Artificial Intelligence in Second Language Learning

Raising Error Awareness

Marina Dodigovic

• Publisher: Multilingual Matters
• Year: 2005
• Print ISBN: 9781853598296
• eBook ISBN: 9781788920346
• Topic: Languages & Linguistics
• Subtopic: Literacy
• Index: Languages & Linguistics

This volume argues that adults can learn English as a second language if their typical errors are corrected systematically and in line with their preferred style of learning. The remedy designed for this purpose relies on artificial intelligence. The book describes original research which demonstrates the success of this approach.

Chapter 1

Can Another Language Be Learnt?

Background

Occasionally one hears that so-and-so speaks five or seven languages fluently, which seems like a truly remarkable achievement, except for the fact that the speaker as a rule fails to deliver any accurate measure of so-and-so’s proficiency in any of those languages. Those who have struggled with another language to little or no avail would most certainly greet such an unqualified statement with a sound dose of scepticism, as they might hold the belief that another language cannot be really mastered to perfection. Is it then possible to learn and use a language other than one’s mother tongue with a native like proficiency? If so, at what age would one be most likely to achieve this? Moreover, would that mastery extend to every aspect of the target language, including lexicon and grammar; all language skills including speaking, listening, reading, writing; features such as idiomatic expressions and language based humour; and the command of functional varieties of that language like sociolects or registers? Given the above questions, second language learnability appears to be somewhat of an undefined term, which we will seek to clarify in the following review of literature. In the interim, the stance taken here is that speakers of languages other than English can achieve error free use of written academic English, regardless of their age, especially if provided with a learning aid that accommodates their specific learning needs.

The situation that has produced the above assertion is the following. Imagine a university in an English speaking country, e.g. Australia, Canada, New Zealand, UK or US, which enrols considerably large numbers of students for whom English is a second language. Prior to enrolment, these students would have had to demonstrate some sort of acceptable standard of English, be it through one of the international proficiency tests, such as IELTS 6.5 or TOEFL 550/580, the institution’s own internal test or an equivalent language proficiency score. Based on those scores, the institution’s assumption is frequently that this population of students is sufficiently equipped to attend to the content delivered in the classes (Severino, 2001). Under the increasing pressure of two factors, this assumption has however begun to weaken, the factors being the varying academic success of this group of students and the increasing body of research in second language acquisition. Consequently, a number of universities are now offering additional language assistance to non-native speakers (NNS) of English, be it English for Academic Purposes (EAP) programs, Writing/Language Across the Curriculum (WAC/LAC), writing centres, study skills or similar (Johns, 1997).

The specific multi-national, predominantly South-East Asian, student population discussed in this book was enrolled in an Australian university and came from a variety of disciplines and modes of study. They had access to a battery of EAP programs, consisting of two credit courses in academic writing, an array of non-credit courses on varying topics of interest, only one of which – and that poorly attended – was devoted to grammar, in addition to individual consultations available at the time of their choice. Careful monitoring of the effect of all of the above measures yielded encouraging results in terms of overall achievement. The students generally seemed to have acquired more efficient reading skills and more successful approaches to writing, leading to an overall better academic performance as demonstrated by improved grades in their content-related subjects. However, one thing remained – language errors, which sometimes obscured the meaning and distracted from the message (Eskenazi, 1999) even in a well structured and most carefully researched assignment. Even though a local survey (Simmons & Thurstun, 1995) had established that the lecturers valued content much more than the linguistic form, the grades they awarded painted a slightly different picture. It is arguable that non-native-like and therefore unexpected linguistic structures would make the comprehension of the message (and therefore content) more difficult. Hence, the lecturers might have in reality been struggling with the meaning affected by form. Be that as it may, language errors became identified as a problem and their eradication a desired outcome.

The students themselves are either overseas students or recent migrants, mostly very self-conscious and uncertain when it comes to the issue of language errors. Most are struggling with a new culture and a new academic tradition, in which the participants such as students and teachers have new and different social roles (Scollon & Scollon, 1995). Not only is the teaching process new and different (Ballard & Clanchy, 1984), but the finances are not always certain or sufficient. Financial crises in their countries of origin as well as civil unrests are a constant source of concern. Nevertheless, most are determined to succeed.

