|
|
Individualised Recommendations for Learning Strategy Use
ARIES Lab, Dept. of Computer Science,
University of Saskatchewan,
Saskatoon, Saskatchewan, S7N 5A9, Canada.
Abstract. This paper describes
LS-LS: a system to raise awareness of language learning strategies to help
students become more effective learners. The focus is the student model, which
contains representations of learning style and current strategy use:
information provided explicitly by the learner. LS-LS infers additional
strategies of potential interest to an individual, based on the contents of
their student model. It also suggests computational learning environments that
a student might find useful to practise these new strategies, based on
information provided by the (human) tutor about locally available software.
|
|
1 Introduction
Linton observes
that intelligent tutoring systems (ITS) are often judged by their ability to make
tutoring decisions for the learner, despite the fact that self-directed
learners actually possess a valuable skill [1]. Learner autonomy is also
important in foreign language learning [2]. Much research has indicated that
appropriate use of language learning strategies can contribute to autonomy and
success in learning a second language [3]. There are various definitions of
language learning strategies: some relate to conscious application of
techniques to help a learner [4]; others allow the possibility of unconscious
strategy use [5]. Kohonen states learners can be made aware of their strategy
use, and that they may modify it with 'conscious effort' [5].
Early work
suggested there are successful language learning strategies, and teaching these
strategies to less successful students might help them improve their
performance [6]. Later research found some unsuccessful learners actually use
many of the same strategies as more successful peers [7]. Such students need to
learn how to apply strategies appropriately.
Further, it is not the case that all good learners use the same strategies [8].
Strategy choice may depend to some extent on learning style [9]. It seems that
while tailored application of learning strategies is useful, there is no single
set of strategies appropriate for recommendation to all learners [10]. Indeed,
Oxford recommends "strategy training should be somewhat
individualized" [3]. However, this is difficult in the typical language
learning situation, where there is a single teacher working with a foreign
language class, for a limited time period.
A few
tutoring systems encouraging the use of a variety of language learning
strategies have been implemented [11,12], to foster the kind of self-direction
proposed by Linton [1] for an ITS. Nevertheless, these systems are tied to
their own contexts. Implementation of a more widely applicable system to foster
the acquisition of language learning strategies has also been undertaken [13].
However, this requires expensive hardware often not available in student
Language Centres.
This paper
introduces LS-LS (learning style–learning strategies): an environment to raise
student awareness of language learning strategies, to help them become more
autonomous learners. LS-LS recommends potentially useful additional strategies
to an individual, according to their learning style and current strategy use,
and suggests ways students may practise these new strategies. It is designed
primarily for use in contexts where resources are restricted, and where individualised,
teacher-led strategy training programmes are infeasible. LS-LS runs on most
Macintosh computers.
LS-LS is
centred around a student model constructed with the help of explicit student
contributions, following a recent trend in student modelling [14-19]. In
addition to providing information for the student model, in LS-LS this approach
has the function of promoting learner reflection on both the student's own
specific approaches to learning, and on different ways of learning in general.
Thus, even before receiving strategy suggestions, students are thinking about
their learning.
LS-LS is
unusual in the sense that the student model is not part of a larger tutoring
system. The suggestions made by LS-LS refer in the main to activities outside the LS-LS system. Some of these
recommendations will include suggestions for computer-based interaction to
practise strategy application. This requires some additional information about
the local situation, which must be provided by the teacher.
2 Theoretical Basis of LS-LS
LS-LS aims to help learners become more self-directed by introducing
new learning strategies which fit with their learning style and current
strategy use. These two types of information form the student model. The
initial representations are provided directly by the student, by indicating
which aspects of learning style descriptions are applicable to their own
learning, and which learning strategy descriptions apply.
Various learning style inventories
have been developed [20-22]. That used in LS-LS is adapted from the Myers-Briggs Type Indicator (MBTI) [22],
as the MBTI was found to correlate with students' choice of language learning
strategies [9]. The MBTI is based on Jung's theory of psychological types [23].
