Posts Tagged ‘Google Translate’

Since I wrote my book of language-learning / teaching activities that involve the use of the learners’ own language (Kerr, 2014), one significant change has taken place. Some of these activities focused on machine translation tools, like Google Translate. The main concern at the time was the lack of reliability of these tools, and many teachers were strongly opposed to their students using them. It was easy to find examples of bad translation and to laugh at them. My favourite was an image of a crowd welcoming Pope Francis to Cuba, where a banner saying ‘Welcome Potato’ was supposedly a mistranslation of the Spanish ‘papa’, which can mean both ‘pope’ and ‘potato’. It’s a pity the image was Photoshopped.

My approach, feeling that it was impracticable and counter-productive to ban Google Translate altogether, was to exploit the poor quality of many of the translations as a way of training learners to use them more critically and more effectively. But, in the intervening years, the accuracy of online translation has much improved. One study (Aiken, 2019) found that Google Translate had improved by 34% over an 8-year period, although there were still significant differences in the accuracy of particular language pairings. Improvements will continue, and there are new services like DeepL Translator, which was launched in 2017, and, in my view, generally outperforms Google Translate, although fewer language pairings are available. 100% translation accuracy (if such a thing actually exists) may never be achievable, but for some kinds of texts with some language pairings, we are effectively there.

Training in using online translation is, however, still needed for some language pairings. There are two good ways of starting this.

1 Take a text in the learners’ L1 and machine-translate it into English. Highlight the errors and give it to the learners along with the original and a list of common error types (see below). The learners work together, looking at the highlighted errors and attempting to match them to one of the error types on the list.

2 Take a text in English and machine-translate it into the learners’ L1. The learners work together, first identifying and highlighting the errors they find, then comparing the translation with the original and attempting to identify the reasons for the error having happened.

At the time that I wrote this book, I would have advised against using Google Translate as a dictionary to look up single words, on the grounds that (1) the tool worked better the more context / co-text it had, and (2) there were usually better bilingual dictionaries available. My position has shifted somewhat, primarily because the features that Google Translate now offers have improved. There’s a video by Russell Stannard, called ‘Using Google Translate in Language Teaching -Tips and Ideas’,where Russell basically uses the software as a dictionary tool, and enthuses about the possibilities for pronunciation and listening work, for using the ‘favourites’ feature, and for exporting, via a spreadsheet, wordlists that are selected so they can be used with a spaced-repetition memory trainer.

You can find more ideas for using Google Translate as a pronunciation training tool in Minh Trang (2019).

One of the most common uses of machine translation by learners is undoubtedly in the production of written work. One recent piece of research (Tsai, 2019) came to the less than surprising conclusion that learners produced better drafts when doing so, and were happy to use it. Whether or not more learning took place when machine translation was used is another matter. O’Neill (2019) came to a similar conclusion, but found that students performed better with prior training. This training consisted of two 20-minutes sessions, where students tested the tool with examples before reviewing its strengths and weaknesses. More ideas for machine translation literacy training can be found in Bowker (2020).

I’d like to suggest a couple of further activities where Google Translate or DeepL can be used in the preparation of activities. In both cases, I’ll illustrate with the short original text from a newspaper (Der Standard) below:

Eine Passage in der neuen Covid-19-Verordnung erregt seit letzter Nacht besondere Aufmerksamkeit: das Alkoholverbot nach der Sperrstunde im Umfeld von Bars. Weil kein Ende definiert ist, sind manche in Sorge: Sind wir auf dem Weg in eine Prohibition? Konkret heißt es in der Novelle, die am Sonntag in Kraft tritt: „Nach der Sperrstunde dürfen im Umkreis von 50 Metern um Betriebsstätten der Gastgewerbe (sic!) keine alkoholischen Getränke konsumiert werden.“ Die Sperrstunde liegt in den meisten Lokalen bei 1.00 Uhr.

For the first activity, the students’ task is to translate this into English. Beforehand, translate the text using DeepL, and scramble the words, giving a copy of this scramble to the students.

1.00 am   50 meters   a   a   after   after   alcohol   alcoholic   amendment   are   are   around   attention   attracting   ban   bars   be   because   been   beverages   closing   come   consumed   Covid 19   curfew   curfew   defined   end   establishments   establishments   force   has   hospitality   in   in   in   into   is   is   last   may   most   new   night   no   no   of   of   on   on   on   one   passage   prohibition   radius   regulation   sic!   since   some   special   specifically   states   Sunday   the   the   the   the   the   the   time   to   vicinity   way   we   which   will   within   worried

The translation becomes a kind of jigsaw.