In order to give the reader a better feel for this student population, I would like to introduce two particular representatives: Eric and Jean. The names are of course aliases designed to protect the real identity of the two students, although both actually used different English aliases. These students are very different from the two mentioned in the Introduction to this volume, in the sense that their coping with a second language environment seemed to be much better. Moreover, they were the students most likely to accept extra help and benefit from it. At the beginning of this study Eric was a 19-year-old undergraduate student of Economics and Finance from Indonesia. His spoken English was fluent; however, in his writing, which was well structured, he often confused parts of speech, omitted the copula and confused finite and non-finite forms. He seemed confident and friendly, but corrections made by a female teacher seemed to be a sensitive issue. He was facing a personal crisis as his funding was interrupted through civil unrests in his country. The university provided him with an interim scholarship, which made him feel somewhat uncomfortable. He was also concerned for the safety of his parents back home. Jean was 23 and came from mainland China. She was a graduate in music, very disciplined and determined. However, communicating with her on campus supervisor was a challenge. As language was not the main area of her expertise, writing presented a major challenge. Verb inflection and transitivity were the grammar areas in which she willingly sought improvement. Funding was not an issue for Jean and she was satisfied that both her parents were safe and sound. Both of these students were keen on improving their English in anticipation of perceived advantages they might gain with the eradication of language errors. Eric’s and Jean’s profiles will hopefully come to mind when the phrase ‘our EAP students’ is mentioned in the course of this book.

Thus the small number of academics in charge of the EAP programs at the university where our EAP students were enrolled faced the daunting challenge of understanding and successfully combating the linguistic errors of a rather large and diversified student population. Besides gaining a thorough understanding of theoretical explanations for the persistence of such errors and their rather diametrally opposite prognoses of success, potentially successful remediation strategies had to be identified and a vehicle for their implementation selected. As the latter required nothing short of a miracle, artificial intelligence was isolated as the only concept that might work. Several chapters of this book grapple with the issue of learner errors in writing and finding ways of their remediation, while one of them is specifically dedicated to the use of artificial intelligence for this purpose. For now, however, we will return to the more general matters addressed in this volume.

Terminology

So far we have mentioned terms such as ‘mother tongue’, ‘another language’ and ‘second language’ to distinguish between different relationships one can assume toward languages one has had exposure to. It is now time to establish a uniform nomenclature, the one that is generally accepted in the relevant literature (Larsen-Freeman & Long, 1991; Cook, 1993; Ellis, 1997), namely the distinction between the first language (L1) and the second language (L2). The first language can be equated with the mother tongue, native language or simply the language one has acquired first. A second language is any language other than L1 a learner is seeking to acquire, even though this might in reality be their fourth or fifth language (Larsen-Freeman & Long, 1991). However, the distinction between L1 and L2becomes problematic in the case of simultaneous bilingualism which, according to Larsen-Freeman & Long (1991: 7), means mastery of two first languages. Sometimes, distinction is made according to the setting in which L2 is learnt. Thus we speak of a foreign language if it is learnt outside the country or countries where it is spoken as the first language, and of a second language if the acquisition takes place in an environment where the target language features as the first language (Oxford, 1990; Mitchell & Myles, 1998; Grabe & Kaplan, 1996). For instance a Korean student learning English at a school in Korea would be learning a foreign language, whereas the same student learning English in Australia would be learning it as a second language. It could also be assumed that the learning purposes would be different in these two cases: the purpose of the former would most probably be to enhance the general education of the student, whereas the latter would be to communicate and function successfully in another society. We will follow the now established tradition of the discipline and subsume both instances under L2, since even though learning purposes and therefore the teaching context may be different (Grabe & Kaplan, 1996), the learning processes are perceived as identical (Mitchell & Myles, 1998).

The discipline referred to in the previous paragraph is the study of second language acquisition (SLA). Nunan (1992: 232) defines SLA as ‘the process through which individuals develop skills in a second or foreign language in tutored or untutored environments’. This definition circumvents the distinction between ‘learning’ and ‘acquisition’, which is how the term will be used throughout the book. It is however worth mentioning that the distinction, if not in terminology at least in principle, does have an almost 80 year long history. Thus Palmer (1926, cited in Cook, 1993: 63) proposes a distinction between two different capacities for language learning: ‘spontaneous’ and ‘studious’. Krashen (1987), probably the best known proponent of this distinction, postulates an acquisition/learning hypothesis, which presumes that acquisition is a spontaneous process, similar to the one that children engage in when learning their first language. Learning (Krashen, 1987), on the other hand, is considered to be a conscious process, used exclusively for monitoring the output. Of the two, only the former is really essential, leading to acquired knowledge, deemed to be productive in L2 use (Krashen, 1987). In this book learning and acquisition will be used interchangeably.