It describes people in terms of four characteristics:
introversion/extraversion; sensing/intuitive; thinking/feeling;
perceptive/judging. However, the MBTI questionnaire is extensive, and in the
context of LS-LS learners may not be prepared to spend much time. Therefore a
much simplified adaptation is used, whereby students select amongst brief
descriptions of learning style components [24]. To compensate for the lack of
detail, students may indicate that two poles (e.g. thinking/feeling) are both
applicable. This results in a less precise learning style descriptor, but it
does ensure that students are not forced into providing information about which
they are unsure. Indeed, lack of preference in any of the four descriptor pairs
is not necessarily negative. It may indicate that the student does not lie at
either extreme of the continuum: their individual learning style may encompass
both aspects of the paired descriptors. Allowing this possibility in LS-LS
ensures that potentially useful learning strategy suggestions are not
suppressed by the system as a result of forced selection of one aspect of
learning style over another.
The
language learning strategy classification system used in LS-LS is adapted from
Oxford's Strategy Inventory for Language
Learning (SILL) [25]. The SILL has been used extensively by researchers,
and has been found to have high validity, reliability and utility [26]. It is
administered to students as a questionnaire, and measures the frequency with
which a student uses memory, cognitive, compensation, metacognitive,
affective and social language learning strategies, giving the result: low, medium
or high use, for each category. As with learning style, in LS-LS students
identify their current strategy use from short strategy descriptions. As
results have shown, information about individuals' use of strategies from
different strategy classes can be very useful for research purposes. However,
this kind of information is less meaningful to learners. For example, what does
it mean to a student to be told that they have 'a medium use of compensation
strategies'? Thus LS-LS requires additional information to be overlaid on
Oxford's classification scheme. This is provided by a Strategy Similarity Measure (based on [11]). This similarity
measure is a theoretical construct indicating conceptual similarities amongst
strategies. This allows new strategies to be introduced with reference to
strategies already used, so suggestions are more meaningful to learners. It
also enables strategies to be considered individually, rather than only in the
six strategy groups identified by Oxford.
This
approach requires learners to be able to identify their current strategy use,
as LS-LS obtains initial representations by self-report. A previous study found
adults were indeed able to identify their strategy use in a manner similar to
that used to acquire the LS-LS student model. Furthermore, most were interested
in doing so [27].
In summary,
LS-LS is based on four areas of previous research:
• learning
strategy classification [25];
• the ability
of students to identify their learning strategy use through self-report in a
computational environment [27];
• relationships
between learning style and strategy choice [9];
• conceptual
similarities between learning strategies [11].
The first area concerns the representations for the student model. The
second relates to the method of obtaining this information. Points 3 and 4 form
the knowledge base: representations used by LS-LS to infer appropriate strategies
to recommend to an individual, according to the contents of their student
model.
3 Individualised Suggestions of Learning
Strategies
As stated above,
to build the LS-LS student model learners provide information about their
learning style and currently used learning strategies. This is accomplished by
viewing descriptions (for an example see part 3 of Figure 1), and selecting the
options which apply. The resulting contents for the student model are
illustrated in Table 1.
Table 1
shows the student model of an adult male Mainland Chinese learner of English
(advanced level), studying English in the U.K. This is presented in full to
illustrate a plausible range of learning strategies an individual may use, and
the kinds of strategy that might be suggested to others. This includes 26 of
the 62 strategies in Oxford's classification [25]. The student model
representations are in Prolog:
learning_style([Style_Components]).
learning_style([extravert, sensing, thinking, perceptive]).
learning_strategies(Strategy_Group,
[Strategy_List]).
learning_strategies(cognitive, [skimming,
analysing_expressions, translation, notes]).
It can be seen
that the student model is quite straightforward, both in terms of its contents
and, as discussed above, in the model acquisition process.