The second activity, only appropriate for more advanced learners, takes a text in the L1. Use two different translation tools to create separate translations, and correct any obvious errors (if there are any). Distribute these, along with the original to the students. Their task is, first, to identify and highlight any differences between the two versions. After that, they discuss each difference, saying which version they prefer (and why) or whether they have no preference.

Google Translate: One passage in the new Covid-19 regulation has been attracting special attention since last night: the ban on alcohol after the curfew in the vicinity of bars. Because no end is defined, some are concerned: are we on the way to prohibition? Specifically, the amendment, which comes into force on Sunday, says: “After the curfew, alcoholic beverages may not be consumed within 50 meters of the hospitality industry (sic!).” The curfew is at 1.00 a.m. in most restaurants.

Deepl: One passage in the new Covid 19 regulation has been attracting special attention since last night: the ban on alcohol after curfew in the vicinity of bars. Because no end is defined, some are worried: Are we on the way to a prohibition? Specifically, the amendment, which will come into force on Sunday, states: “After curfew, no alcoholic beverages may be consumed within a radius of 50 meters around hospitality establishments (sic!). The closing time is 1.00 am in most establishments.

One further activity that I would like to suggest makes use of the way that Google Translate translates each word as it goes, but amends previously translated words in the light of what follows. This is only suitable when Google Translate is accurate! The cleft example below (The thing that bothers me most is how long it will take) neatly illustrates the process. The following is a game-like exploitation. Project (or screen-share) Google Translate, set up to English and the learners’ own language. Tell the students that you are going to do a translation together. Tell them that the first word will be ‘the’, and ask them to predict how Google will translate it. Then, type in the word and everyone can see how Google translates it. Tell the students the next word (‘thing’) and again ask for their suggestions before typing it in. Carry on in the same way.



The thing

Die Sache

The thing that

Die Sache, die

The thing that bothers

Das, was stört

The thing that bothers me

Das, was mich stört

The thing that bothers me most

Das, was mich am mesiten stört

The thing that bothers me most is

Das, was mich am mesiten stört, ist

The thing that bothers me most is how

Was mich am meisten stört, ist wie

The thing that bothers me most is how long

Was mich am meisten stört, ist wie lange

The thing that bothers me most is how long it

Das, was mich am meisten stört, ist, wie lange es dauert

The thing that bothers me most is how long it will

Was mich am meisten stört, ist, wie lange es dauern wird

The thing that bothers me most is how long it will take.

Was mich am meisten stört, ist, wie lange es dauern wird.


Aiken, M. (2019). An Updated Evaluation of Google Translate Accuracy. Studies in Linguistics and Literature, 3 (3)

Bowker, L. (2020) Machine translation literacy instruction for international business students and business English instructors. Journal of Business & Finance Librarianship 25 (1):1-19

Kerr, P. (2014) Translation and Own-Language Activities. Cambridge: Cambridge University Press

Minh Trang, N. (2019) Using Google Translate as a Pronunciation Training Tool. LangLit, 5 (4), May 2019

O’Neill, E. M. (2019) Training students to use online translators and dictionaries: The impact on second language writing scores. International Journal of Research Studies in Language Learning, 8(2), 47-65

Tsai, S. (2019) Using google translate in EFL drafts: a preliminary investigation. Computer Assisted Language Learning, 32 (5-6): pp. 510–526.

The most widely-used and popular tool for language learners is the bilingual dictionary (Levy & Steel, 2015), and the first of its kind appeared about 4,000 years ago (2,000 years earlier than the first monolingual dictionaries), offering wordlists in Sumerian and Akkadian (Wheeler, 2013: 9 -11). Technology has come a long way since the clay tablets of the Bronze Age. Good online dictionaries now contain substantially more information (in particular audio recordings) than their print equivalents of a few decades ago. In addition, they are usually quicker and easier to use, more popular, and lead to retention rates that are comparable to, or better than, those achieved with print (Töpel, 2014). The future of dictionaries is likely to be digital, and paper dictionaries may well disappear before very long (Granger, 2012: 2).

English language learners are better served than learners of other languages, and the number of free, online bilingual dictionaries is now enormous. Speakers of less widely-spoken languages may still struggle to find a good quality service, but speakers of, for example, Polish (with approximately 40 million speakers, and a ranking of #33 in the list of the world’s most widely spoken languages) will find over twenty free, online dictionaries to choose from (Lew & Szarowska, 2017). Speakers of languages that are more widely spoken (Chinese, Spanish or Portuguese, for example) will usually find an even greater range. The choice can be bewildering and neither search engine results nor rankings from app stores can be relied on to suggest the product of the highest quality.

