Posts Tagged ‘corpora’

A personalized language learning programme that is worth its name needs to offer a wide variety of paths to accommodate the varying interests, priorities, levels and preferred approaches to learning of the users of the programme. For this to be possible, a huge quantity of learning material is needed (Iwata et al., 2011: 1): the preparation and curation of this material is extremely time-consuming and expensive (despite the pittance that is paid to writers and editors). It’s not surprising, then, that a growing amount of research is being devoted to the exploration of ways of automatically generating language learning material. One area that has attracted a lot of attention is the learning of vocabulary.

Memrise screenshot 2Many simple vocabulary learning tasks are relatively simple to generate automatically. These include matching tasks of various kinds, such as the matching of words or phrases to meanings (either in English or the L1), pictures or collocations, as in many flashcard apps. Doing it well is rather harder: the definitions or translations have to be good and appropriate for learners of the level, the pictures need to be appropriate. If, as is often the case, the lexical items have come from a text or form part of a group of some kind, sense disambiguation software will be needed to ensure that the right meaning is being practised. Anyone who has used flashcard apps knows that the major problem is usually the quality of the content (whether it has been automatically generated or written by someone).

A further challenge is the generation of distractors. In the example here (from Memrise), the distractors have been so badly generated as to render the task more or less a complete waste of time. Distractors must, in some way, be viable alternatives (Smith et al., 2010) but still clearly wrong. That means they should normally be the same part of speech and true cognates should be avoided. Research into the automatic generation of distractors is well-advanced (see, for instance, Kumar at al., 2015) with Smith et al (2010), for example, using a very large corpus and various functions of Sketch Engine (the most well-known corpus query tool) to find collocates and other distractors. Their TEDDCLOG (Testing English with Data-Driven CLOze Generation) system produced distractors that were deemed acceptable 91% of the time. Whilst impressive, there is still a long way to go before human editing / rewriting is no longer needed.

One area that has attracted attention is, of course, tests, and some tasks, such as those in TOEFL (see image). Susanti et al (2015, 2017) were able, given a target word, to automatically generate a reading passage from web sources along with questions of the TOEFL kind. However, only about half of them were considered good enough to be used in actual tests. Again, that is some way off avoiding human intervention altogether, but the automatically generated texts and questions can greatly facilitate the work of human item writers.

toefl task

 

Other tools that might be useful include the University of Nottingham AWL (Academic Word List) Gapmaker . This allows users to type or paste in a text, from which items from the AWL are extracted and replaced as a gap. See the example below. It would, presumably, not be too difficult, to combine this approach with automatic distractor generation and to create multiple choice tasks.

Nottingham_AWL_Gapmaster

WordGapThere are a number of applications that offer the possibility of generating cloze tasks from texts selected by the user (learner or teacher). These have not always been designed with the language learner in mind but one that was is the Android app, WordGap (Knoop & Wilske, 2013). Described by its developers as a tool that ‘provides highly individualized exercises to support contextualized mobile vocabulary learning …. It matches the interests of the learner and increases the motivation to learn’. It may well do all that, but then again, perhaps not. As Knoop & Wilske acknowledge, it is only appropriate for adult, advanced learners and its value as a learning task is questionable. The target item that has been automatically selected is ‘novel’, a word that features in the list Oxford 2000 Keywords (as do all three distractors), and therefore ought to be well below the level of the users. Some people might find this fun, but, in terms of learning, they would probably be better off using an app that made instant look-up of words in the text possible.

More interesting, in my view, is TEDDCLOG (Smith et al., 2010), a system that, given a target learning item (here the focus is on collocations), trawls a large corpus to find the best sentence that illustrates it. ‘Good sentences’ were defined as those which were short (but not too short, or there is not enough useful context, begins with a capital letter and ends with a full stop, has a maximum of two commas; and otherwise contains only the 26 lowercase letters. It must be at a lexical and grammatical level that an intermediate level learner of English could be expected to understand. It must be well-formed and without too much superfluous material. All others were rejected. TEDDCLOG uses Sketch Engine’s GDEX function (Good Dictionary Example Extractor, Kilgarriff et al 2008) to do this.

My own interest in this area came about as a result of my work in the development of the Oxford Vocabulary Trainer . The app offers the possibility of studying both pre-determined lexical items (e.g. the vocabulary list of a coursebook that the learner is using) and free choice (any item could be activated and sent to a learning queue). In both cases, practice takes the form of sentences with the target item gapped. There are a range of hints and help options available to the learner, and feedback is both automatic and formative (i.e. if the supplied answer is not correct, hints are given to push the learner to do better on a second attempt). Leveraging some fairly heavy technology, we were able to achieve a fair amount of success in the automation of intelligent feedback, but what had, at first sight, seemed a lesser challenge – the generation of suitable ‘carrier sentences’, proved more difficult.