Table 1. Representations in the LS-LS student model
|
Learning
Style |
Strategy Group |
Strategy Name |
|
extravert sensing thinking perceptive |
cognitive |
skimming, analysing expressions, translation,
notes. |
|
metacognitive |
overviewing/linking with known, delaying speech to
focus on listening, setting goals, planning, seeking practice opportunities,
self-monitoring, self-evaluation. |
|
|
memory |
grouping information, associating/elaborating,
structured reviewing, mechanical techniques. |
|
|
compensation |
using linguistic cues, language switching, getting
help, circumlocution/synonyms. |
|
|
social |
requesting correction, cooperation with peers,
cultural understanding, awareness of others' feelings. |
|
|
affective |
using relaxation/deep breathing/meditation, using
music to relax, using a checklist about feelings. |
Once
representations for the student model are completed, students may receive suggestions
of additional strategies that may be useful. Suggested strategies must fulfil
two conditions: (1) they may not conflict with the student's learning style;
(2) they must have something in common with at least one used strategy. The
former is based on Ehrman and Oxford's finding that learning style appears to
influence strategy choice [9]. LS-LS therefore contains representations of
permitted learning style–learning strategy links. For example, an ISTJ
(Introvert, Sensing, Thinking, Judging) learner will be primarily recommended
strategies from the groups metacognitive, cognitive and memory. This is because
Ehrman and Oxford's data suggested Introverts and Thinkers are generally
uncomfortable with social strategies; Sensers and Judgers disliked compensation
strategies; Introverts did not like affective strategies. On the positive side:
Introverts were very much in favour of metacognitive strategies; Sensers liked
cognitive, metacognitive, and in particular, memory strategies; Thinkers were
very positive about cognitive strategies, and also liked metacognitive
strategies; Judgers liked social, and especially metacognitive strategies.
Point 2
above fulfils the requirement that strategy recommendations be made with
reference to something the learner can readily understand. This is accomplished
through a database of strategy links based on the strategy similarity measure. Table 2 shows excerpts from the
database of strategy links in three of the six strategy groups.
The first two examples of Table 2 indicate that there is some similarity
between the concepts of the memory strategies representing sounds in memory and imagery. Thus, a student who uses one of these strategies but not
the other, will probably appreciate the potential utility of the new strategy
due to the similarity of the function of the pair.
The next two entries in Table 2, analysing expressions and contrastive analysis, show a similar
bidirectional relationship, but in the cognitive group. The fifth entry, also
concerning cognitive strategies, illustrates how the suggestion of a new
strategy may be based on more than one currently used strategy. If a student
uses contrastive analysis and deduction, but not analysing expressions, the latter will be suggested with reference
to both contrastive analysis and deduction (assuming there are no
objections from the learning style component). The link between deduction and analysing expressions is also bidirectional, as indicated by entry
number 6, as is the link between contrastive
analysis and deduction (not
shown).
Table 2. Excerpt
from database of strategy links
|
Used
Strategy |
Strategy
Suggestion |
|
mem: representing sounds in memory |
mem: imagery |
|
mem: imagery |
mem: representing sounds in memory |
|
cog: analysing expressions |
cog: contrastive analysis |
|
cog: contrastive analysis |
cog: analysing expressions |
|
cog: deduction |
cog: analysing expressions |
|
cog: analysing expressions |
cog: deduction |
|
cog: making notes |
mem: grouping |
|
comp: avoiding communication |
comp: selecting the topic |
|
comp: avoiding communication |
comp: adjusting the message |
The next
entry illustrates that links, and hence recommendations, occur not only between
strategies within a strategy group, but also occur across groups. Making notes
is a common cognitive strategy. However, some students do not organise their
notes effectively. For such learners, the memory strategy grouping may be suggested.
The last
two entries show that a single strategy may be used as support for recommending
more than one new strategy. This example also illustrates that links are not
always bidirectional. Some students avoid
communication, a compensation strategy, when a topic is problematic.
Alternatives may be suggested, e.g. selecting
the topic or adjusting the message.
However, the reverse does not occur: a student who uses one or both of these
will not receive the suggestion to avoid
communication.
1 You already use visual imagery to help you remember
vocabulary. There may be times when imagery is difficult. This might occur, for
example, when you need to learn abstract words.
2 You may find using sound a good substitute for imagery, as
these strategies have the same function of using the senses to learn
vocabulary. They are both memory strategies.
3 Representing sounds in
memory involves creating an association between new
and known material by using sound. For example, there may be a word in your
native language that sounds similar to the new word you are trying to remember.
Or the new word may sound similar to another word that you already know in the
target language.
Fig.
1. Example of a
strategy recommendation
Figure 1
illustrates a strategy recommendation (generated from templates). It first
refers to a strategy the student already uses. It then links this to the new
strategy. Finally, the new strategy is described. Note that there is no
implication that the suggested strategy should replace any strategy already used. It is simply stated that it
might be a useful strategy when it is difficult to use an existing one. It is
up to the learner to decide whether the new strategy is, in fact, more helpful
than any they currently apply in a particular situation.