Language teachers are not always as enthusiastic about bilingual dictionaries as their learners. Folse (2004: 114 – 120) reports on an informal survey of English teachers which indicated that 11% did not allow any dictionaries in class at all, 37% allowed monolingual dictionaries and only 5% allowed bilingual dictionaries. Other researchers (e.g. Boonmoh & Nesi, 2008), have found a similar situation, with teachers overwhelmingly recommending the use of a monolingual learner’s dictionary: almost all of their students bought one, but the great majority hardly ever used it, preferring instead a digital bilingual version.

Teachers’ preferences for monolingual dictionaries are usually motivated in part by a fear that their students will become too reliant on translation. Whilst this concern remains widespread, much recent suggests that this fear is misguided (Nation, 2013: 424) and that monolingual dictionaries do not actually lead to greater learning gains than their bilingual counterparts. This is, in part, due to the fact that learners typically use these dictionaries in very limited ways – to see if a word exists, check spelling or look up meaning (Harvey & Yuill, 1997). If they made fuller use of the information (about frequency, collocations, syntactic patterns, etc.) on offer, it is likely that learning gains would be greater: ‘it is accessing multiplicity of information that is likely to enhance retention’ (Laufer & Hill, 2000: 77). Without training, however, this is rarely the case.  With lower-level learners, a monolingual learner’s dictionary (even one designed for Elementary level students) can be a frustrating experience, because until they have reached a vocabulary size of around 2,000 – 3,000 words, they will struggle to understand the definitions (Webb & Nation, 2017: 119).

The second reason for teachers’ preference for monolingual dictionaries is that the quality of many bilingual dictionaries is undoubtedly very poor, compared to monolingual learner’s dictionaries such as those produced by Oxford University Press, Cambridge University Press, Longman Pearson, Collins Cobuild, Merriam-Webster and Macmillan, among others. The situation has changed, however, with the rapid growth of bilingualized dictionaries. These contain all the features of a monolingual learner’s dictionary, but also include translations into the learner’s own language. Because of the wealth of information provided by a good bilingualized dictionary, researchers (e.g. Laufer & Hadar, 1997; Chen, 2011) generally consider them preferable to monolingual or normal bilingual dictionaries. They are also popular with learners. Good bilingualized online dictionaries (such as the Oxford Advanced Learner’s English-Chinese Dictionary) are not always free, but many are, and with some language pairings free software can be of a higher quality than services that incur a subscription charge.

If a good bilingualized dictionary is available, there is no longer any compelling reason to use a monolingual learner’s dictionary, unless it contains features which cannot be found elsewhere. In order to compete in a crowded marketplace, many of the established monolingual learner’s dictionaries do precisely that. Examples of good, free online dictionaries include:

Students need help in selecting a dictionary that is right for them. Without this, many end up using as a dictionary a tool such as Google Translate , which, for all its value, is of very limited use as a dictionary. They need to understand that the most appropriate dictionary will depend on what they want to use it for (receptive, reading purposes or productive, writing purposes). Teachers can help in this decision-making process by addressing the issue in class (see the activity below).

In addition to the problem of selecting an appropriate dictionary, it appears that many learners have inadequate dictionary skills (Niitemaa & Pietilä, 2018). In one experiment (Tono, 2011), only one third of the vocabulary searches in a dictionary that were carried out by learners resulted in success. The reasons for failure include focussing on only the first meaning (or translation) of a word that is provided, difficulty in finding the relevant information in long word entries, an inability to find the lemma that is needed, and spelling errors (when they had to type in the word) (Töpel, 2014). As with monolingual dictionaries, learners often only check the meaning of a word in a bilingual dictionary and fail to explore the wider range of information (e.g. collocation, grammatical patterns, example sentences, synonyms) that is available (Laufer & Kimmel, 1997; Laufer & Hill, 2000; Chen, 2010). This information is both useful and may lead to improved retention.

Most learners receive no training in dictionary skills, but would clearly benefit from it. Nation (2013: 333) suggests that at least four or five hours, spread out over a few weeks, would be appropriate. He suggests (ibid: 419 – 421) that training should encourage learners, first, to look closely at the context in which an unknown word is encountered (in order to identify the part of speech, the lemma that needs to be looked up, its possible meaning and to decide whether it is worth looking up at all), then to help learners in finding the relevant entry or sub-entry (by providing information about common dictionary abbreviations (e.g. for parts of speech, style and register)), and, finally, to check this information against the original context.