The sentences which ‘carry’ the gap should, ideally, be authentic: invented examples often ‘do not replicate the phraseology and collocational preferences of naturally-occurring text’ (Smith et al., 2010). The technology of corpus search tools should allow us to do a better job than human item writers. For that to be the case, we need not only good search tools but a good corpus … and some are better than others for the purposes of language learning. As Fenogenova & Kuzmenko (2016) discovered when using different corpora to automatically generate multiple choice vocabulary exercises, the British Academic Written English corpus (BAWE) was almost 50% more useful than the British National Corpus (BNC). In the development of the Oxford Vocabulary Trainer, we thought we had the best corpus we could get our hands on – the tagged corpus used for the production of the Oxford suite of dictionaries. We could, in addition and when necessary, turn to other corpora, including the BAWE and the BNC. Our requirements for acceptable carrier sentences were similar to those of Smith et al (2010), but were considerably more stringent.

To cut quite a long story short, we learnt fairly quickly that we simply couldn’t automate the generation of carrier sentences with sufficient consistency or reliability. As with some of the other examples discussed in this post, we were able to use the technology to help the writers in their work. We also learnt (rather belatedly, it has to be admitted) that we were trying to find technological solutions to problems that we hadn’t adequately analysed at the start. We hadn’t, for example, given sufficient thought to learner differences, especially the role of L1 (and other languages) in learning English. We hadn’t thought enough about the ‘messiness’ of either language or language learning. It’s possible, given enough resources, that we could have found ways of improving the algorithms, of leveraging other tools, or of deploying additional databases (especially learner corpora) in our quest for a personalised vocabulary learning system. But, in the end, it became clear to me that we were only nibbling at the problem of vocabulary learning. Deliberate learning of vocabulary may be an important part of acquiring a language, but it remains only a relatively small part. Technology may be able to help us in a variety of ways (and much more so in testing than learning), but the dreams of the data scientists (who wrote much of the research cited here) are likely to be short-lived. Experienced writers and editors of learning materials will be needed for the foreseeable future. And truly personalized vocabulary learning, fully supported by technology, will not be happening any time soon.

 

References

Fenogenova, A. & Kuzmenko, E. 2016. Automatic Generation of Lexical Exercises Available online at http://www.dialog-21.ru/media/3477/fenogenova.pdf

Iwata, T., Goto, T., Kojiri, T., Watanabe, T. & T. Yamada. 2011. ‘Automatic Generation of English Cloze Questions Based on Machine Learning’. NTT Technical Review Vol. 9 No. 10 Oct. 2011

Kilgarriff, A. et al. 2008. ‘GDEX: Automatically Finding Good Dictionary Examples in a Corpus.’ In E. Bernal and J. DeCesaris (eds.), Proceedings of the XIII EURALEX International Congress: Barcelona, 15-19 July 2008. Barcelona: l’Institut Universitari de Lingüística Aplicada (IULA) dela Universitat Pompeu Fabra, 425–432.

Knoop, S. & Wilske, S. 2013. ‘WordGap – Automatic generation of gap-filling vocabulary exercises for mobile learning’. Proceedings of the second workshop on NLP for computer-assisted language learning at NODALIDA 2013. NEALT Proceedings Series 17 / Linköping Electronic Conference Proceedings 86: 39–47. Available online at http://www.ep.liu.se/ecp/086/004/ecp13086004.pdf

Kumar, G., Banchs, R.E. & D’Haro, L.F. 2015. ‘RevUP: Automatic Gap-Fill Question Generation from Educational Texts’. Proceedings of the Tenth Workshop on Innovative Use of NLP for Building Educational Applications, 2015, pp. 154–161, Denver, Colorado, June 4, Association for Computational Linguistics

Smith, S., Avinesh, P.V.S. & Kilgariff, A. 2010. ‘Gap-fill tests for Language Learners: Corpus-Driven Item Generation’. Proceedings of ICON-2010: 8th International Conference on Natural Language Processing, Macmillan Publishers, India. Available online at https://curve.coventry.ac.uk/open/file/2b755b39-a0fa-4171-b5ae-5d39568874e5/1/smithcomb2.pdf

Susanti, Y., Iida, R. & Tokunaga, T. 2015. ‘Automatic Generation of English Vocabulary Tests’. Proceedings of 7th International Conference on Computer Supported Education. Available online https://pdfs.semanticscholar.org/aead/415c1e07803756902b859e8b6e47ce312d96.pdf