Theory

The use of terminology discussed in the previous section was sparked off by a debate not only concerning the nature of the language learning process, but also of the product, the knowledge of language itself. Nowadays, we can choose from an array of theories, some of which include a more comprehensive theory of language. What is meant by theory is a constructed framework intended to explain and predict the real life phenomena (Larsen-Freeman & Long, 1991). Having a theory allows the researcher to ask research questions and postulate hypotheses before the commencement of research (Ellis, 1997; Nunan, 1992; McDonough & McDonough, 1997; Gregg, 2001). This is an approach opposite to data-driven research (McDonough & McDonough, 1997), which collects data first and looks for emerging patterns to ask questions about. Theory is however valued for its contribution of systematicity that it can provide research and practise with (Ellis,1997; Grabe&Kaplan,1996; Gregg,2001).

Cook (1993: 246) subdivides SLA theoretical approaches into two main groups: (1) those that assume that language is acquired and represented by the human mind in a way which is unique to it, and (2) those that believe that language is no exception to the way people acquire and store knowledge in general. The first group is mainly under the influence of Chomsky’s Universal Grammar, whereas the second group is recruited from a number of different psychological, psycholinguistic and sociolinguistic camps. Larsen-Freeman and Long (1991) subdivide all SLA theories into nativist, environmentalist and interactionist. The nativist theories are those claiming that the ability to learn a language is innate, specific to language and different to any other mental ability. Thus these are identical with Cook’s (1993) first group. Environmentalists believe that nurture, rather than nature is key to learning, whereas interactionists acknowledge the role of both nature and nurture, the innate and environmental factors. Oxford (1995) talks about novice-to-expert, constructivist and individual difference paradigms. While the first paradigm is one of progress from the stage of being a novice to the stage of being an expert, the second paradigm is associated with constructivism in psychology and will be discussed in more detail later. The individual difference theory suggests that learners learn in different ways and therefore no single methodology will serve all learners equally well.

While SLA overviews, such as that of Mitchell and Myles (1998), observe the theoretical paradigms in light of their relationship to language in general, the human mind, L1 and L2, individual differences and the learning context, it seems that different theories are not simply different statements about the same phenomena, but more often than not differing statements regarding different phenomena (Oxford, 1995). Thus, while UG linguistics is more interested in the explanation of language itself, the cognitivist theories are primarily interested in the internal processes of the mind that enable learning, whereas the sociolinguistic perspective looks at language in its social use. We will therefore have to view the theories in light of the specific answers they provide to our very specific question: Can adult learners of English from various linguistic backgrounds eradicate grammatical errors in their L2 academic writing? The specific variables we are interested in are the learner’s age, the writing skills and the development of grammar. The development of grammar entails the identification of common errors and their intended decrease in frequency and number as a result of an intervention.

In order to answer the above question, it might be purposeful to look at the general issues at hand in any language learning situation. The process of learning has to do with the language learnt, the learner herself and the context, whether social or not (Levy, 1997b). The underlying notion in most SLA theories, be they linguistic, psychological, psycholinguistic, sociolinguistic or sociological, is a particular understanding of the ownership of language. Thus, Chomskyan linguistics places the ownership of language outside the human being, either as an individual or a society. Language is rather seen as ‘a separate entity because it has an independent existence unrelated to human production or use’ (Grabe & Kaplan, 1996: 176). Even though this linguistic approach claims that the language acquisition device (LAD), a special and uniquely human partition of the mind responsible for language learning, is inborn or genetic, it is not clear how it came into being. LAD seems given, a premise that is not only in accord with idealist philosophy, but which also suggests a creationist view of the human mind. The very word ‘genetic’, used to describe it, comes from the term genesis, which in its meaning includes the act of creation (Petkovic, 1984), rather than referring to a long process of evolution. A radically different view of language, supported by cognitivists, who see no difference between language learning and other problem-solving processes in the mind, is that language is owned by the individual in the form of neural pathways established through that individual’s unique mental activity, which only reflect a complex linguistic environment surrounding the individual (N. Ellis, 2001). The third view of language is that of social interactionists, functional linguists and sociolinguists (Long, 1996; Halliday, 1999; Givon, 1979) who believe that language is the property of a community, as its meaning – its raison d’être – is developed and negotiated in social interaction.