Two good resource books full of practical activities for dictionary training are available: ‘Dictionary Activities’ by Cindy Leaney (Cambridge: Cambridge University Press, 2007) and ‘Dictionaries’ by Jon Wright (Oxford: Oxford University Press, 1998). Many of the good monolingual dictionaries offer activity guides to promote effective dictionary use and I have suggested a few activities here.

Activity: Understanding a dictionary

Outline: Students explore the use of different symbols in good online dictionaries.

Level: All levels, but not appropriate for very young learners. The activity ‘Choosing a dictionary’ is a good follow-up to this activity.

1 Distribute the worksheet and ask students to follow the instructions.


2 Check the answers.


Activity: Choosing a dictionary

Outline: Students explore and evaluate the features of different free, online bilingual dictionaries.

Level: All levels, but not appropriate for very young learners. The text in stage 3 is appropriate for use with levels A2 and B1. For some groups of learners, you may want to adapt (or even translate) the list of features. It may be useful to do the activity ‘Understanding a dictionary’ before this activity.

1 Ask the class which free, online bilingual dictionaries they like to use. Write some of their suggestions on the board.

2 Distribute the list of features. Ask students to work individually and tick the boxes that are important for them. Ask students to work with a partner to compare their answers.


3 Give students a list of free, online bilingual (English and the students’ own language) dictionaries. You can use suggestions from the list below, add the suggestions that your students made in stage 1, or add your own ideas. (For many language pairings, better resources are available than those in the list below.) Give the students the following short text and ask the students to use two of these dictionaries to look up the underlined words. Ask the students to decide which dictionary they found most useful and / or easiest to use.



4 Conduct feedback with the whole class.

Activity: Getting more out of a dictionary

Outline: Students use a dictionary to help them to correct a text

Level: Levels B1 and B2, but not appropriate for very young learners. For higher levels, a more complex text (with less obvious errors) would be appropriate.

1 Distribute the worksheet below and ask students to follow the instructions.


2 Check answers with the whole class. Ask how easy it was to find the information in the dictionary that they were using.


When you are reading, you probably only need a dictionary when you don’t know the meaning of a word and you want to look it up. For this, a simple bilingual dictionary is good enough. But when you are writing or editing your writing, you will need something that gives you more information about a word: grammatical patterns, collocations (the words that usually go with other words), how formal the word is, and so on. For this, you will need a better dictionary. Many of the better dictionaries are monolingual (see the box), but there are also some good bilingual ones.

Use one (or more) of the online dictionaries in the box (or a good bilingual dictionary) and make corrections to this text. There are eleven mistakes (they have been underlined) in total.


Boonmoh, A. & Nesi, H. 2008. ‘A survey of dictionary use by Thai university staff and students with special reference to pocket electronic dictionaries’ Horizontes de Linguística Aplicada , 6(2), 79 – 90

Chen, Y. 2011. ‘Studies on Bilingualized Dictionaries: The User Perspective’. International Journal of Lexicography, 24 (2): 161–197

Folse, K. 2004. Vocabulary Myths. Ann Arbor: University of Michigan Press

Granger, S. 2012. Electronic Lexicography. Oxford: Oxford University Press

Harvey, K. & Yuill, D. 1997. ‘A study of the use of a monolingual pedagogical dictionary by learners of English engaged in writing’ Applied Linguistics, 51 (1): 253 – 78

Laufer, B. & Hadar, L. 1997. ‘Assessing the effectiveness of monolingual, bilingual and ‘bilingualized’ dictionaries in the comprehension and production of new words’. Modern Language Journal, 81 (2): 189 – 96

Laufer, B. & M. Hill 2000. ‘What lexical information do L2 learners select in a CALL dictionary and how does it affect word retention?’ Language Learning & Technology 3 (2): 58–76

Laufer, B. & Kimmel, M. 1997. ‘Bilingualised dictionaries: How learners really use them’, System, 25 (3): 361 -369

Leaney, C. 2007. Dictionary Activities. Cambridge: Cambridge University Press

Levy, M. and Steel, C. 2015. ‘Language learner perspectives on the functionality and use of electronic language dictionaries’. ReCALL, 27(2): 177–196

Lew, R. & Szarowska, A. 2017. ‘Evaluating online bilingual dictionaries: The case of popular free English-Polish dictionaries’ ReCALL 29(2): 138–159

