy. 2020. Technology and ansla o aining. In Minako O’Hagan
(ed.), The Rou ledge handbook o ansla ion and echnology, 498–515. London,
UK: Rou ledge.
Koponen, Maa i & Ma y Nu minen. 2024. Risk managemen o con en deli e y
ia aw machine ansla ion. In Ma ion Win e s, Sha on Deane-Cox & U sula
Böse (eds.), T ansla ion, in e p e a ion and echnological change, 111–135. Lon-
don, UK: Bloomsbu y.
Ko kon zelos, Ioannis & Sophia Ananiadou. 2022. Te m ex ac ion. In Ruslan
Mi ko (ed.), The Ox o d handbook o compu a ional linguis ics, 2nd edn., 991–
1012. Ox o d, UK: Ox o d Uni e si y P ess.
Lank o d, Séamus, Hai hem A li & Andy Way. 2023. adap NMT: An open-sou ce,
language-agnos ic de elopmen en i onmen o neu al machine ansla ion.
Language Resou ces and E alua ion 57. 1671–1696. DOI: 10.1007/s10579-023-
09671-2.
Ma shman, Elizabe h & Lynne Bowke . 2012. T ansla ion echnologies as seen
h ough he eyes o educa o s and s uden s: Ha monizing iews wi h he
help o a cen alized eaching and lea ning esou ce. In Sé e ine Hubsche -
Da idson & Michał Bo odo (eds.), Global ends in ansla o and in e p e e
aining, 69–95. London, UK: Con inuum.
McEne y, Tony. 2022. Co po a. In Ruslan Mi ko (ed.), The Ox o d handbook o
compu a ional linguis ics, 2nd edn., 494–507. Ox o d, UK: Ox o d Uni e si y
P ess.
Melby, Alan K. & Sue Ellen W igh . 2023. T ansla ion memo y. In Chan Sin-
wai (ed.), Rou ledge encyclopedia o ansla ion echnology, 2nd edn., 766–781.
London, UK: Rou ledge.
83
Lynne Bowke
Moo kens, Joss. 2022. E hics and machine ansla ion. In Do o hy Kenny (ed.),
Machine ansla ion o e e yone, 121–140. Be lin, Ge many: Language Science
P ess.
Na la, Akhila, B e Kup el, Ka i a Sa in, Robe o No oa & Jus in Ko. 2018. Au-
oma ed classi ica ion o skin lesions: F om pixels o p ac ice. Jou nal o In es-
iga i e De ma ology 138(10). 2108–2110. DOI: 10.1016/j.jid.2018.06.175.
Na ional Academies o Sciences, Enginee ing, and Medicine. 2017. Communica -
ing science e ec i ely: A esea ch agenda. Washing on DC, USA: The Na ional
Academies P ess. DOI: 10.17226/23674. h ps://nap.na ionalacademies.o g/
ca alog/23674/communica ing-science-e ec i ely-a- esea ch-agenda.
Ribei o, Ma co T., Samee Singh & Ca los Gues in. 2016. Why should I us you?
Explaining he p edic ions o any classi ie . In P oceedings o he 22nd ACM
SIGKDD In e na ional Con e ence on Knowledge Disco e y and Da a Mining:
Associa ion o compu ing machine y, 1135–1144. New Yo k, USA. DOI: 10.1145/
2939672.2939778.
Ridley, Michael. 2025. Human-cen e ed explainable a i icial in elligence: An an-
nual e iew o in o ma ion science and echnology (ARIST) pape . Jou nal
o he Associa ion o In o ma ion Science and Technology 76(1). 98–120. DOI:
10.1002/asi.24889.
Robe s, Michael, De ek D iggs, Ma hew Tho pe, Julian Gilbey, Michael Ye-
ung, S ephan U sp ung, Angelica I. A iles-Ri e o, Ch is ian E mann, Ca hal
McCague, Lucian Bee , Jona han R. Wei -McCall, Zhongzhao Teng, Aix-
Co ne E ossyni Gk ania-Klos as, James H. F. Rudd, E is Sala & Ca ola-
Bibiane Schönlieb. 2021. Common pi alls and ecommenda ions o using ma-
chine lea ning o de ec and p ognos ica e o COVID-19 using ches adio-
g aphs and CT scans. Na u e Machine In elligence 3. 199–217. DOI: 10.1038/
s42256-021-00307-0.
Senn ich, Rico, Ba y Haddow & Alexand a Bi ch. 2016. Imp o ing neu al ma-
chine ansla ion models wi h monolingual da a. In P oceedings o he 54 h
annual mee ing o he Associa ion o Compu a ional Linguis ics ( olume 1): As-
socia ion o compu a ional linguis ics, 86–96.
Sima d, Michel. 2020. Building and using pa allel ex o ansla ion. In Minako
O’Hagan (ed.), The Rou ledge handbook o ansla ion and echnology, 78–90.
London, UK: Rou ledge.
Siu, Sai Cheong. 2024. Re olu ionising ansla ion wi h AI: Un a elling neu-
al machine ansla ion and gene a i e p e- ained la ge language models. In
Yuhong Peng, Huihui Huang & De eng Li (eds.), New ad ances in ansla ion
echnology, 29–54. Singapo e: Sp inge . DOI: 10.1007/978-981-97-2958-6_3.
84
4 Teaching ansla ion s uden s abou da a in he age o gene a i e AI
Vanmassenho e, E a, Ch is ian Ha dmeie & Andy Way. 2018. Ge ing gende
igh in neu al machine ansla ion. In P oceedings o he 2018 Con e ence on
Empi ical Me hods in Na u al Language P ocessing, 3003–3008. B ussels, Bel-
gium: Associa ion o Compu a ional Linguis ics.
Wang, Hai eng, Hua Wu, Zhongjun He, Liang Huang & Kenne h W. Chu ch. 2022.
P og ess in machine ansla ion. Enginee ing 18. 143–153. DOI: 10.1016/j.eng.
2021.03.023.
Way, Andy. 2013. T adi ional and eme ging use-cases o machine ansla ion.
In P oceedings o T ansla ing and he Compu e 35. London, UK: ASLIB. h ps:
//aclan hology.o g/2013. c-1.12/.
Way, Andy. 2020. Machine ansla ion: Whe e a e we a oday? In E ik Angelone,
Mau een Eh ensbe ge -Dow & Ga y Massey (eds.), The Bloomsbu y companion
o language indus y s udies, 311–332. London, UK: Bloomsbu y.
Way, Andy. 2025. Wha does he u u e hold o ansla ion echnologies? In S e-
an Baumga en & Michael Tiebe (eds.), The Rou ledge handbook o ansla ion
echnology and socie y, 448–461. London, UK: Rou ledge.
Zane in, Fede ico. 2023. Conco dancing. In Chan Sin-wai (ed.), Rou ledge ency-
clopedia o ansla ion echnology, 2nd edn., 498–511. London, UK: Rou ledge.
85