It may be
that a new strategy is not suitable: it will not always hold that a visual learner
will benefit from sound to the extent other learners might. Hence words
like 'might' and 'may' in the recommendation. However, recall only those
strategies which do not conflict with learning style pass the 'strategy
suggestion threshold'.
Once new
strategies have been experienced, the learner may return to LS-LS for further
suggestions, which can take recently acquired strategies into account.
4 Recommending Computer-Based Environments
Thus far discussion
has centred on the first set of suggestions received by students: general
recommendations which may be applicable to a variety of language learning
contexts. The second set of proposals concerns these new strategies, but
includes suggestions of specific computer assisted language learning (CALL)
software available at their institution, where some of these
strategies can be practised.
Jones
explains how the institutional context is a major factor in the design of the
majority of CALL programs [28]:
Most
CALL programs are developed at universities… CALL software is usually intended
for a particular course at a particular institution with a particular sort of
student with particular needs. This exact matching of needs is what makes
computer-based courseware so successful for its intended audience, but which
can impair its marketability. [28]
This implies that
LS-LS would be severely restricting its applicability if it were not to take
into account potentially numerous in-house developments when suggesting CALL
activities for a student. Because much of the courseware may have been
developed by language teachers, in many situations this will not include
intelligent CALL. Nevertheless, because of its design focus on local students,
and its recommendation by LS-LS to students because of the potential for them
to practise the application of learning strategies which are appropriate for them, any lack of individualisation in the CALL software will
be less crucial. The 'intelligence' in this approach is found in LS-LS's
inferring suitable programs to recommend, depending on characteristics
(learning style and current strategy use) of the individual learner.
Including
local information requires input from the local tutor. It must be assumed that
the tutor is aware of the CALL options available at their institution as,
indeed, a good teacher should be. However, it is not assumed that tutors will
already be aware of Oxford's language learning strategy classification system:
they may learn about these strategies by reviewing the strategy descriptions in
LS-LS (as does the student). Figure 2 shows how tutors provide information
about available CALL opportunities. This method of inputting information covers
a range of CALL types: e.g. concordancers (cognitive–recognising forms and
patterns, analysing expressions, contrastive analysis, deductive reasoning,
resourcing); traditional drills (cognitive–repetition, recognising forms
and patterns, deductive reasoning); foreign language chat rooms (cognitive–practising
naturalistically; compensation–selecting the topic, adjusting the
message, coining words, circumlocution/synonyms; metacognitive–seeking
practice; affective–risk-taking; social–cooperation). It can
be seen that for a single strategy, there might be several different kinds of
program that may be used to experience it. Therefore learners will often be
able to select the kind of software they prefer, or use more than one type of
CALL to consolidate the use of their new skill.
Strategies
expected to be useful in many implementations are listed for two reasons: (1)
to make it easy for teachers to input required information; and (2) to
encourage tutors to consider the applicability of the most likely strategies
(i.e. not overlook them). Further strategies may be entered if the local
situation encompasses them. Space is also available to describe how learning strategies may be applied.
Specific instances of CALL can also be referred to. Recommendations of CALL
environments are presented to students exactly as described by the tutor: the
relevant strategies are listed, followed by the teacher's textual description.
A few examples of CALL programs are given below, to demonstrate the value of
this practice facility.
Fig.
2. Tutor input
of information about available CALL options
Milton's
Electronic Learning and Production Environment aims to help students write
appropriately by using a concordancer together with error recognition tasks, a
hypertext grammar and databases of underused phrases [29]. Thus it has the potential
of encouraging the above-mentioned strategies connected with concordancing, but
the resourcing and deduction opportunities are broader than with
many concordancers.
In the
context of translation, Metatext is a HyperCard development which has links from
the main card containing the source text to datacards where learners may view
or send information [30]. Therefore, in addition to improving translation
skills, it can be used to explicitly practise the cognitive strategies of noting
and resourcing; and the memory strategies of grouping and using
mechanical techniques.
Sawada et
al describe a system with which students may practise writing Japanese Kanji
characters and phrases [31]. They can also test their sequencing of strokes in
a character, to help them understand the structure of Kanji patterns. Thus
learners may extensively practise the cognitive strategy of formally
practising with writing systems.