Nation, I.S.P. 2013. Learning Vocabulary in Another Language 2nd edition. Cambridge: Cambridge University Press

Niitemaa, M.-L. & Pietilä, P. 2018. ‘Vocabulary Skills and Online Dictionaries: A Study on EFL Learners’ Receptive Vocabulary Knowledge and Success in Searching Electronic Sources for Information’, Journal of Language Teaching and Research, 9 (3): 453-462

Tono, Y. 2011. ‘Application of eye-tracking in EFL learners’ dictionary look-up process research’, International Journal of Lexicography 24 (1): 124–153

Töpel, A. 2014. ‘Review of research into the use of electronic dictionaries’ in Müller-Spitzer, C. (Ed.) 2014. Using Online Dictionaries. Berlin: De Gruyter, pp. 13 – 54

Webb, S. & Nation, P. 2017. How Vocabulary is Learned. Oxford: Oxford University Press

Wheeler, G. 2013. Language Teaching through the Ages. New York: Routledge

Wright, J. 1998. Dictionaries. Oxford: Oxford University Press

At a recent ELT conference, a plenary presentation entitled ‘Getting it right with edtech’ (sponsored by a vendor of – increasingly digital – ELT products) began with the speaker suggesting that technology was basically neutral, that what you do with educational technology matters far more than the nature of the technology itself. The idea that technology is a ‘neutral tool’ has a long pedigree and often accompanies exhortations to embrace edtech in one form or another (see for example Fox, 2001). It is an idea that is supported by no less a luminary than Chomsky, who, in a 2012 video entitled ‘The Purpose of Education’ (Chomsky, 2012), said that:

As far as […] technology […] and education is concerned, technology is basically neutral. It’s kind of like a hammer. I mean, […] the hammer doesn’t care whether you use it to build a house or whether a torturer uses it to crush somebody’s skull; a hammer can do either. The same with the modern technology; say, the Internet, and so on.

Womans hammerAlthough hammers are not usually classic examples of educational technology, they are worthy of a short discussion. Hammers come in all shapes and sizes and when you choose one, you need to consider its head weight (usually between 16 and 20 ounces), the length of the handle, the shape of the grip, etc. Appropriate specifications for particular hammering tasks have been calculated in great detail. The data on which these specifications is based on an analysis of the hand size and upper body strength of the typical user. The typical user is a man, and the typical hammer has been designed for a man. The average male hand length is 177.9 mm, that of the average woman is 10 mm shorter (Wang & Cai, 2017). Women typically have about half the upper body strength of men (Miller et al., 1993). It’s possible, but not easy to find hammers designed for women (they are referred to as ‘Ladies hammers’ on Amazon). They have a much lighter head weight, a shorter handle length, and many come in pink or floral designs. Hammers, in other words, are far from neutral: they are highly gendered.

Moving closer to educational purposes and ways in which we might ‘get it right with edtech’, it is useful to look at the smart phone. The average size of these devices has risen in recent years, and is now 5.5 inches, with the market for 6 inch screens growing fast. Why is this an issue? Well, as Caroline Criado Perez (2019: 159) notes, ‘while we’re all admittedly impressed by the size of your screen, it’s a slightly different matter when it comes to fitting into half the population’s hands. The average man can fairly comfortably use his device one-handed – but the average woman’s hand is not much bigger than the handset itself’. This is despite the fact the fact that women are more likely to own an iPhone than men  .

It is not, of course, just technological artefacts that are gendered. Voice-recognition software is also very biased. One researcher (Tatman, 2017) has found that Google’s speech recognition tool is 13% more accurate for men than it is for women. There are also significant biases for race and social class. The reason lies in the dataset that the tool is trained on: the algorithms may be gender- and socio-culturally-neutral, but the dataset is not. It would not be difficult to redress this bias by training the tool on a different dataset.

The same bias can be found in automatic translation software. Because corpora such as the BNC or COCA have twice as many male pronouns as female ones (as a result of the kinds of text that are selected for the corpora), translation software reflects the bias. With Google Translate, a sentence in a language with a gender-neutral pronoun, such as ‘S/he is a doctor’ is rendered into English as ‘He is a doctor’. Meanwhile, ‘S/he is a nurse’ is translated as ‘She is a nurse’ (Criado Perez, 2019: 166).

Datasets, then, are often very far from neutral. Algorithms are not necessarily any more neutral than the datasets, and Cathy O’Neil’s best-seller ‘Weapons of Math Destruction’ catalogues the many, many ways in which algorithms, posing as neutral mathematical tools, can increase racial, social and gender inequalities.