Susanti, Y., Tokunaga, T., Nishikawa, H. & H. Obari 2017. ‘Evaluation of automatically generated English vocabulary questions’ Research and Practice in Technology Enhanced Learning 12 / 11

 

Advertisements

Every now and then, someone recommends me to take a look at a flashcard app. It’s often interesting to see what developers have done with design, gamification and UX features, but the content is almost invariably awful. Most recently, I was encouraged to look at Word Pash. The screenshots below are from their promotional video.

word-pash-1 word-pash-2 word-pash-3 word-pash-4

The content problems are immediately apparent: an apparently random selection of target items, an apparently random mix of high and low frequency items, unidiomatic language examples, along with definitions and distractors that are less frequent than the target item. I don’t know if these are representative of the rest of the content. The examples seem to come from ‘Stage 1 Level 3’, whatever that means. (My confidence in the product was also damaged by the fact that the Word Pash website includes one testimonial from a certain ‘Janet Reed – Proud Mom’, whose son ‘was able to increase his score and qualify for academic scholarships at major universities’ after using the app. The picture accompanying ‘Janet Reed’ is a free stock image from Pexels and ‘Janet Reed’ is presumably fictional.)

According to the website, ‘WordPash is a free-to-play mobile app game for everyone in the global audience whether you are a 3rd grader or PhD, wordbuff or a student studying for their SATs, foreign student or international business person, you will become addicted to this fast paced word game’. On the basis of the promotional video, the app couldn’t be less appropriate for English language learners. It seems unlikely that it would help anyone improve their ACT or SAT test scores. The suggestion that the vocabulary development needs of 9-year-olds and doctoral students are comparable is pure chutzpah.

The deliberate study of more or less random words may be entertaining, but it’s unlikely to lead to very much in practical terms. For general purposes, the deliberate learning of the highest frequency words, up to about a frequency ranking of #7500, makes sense, because there’s a reasonably high probability that you’ll come across these items again before you’ve forgotten them. Beyond that frequency level, the value of the acquisition of an additional 1000 words tails off very quickly. Adding 1000 words from frequency ranking #8000 to #9000 is likely to result in an increase in lexical understanding of general purpose texts of about 0.2%. When we get to frequency ranks #19,000 to #20,000, the gain in understanding decreases to 0.01%[1]. In other words, deliberate vocabulary learning needs to be targeted. The data is relatively recent, but the principle goes back to at least the middle of the last century when Michael West argued that a principled approach to vocabulary development should be driven by a comparison of the usefulness of a word and its ‘learning cost’[2]. Three hundred years before that, Comenius had articulated something very similar: ‘in compiling vocabularies, my […] concern was to select the words in most frequent use[3].

I’ll return to ‘general purposes’ later in this post, but, for now, we should remember that very few language learners actually study a language for general purposes. Globally, the vast majority of English language learners study English in an academic (school) context and their immediate needs are usually exam-specific. For them, general purpose frequency lists are unlikely to be adequate. If they are studying with a coursebook and are going to be tested on the lexical content of that book, they will need to use the wordlist that matches the book. Increasingly, publishers make such lists available and content producers for vocabulary apps like Quizlet and Memrise often use them. Many examinations, both national and international, also have accompanying wordlists. Examples of such lists produced by examination boards include the Cambridge English young learners’ exams (Starters, Movers and Flyers) and Cambridge English Preliminary. Other exams do not have official word lists, but reasonably reliable lists have been produced by third parties. Examples include Cambridge First, IELTS and SAT. There are, in addition, well-researched wordlists for academic English, including the Academic Word List (AWL)  and the Academic Vocabulary List  (AVL). All of these make sensible starting points for deliberate vocabulary learning.

When we turn to other, out-of-school learners the number of reasons for studying English is huge. Different learners have different lexical needs, and working with a general purpose frequency list may be, at least in part, a waste of time. EFL and ESL learners are likely to have very different needs, as will EFL and ESP learners, as will older and younger learners, learners in different parts of the world, learners who will find themselves in English-speaking countries and those who won’t, etc., etc. For some of these demographics, specialised corpora (from which frequency-based wordlists can be drawn) exist. For most learners, though, the ideal list simply does not exist. Either it will have to be created (requiring a significant amount of time and expertise[4]) or an available best-fit will have to suffice. Paul Nation, in his recent ‘Making and Using Word Lists for Language Learning and Testing’ (John Benjamins, 2016) includes a useful chapter on critiquing wordlists. For anyone interested in better understanding the issues surrounding the development and use of wordlists, three good articles are freely available online. These are:making-and-using-word-lists-for-language-learning-and-testing

Lessard-Clouston, M. 2012 / 2013. ‘Word Lists for Vocabulary Learning and Teaching’ The CATESOL Journal 24.1: 287- 304

Lessard-Clouston, M. 2016. ‘Word lists and vocabulary teaching: options and suggestions’ Cornerstone ESL Conference 2016

Sorell, C. J. 2013. A study of issues and techniques for creating core vocabulary lists for English as an International Language. Doctoral thesis.