Determining whether the locus of language ownership is an individual consciousness, the society or the genes, seems crucial to answering our initial question: Can our adult NNS university students learn how to improve their erroneous academic English? Thus, if the view is that language is not subject to the will and consciousness of an individual or a group, but rather to the predetermined coding of the genes, then this potentially leads to a negative answer to our question, e.g. in case the genes are not programmed for SLA in adulthood. The answer is likely to be negative since our NNS students are still making errors despite the fact that their consciousness as well as the respective social group, i.e. the university, may be focused on the goal of their language learning progress. On the other hand, if the view is that language is owned by the individual, then this becomes an empowering impetus for the learner to persevere, as she can expect to be successful. Responsibility for the outcome is however entirely her own. Finally, if the view is that the language is owned by a social group, then learning will only be possible in social interaction, thus making the group responsible for the success of the individual. Let us now examine to what extent the research conducted on each of the three premises conforms to the hypotheses we have just expressed.

Universal Grammar (UG) and language learning

Let us first examine the success prognosis given to the adult L2 learner from the UG perspective of genes as language owners. As an explanation of language acquisition, which is not its primary purpose, Chomsky (1965) as the founder of UG claims that each child is equipped with a highly specialised language acquisition device (LAD) that does not serve any other cognitive purposes but language acquisition alone. This device in a later version becomes equated with the universal grammar. UG consists of principles which are universal to all human languages, e.g. that each phrase has a head, and the parameters that differ from language to language, e.g. the position of the head in a phrase. The way LAD works is by setting the language specific parameters when a certain language structure in L1 is encountered, presumably in early childhood (Manning, 1991). The main argument in favour of such a device is the apparently insufficient input a child receives in its language development phase. Thus, so the argument goes, as the child cannot learn the language by imitating its environment, it has to have an innate ability that eventually results in a very creative use of language, enabling the child to produce a number of sentences they have never heard before. With age, this ability to set the parameters decays, which does not raise much hope for adult learners of L2, including our EAP students.

Indeed, research influenced by UG has yielded four different hypotheses about second language learning (Mitchell & Myles, 1998; Doughty & Williams, 1998b) and only one of them seems really optimistic. This one postulates that when learning an L2, the learner has access to the same UG features (principles and parameters) responsible for their learning L1. The other three hypotheses range from total dismissal of UG availability over partial UG availability for L2 learning to indirect availability of UG via L1. The prognosis for any L2 learning outcomes without direct access to UG must be grim coming from this group of linguists, as we must remember that they do believe that LAD, sometimes equated with UG, is solely responsible for language acquisition.

The first UG-based SLA hypothesis, the one that allows for the availability of UG to adult learners, comes from Flynn (1996) and is the most recent development. The second one, involving total denial, was based on the observation of immigrants to the USA, who only seemed to attain nativelike proficiency in English if they arrived not later than the age of seven (Mitchell & Myles, 1998). Partial access hypothesis is based on the observation of partial success of L2 learners, while the indirect access hypothesis is associated with the notion of critical period, during which only at a young age does an L2 learner have a window of opportunity to access UG and be successful (Mitchell & Myles, 1998). Thereafter, L2 can only be accessed indirectly through L1 and native-like competence can therefore not be attained. According to UG theorists, indeed, adult learners do not seem to have fair chances of L2 acquisition, perhaps because, as we noticed, within the framework of this theory language is not viewed as being fully owned by the individual or social consciousness. Even though this theory does not promise much hope for our adult NNS students, it is, according to Gregg (2001), the only SLA theory based on a linguistic theoretical framework, which has its own merits.

It is in the context of nativism and UG that Stephen Krashen’s (1987) take on SLA is sometimes accounted for (see e.g. Cook, 1993 and Larsen-Freeman & Long, 1991). On the surface, it would seem that Krashen (1987) would have something in common with UG, since he evokes LAD as the aspect of mind responsible for language acquisition in general, but he does not explain its internal workings. Krashen’s (1987) general theory is based on five premises called ‘hypotheses’, even though Larsen-Freeman and Long (1991) rightly point out that they cannot really be considered hypotheses in the technical sense of the term, being neither falsifiable nor verifiable. The five ‘hypotheses’ are thus Acquisition-Learning Hypothesis, the Monitor Hypothesis, the Natural Order Hypothesis, the Input Hypothesis and the Affective Filter Hypothesis. We shall briefly outline each one in turn.

Krashen (1987) distinguishes between ‘acquisition’, which is a subconscio...

Table of contents

• Cover
• Title page
• Copyright
• Contents
• Acknowledgements
• Introduction
• 1 Can Another Language Be Learnt?
• 2 Where Does Research End and CALL Development Begin?
• 3 Why the Web?
• 4 Can Computers Correct Language Errors?
• 5 How to Develop an Artificially Intelligent Language Tutor?
• 6 How Does it Work?
• Conclusion
• Appendices
• Bibliography
• Index