It would not be hard to provide many more examples, but the selection above is probably enough. Technology, as Langdon Winner (Winner, 1980) observed almost forty years ago, is ‘deeply interwoven in the conditions of modern politics’. Technology cannot be neutral: it has politics.

So far, I have focused primarily on the non-neutrality of technology in terms of gender (and, in passing, race and class). Before returning to broader societal issues, I would like to make a relatively brief mention of another kind of non-neutrality: the pedagogic. Language learning materials necessarily contain content of some kind: texts, topics, the choice of values or role models, language examples, and so on. These cannot be value-free. In the early days of educational computer software, one researcher (Biraimah, 1993) found that it was ‘at least, if not more, biased than the printed page it may one day replace’. My own impression is that this remains true today.

Equally interesting to my mind is the fact that all educational technologies, ranging from the writing slate to the blackboard (see Buzbee, 2014), from the overhead projector to the interactive whiteboard, always privilege a particular kind of teaching (and learning). ‘Technologies are inherently biased because they are built to accomplish certain very specific goals which means that some technologies are good for some tasks while not so good for other tasks’ (Zhao et al., 2004: 25). Digital flashcards, for example, inevitably encourage a focus on rote learning. Contemporary LMSs have impressive multi-functionality (i.e. they often could be used in a very wide variety of ways), but, in practice, most teachers use them in very conservative ways (Laanpere et al., 2004). This may be a result of teacher and institutional preferences, but it is almost certainly due, at least in part, to the way that LMSs are designed. They are usually ‘based on traditional approaches to instruction dating from the nineteenth century: presentation and assessment [and] this can be seen in the selection of features which are most accessible in the interface, and easiest to use’ (Lane, 2009).

The argument that educational technology is neutral because it could be put to many different uses, good or bad, is problematic because the likelihood of one particular use is usually much greater than another. There is, however, another way of looking at technological neutrality, and that is to look at its origins. Elsewhere on this blog, in post after post, I have given examples of the ways in which educational technology has been developed, marketed and sold primarily for commercial purposes. Educational values, if indeed there are any, are often an afterthought. The research literature in this area is rich and growing: Stephen Ball, Larry Cuban, Neil Selwyn, Joel Spring, Audrey Watters, etc.

Rather than revisit old ground here, this is an opportunity to look at a slightly different origin of educational technology: the US military. The close connection of the early history of the internet and the Advanced Research Projects Agency (now DARPA) of the United States Department of Defense is fairly well-known. Much less well-known are the very close connections between the US military and educational technologies, which are catalogued in the recently reissued ‘The Classroom Arsenal’ by Douglas D. Noble.

Following the twin shocks of the Soviet Sputnik 1 (in 1957) and Yuri Gagarin (in 1961), the United States launched a massive programme of investment in the development of high-tech weaponry. This included ‘computer systems design, time-sharing, graphics displays, conversational programming languages, heuristic problem-solving, artificial intelligence, and cognitive science’ (Noble, 1991: 55), all of which are now crucial components in educational technology. But it also quickly became clear that more sophisticated weapons required much better trained operators, hence the US military’s huge (and continuing) interest in training. Early interest focused on teaching machines and programmed instruction (branches of the US military were by far the biggest purchasers of programmed instruction products). It was essential that training was effective and efficient, and this led to a wide interest in the mathematical modelling of learning and instruction.

What was then called computer-based education (CBE) was developed as a response to military needs. The first experiments in computer-based training took place at the Systems Research Laboratory of the Air Force’s RAND Corporation think tank (Noble, 1991: 73). Research and development in this area accelerated in the 1960s and 1970s and CBE (which has morphed into the platforms of today) ‘assumed particular forms because of the historical, contingent, military contexts for which and within which it was developed’ (Noble, 1991: 83). It is possible to imagine computer-based education having developed in very different directions. Between the 1960s and 1980s, for example, the PLATO (Programmed Logic for Automatic Teaching Operations) project at the University of Illinois focused heavily on computer-mediated social interaction (forums, message boards, email, chat rooms and multi-player games). PLATO was also significantly funded by a variety of US military agencies, but proved to be of much less interest to the generals than the work taking place in other laboratories. As Noble observes, ‘some technologies get developed while others do not, and those that do are shaped by particular interests and by the historical and political circumstances surrounding their development (Noble, 1991: 4).

According to Noble, however, the influence of the military reached far beyond the development of particular technologies. Alongside the investment in technologies, the military were the prime movers in a campaign to promote computer literacy in schools.