But, back to ‘general purposes’ …. Frequency lists are the obvious starting point for preparing a wordlist for deliberate learning, but they are very problematic. Frequency rankings depend on the corpus on which they are based and, since these are different, rankings vary from one list to another. Even drawing on just one corpus, rankings can be a little strange. In the British National Corpus, for example, ‘May’ (the month) is about twice as frequent as ‘August’[5], but we would be foolish to infer from this that the learning of ‘May’ should be prioritised over the learning of ‘August’. An even more striking example from the same corpus is the fact that ‘he’ is about twice as frequent as ‘she’[6]: should, therefore, ‘he’ be learnt before ‘she’?

List compilers have to make a number of judgement calls in their work. There is not space here to consider these in detail, but two particularly tricky questions concerning the way that words are chosen may be mentioned: Is a verb like ‘list’, with two different and unrelated meanings, one word or two? Should inflected forms be considered as separate words? The judgements are not usually informed by considerations of learners’ needs. Learners will probably best approach vocabulary development by building their store of word senses: attempting to learn all the meanings and related forms of any given word is unlikely to be either useful or successful.

Frequency lists, in other words, are not statements of scientific ‘fact’: they are interpretative documents. They have been compiled for descriptive purposes, not as ways of structuring vocabulary learning, and it cannot be assumed they will necessarily be appropriate for a purpose for which they were not designed.

A further major problem concerns the corpus on which the frequency list is based. Large databases, such as the British National Corpus or the Corpus of Contemporary American English, are collections of language used by native speakers in certain parts of the world, usually of a restricted social class. As such, they are of relatively little value to learners who will be using English in contexts that are not covered by the corpus. A context where English is a lingua franca is one such example.

A different kind of corpus is the Cambridge Learner Corpus (CLC), a collection of exam scripts produced by candidates in Cambridge exams. This has led to the development of the English Vocabulary Profile (EVP) , where word senses are tagged as corresponding to particular levels in the Common European Framework scale. At first glance, this looks like a good alternative to frequency lists based on native-speaker corpora. But closer consideration reveals many problems. The design of examination tasks inevitably results in the production of language of a very different kind from that produced in other contexts. Many high frequency words simply do not appear in the CLC because it is unlikely that a candidate would use them in an exam. Other items are very frequent in this corpus just because they are likely to be produced in examination tasks. Unsurprisingly, frequency rankings in EVP do not correlate very well with frequency rankings from other corpora. The EVP, then, like other frequency lists, can only serve, at best, as a rough guide for the drawing up of target item vocabulary lists in general purpose apps or coursebooks[7].

There is no easy solution to the problems involved in devising suitable lexical content for the ‘global audience’. Tagging words to levels (i.e. grouping them into frequency bands) will always be problematic, unless very specific user groups are identified. Writers, like myself, of general purpose English language teaching materials are justifiably irritated by some publishers’ insistence on allocating words to levels with numerical values. The policy, taken to extremes (as is increasingly the case), has little to recommend it in linguistic terms. But it’s still a whole lot better than the aleatory content of apps like Word Pash.

[1] See Nation, I.S.P. 2013. Learning Vocabulary in Another Language 2nd edition. (Cambridge: Cambridge University Press) p. 21 for statistical tables. See also Nation, P. & R. Waring 1997. ‘Vocabulary size, text coverage and word lists’ in Schmitt & McCarthy (eds.) 1997. Vocabulary: Description, Acquisition and Pedagogy. (Cambridge: Cambridge University Press) pp. 6 -19

[2] See Kelly, L.G. 1969. 25 Centuries of Language Teaching. (Rowley, Mass.: Rowley House) p.206 for a discussion of West’s ideas.

[3] Kelly, L.G. 1969. 25 Centuries of Language Teaching. (Rowley, Mass.: Rowley House) p. 184

[4] See Timmis, I. 2015. Corpus Linguistics for ELT (Abingdon: Routledge) for practical advice on doing this.

[5] Nation, I.S.P. 2016. Making and Using Word Lists for Language Learning and Testing. (Amsterdam: John Benjamins) p.58

[6] Taylor, J.R. 2012. The Mental Corpus. (Oxford: Oxford University Press) p.151

[7] For a detailed critique of the limitations of using the CLC as a guide to syllabus design and textbook development, see Swan, M. 2014. ‘A Review of English Profile Studies’ ELTJ 68/1: 89-96