Computer literacy was an ideological campaign rather than an educational initiative – a campaign designed, at bottom, to render people ‘comfortable’ with the ‘inevitable’ new technologies. Its basic intent was to win the reluctant acquiescence of an entire population in a brave new world sculpted in silicon.

The computer campaign also succeeded in getting people in front of that screen and used to having computers around; it made people ‘computer-friendly’, just as computers were being rendered ‘used-friendly’. It also managed to distract the population, suddenly propelled by the urgency of learning about computers, from learning about other things, such as how computers were being used to erode the quality of their working lives, or why they, supposedly the citizens of a democracy, had no say in technological decisions that were determining the shape of their own futures.

Third, it made possible the successful introduction of millions of computers into schools, factories and offices, even homes, with minimal resistance. The nation’s public schools have by now spent over two billion dollars on over a million and a half computers, and this trend still shows no signs of abating. At this time, schools continue to spend one-fifth as much on computers, software, training and staffing as they do on all books and other instructional materials combined. Yet the impact of this enormous expenditure is a stockpile of often idle machines, typically used for quite unimaginative educational applications. Furthermore, the accumulated results of three decades of research on the effectiveness of computer-based instruction remain ‘inconclusive and often contradictory’. (Noble, 1991: x – xi)

Rather than being neutral in any way, it seems more reasonable to argue, along with (I think) most contemporary researchers, that edtech is profoundly value-laden because it has the potential to (i) influence certain values in students; (ii) change educational values in [various] ways; and (iii) change national values (Omotoyinbo & Omotoyinbo, 2016: 173). Most importantly, the growth in the use of educational technology has been accompanied by a change in the way that education itself is viewed: ‘as a tool, a sophisticated supply system of human cognitive resources, in the service of a computerized, technology-driven economy’ (Noble, 1991: 1). These two trends are inextricably linked.


Biraimah, K. 1993. The non-neutrality of educational computer software. Computers and Education 20 / 4: 283 – 290

Buzbee, L. 2014. Blackboard: A Personal History of the Classroom. Minneapolis: Graywolf Press

Chomsky, N. 2012. The Purpose of Education (video). Learning Without Frontiers Conference.

Criado Perez, C. 2019. Invisible Women. London: Chatto & Windus

Fox, R. 2001. Technological neutrality and practice in higher education. In A. Herrmann and M. M. Kulski (Eds), Expanding Horizons in Teaching and Learning. Proceedings of the 10th Annual Teaching Learning Forum, 7-9 February 2001. Perth: Curtin University of Technology.

Laanpere, M., Poldoja, H. & Kikkas, K. 2004. The second thoughts about pedagogical neutrality of LMS. Proceedings of IEEE International Conference on Advanced Learning Technologies, 2004.

Lane, L. 2009. Insidious pedagogy: How course management systems impact teaching. First Monday, 14(10).

Miller, A.E., MacDougall, J.D., Tarnopolsky, M. A. & Sale, D.G. 1993. ‘Gender differences in strength and muscle fiber characteristics’ European Journal of Applied Physiology and Occupational Physiology. 66(3): 254-62

Noble, D. D. 1991. The Classroom Arsenal. Abingdon, Oxon.: Routledge

Omotoyinbo, D. W. & Omotoyinbo, F. R. 2016. Educational Technology and Value Neutrality. Societal Studies, 8 / 2: 163 – 179

O’Neil, C. 2016. Weapons of Math Destruction. London: Penguin

Sundström, P. Interpreting the Notion that Technology is Value Neutral. Medicine, Health Care and Philosophy 1, 1998: 42-44

Tatman, R. 2017. ‘Gender and Dialect Bias in YouTube’s Automatic Captions’ Proceedings of the First Workshop on Ethics in Natural Language Processing, pp. 53–59

Wang, C. & Cai, D. 2017. ‘Hand tool handle design based on hand measurements’ MATEC Web of Conferences 119, 01044 (2017)

Winner, L. 1980. Do Artifacts have Politics? Daedalus 109 / 1: 121 – 136

Zhao, Y, Alvarez-Torres, M. J., Smith, B. & Tan, H. S. 2004. The Non-neutrality of Technology: a Theoretical Analysis and Empirical Study of Computer Mediated Communication Technologies. Journal of Educational Computing Research 30 (1 &2): 23 – 55

Knowble, claims its developers, is a browser extension that will improve English vocabulary and reading comprehension. It also describes itself as an ‘adaptive language learning solution for publishers’. It’s currently beta and free, and sounds right up my street so I decided to give it a run.

Knowble reader

Users are asked to specify a first language (I chose French) and a level (A1 to C2): I chose B1, but this did not seem to impact on anything that subsequently happened. They are then offered a menu of about 30 up-to-date news items, grouped into 5 categories (world, science, business, sport, entertainment). Clicking on one article takes you to the article on the source website. There’s a good selection, including USA Today, CNN, Reuters, the Independent and the Torygraph from Britain, the Times of India, the Independent from Ireland and the Star from Canada. A large number of words are underlined: a single click brings up a translation in the extension box. Double-clicking on all other words will also bring up translations. Apart from that, there is one very short exercise (which has presumably been automatically generated) for each article.

For my trial run, I picked three articles: ‘Woman asks firefighters to help ‘stoned’ raccoon’ (from the BBC, 240 words), ‘Plastic straw and cotton bud ban proposed’ (also from the BBC, 823 words) and ‘London’s first housing market slump since 2009 weighs on UK price growth’ (from the Torygraph, 471 words).


Research suggests that the use of translations, rather than definitions, may lead to more learning gains, but the problem with Knowble is that it relies entirely on Google Translate. Google Translate is fast improving. Take the first sentence of the ‘plastic straw and cotton bud’ article, for example. It’s not a bad translation, but it gets the word ‘bid’ completely wrong, translating it as ‘offre’ (= offer), where ‘tentative’ (= attempt) is needed. So, we can still expect a few problems with Google Translate …

google_translateOne of the reasons that Google Translate has improved is that it no longer treats individual words as individual lexical items. It analyses groups of words and translates chunks or phrases (see, for example, the way it translates ‘as part of’). It doesn’t do word-for-word translation. Knowble, however, have set their software to ask Google for translations of each word as individual items, so the phrase ‘as part of’ is translated ‘comme’ + ‘partie’ + ‘de’. Whilst this example is comprehensible, problems arise very quickly. ‘Cotton buds’ (‘cotons-tiges’) become ‘coton’ + ‘bourgeon’ (= botanical shoots of cotton). Phrases like ‘in time’, ‘run into’, ‘sleep it off’ ‘take its course’, ‘fire station’ or ‘going on’ (all from the stoned raccoon text) all cause problems. In addition, Knowble are not using any parsing tools, so the system does not identify parts of speech, and further translation errors inevitably appear. In the short article of 240 words, about 10% are wrongly translated. Knowble claim to be using NLP tools, but there’s no sign of it here. They’re just using Google Translate rather badly.

Highlighted items

word_listNLP tools of some kind are presumably being used to select the words that get underlined. Exactly how this works is unclear. On the whole, it seems that very high frequency words are ignored and that lower frequency words are underlined. Here, for example, is the list of words that were underlined in the stoned raccoon text. I’ve compared them with (1) the CEFR levels for these words in the English Profile Text Inspector, and (2) the frequency information from the Macmillan dictionary (more stars = more frequent). In the other articles, some extremely high frequency words were underlined (e.g. price, cost, year) while much lower frequency items were not.

It is, of course, extremely difficult to predict which items of vocabulary a learner will know, even if we have a fairly accurate idea of their level. Personal interests play a significant part, so, for example, some people at even a low level will have no problem with ‘cannabis’, ‘stoned’ and ‘high’, even if these are low frequency. First language, however, is a reasonably reliable indicator as cognates can be expected to be easy. A French speaker will have no problem with ‘appreciate’, ‘unique’ and ‘symptom’. A recommendation engine that can meaningfully personalize vocabulary suggestions will, at the very least, need to consider cognates.

In short, the selection and underlining of vocabulary items, as it currently stands in Knowble, appears to serve no clear or useful function.

taskVocabulary learning

Knowble offers a very short exercise for each article. They are of three types: word completion, dictation and drag and drop (see the example). The rationale for the selection of the target items is unclear, but, in any case, these exercises are tokenistic in the extreme and are unlikely to lead to any significant learning gains. More valuable would be the possibility of exporting items into a spaced repetition flash card system.

effectiveThe claim that Knowble’s ‘learning effect is proven scientifically’ seems to me to be without any foundation. If there has been any proper research, it’s not signposted anywhere. Sure, reading lots of news articles (with a look-up function – if it works reliably) can only be beneficial for language learners, but they can do that with any decent dictionary running in the background.

Similar in many ways to, which I reviewed in my last post, Knowble is another example of a technology-driven product that shows little understanding of language learning.