Some ITSs
contain a model of the target language rules, and also the equivalent rules
from a learner's native language, allowing explicit reference to both languages
during an interaction [32-35]. Such systems provide opportunities for learners
to consider the cognitive strategies of contrastive analysis and language
transfer.
Despite the
potential for students to practise a variety of strategies in CALL
environments, it is clear they may need guidance on how this might be
accomplished–although systems have been designed to foster such skills, they
are for the most part not designed with the aim of explicitly tutoring the
strategies concerned. Chapelle and Mizuno recognize that much CALL assumes
learners are already able to regulate their learning effectively, whereas, in
fact, they often do not use the most appropriate strategies [36]. Hence the
importance of allowing tutors the space to describe for students, the use of
these strategies in the particular CALL contexts (see Figure 2). An advantage
students who have used LS-LS might have when using these CALL systems, is that
they are by then already aware of the variety of strategies that exist.
Using the
student from Table 1, the learner was identified as having the personality
attributes ESTP (Extravert, Sensing, Thinking, Perceptive). The rules for
generating the sequence of strategy presentations for student selection of used
strategies are based on these attributes. Mapping personality attributes to the
strategy presentation sequence ensures that learners identify first the
strategies they are most likely to use, in case they choose not to complete the
full sequence of strategy identification. In our example, cognitive strategies
were presented first, as the style components STP each view cognitive
strategies positively (and E is neutral) [9]. Second came metacognitive
strategies, typically viewed positively by EST, but negatively by P, and so on.
Strategy
suggestions were presented, ranked according to personality attributes and
strategy similarity measure. In our example, contrastive analysis was
recommended early, because of the personality component Perceptive, and the
used strategies requesting correction, using linguistic clues, translation
and analysing expressions. In contrast, repetition was suggested
based only on the descriptor Sensing, and came last. This recommendation
according to constraints imposed by learning style and the strategy similarity
measure, and the ranking of suggestions, ensures that strategies are presented
in a sequence reflecting understandability and relevance for the individual.
CALL
programs are then suggested, which also fit the constraints of learning style
and the strategy similarity measure. These are similarly ranked according to
expected utility. For our learner, a bilingual concordancer might be useful,
since it allows practice of contrastive analysis (a strongly suggested
strategy) and also deduction (recommended based on one personality
attribute and two used strategies). Lower on the list come drills to practise deduction
and repetition. The range of strategy suggestions, and hence CALL
suggestions, are likely to vary since even in a small sample of students (5),
total suggested strategies ranged from 6 to 16 [24].
An
interesting situation has occurred, whereby an intelligent learning environment
(LS-LS) will be recommending largely 'unintelligent' programs to students.
LS-LS starts from a quite simple student model, performing some complex
inferencing [24], to then recommending less adaptive systems. Although these
less flexible programs are often criticised for their inability to take into
account learner differences, when recommended by LS-LS, such differences have
already been catered for.
5 Summary
LS-LS aims to raise student
awareness of ways to make their learning more effective, by fostering learner
autonomy in a manner that suits their learning style, and is easily
understandable according to their current strategies. This occurs as in Figure
3.
Fig. 3. CALL
recommendations
LS-LS
prompts students for information about their learning style and approaches to
learning, offering descriptions from which they select those aspects they
believe apply to their learning. The resulting representations form the two
components of the student model. LS-LS also contains a learning strategy
database: one part containing strategy descriptions; a second detailing
information about strategies typically liked and disliked by learners with
different learning styles; a third measuring similarities between pairs of
strategies. LS-LS compares information from the student model to the
constraints implied in the strategy database (parts 2 and 3), and makes
recommendations of potentially helpful strategies to an individual. These
recommendations are general: they describe strategies, with examples, but no
specific learning materials are suggested. Strategy suggestions are fed back into
the student model to be used should the learner later return for a further
interaction with LS-LS.
A second
database contains representations relating to other CALL systems. This has two
parts: a general one detailing kinds of CALL program that can be used to
experience different learning strategies; and a specific part describing
software available locally. This second part is input by the tutor. LS-LS
combines information about strategies it suggested with information in the CALL
database, to suggest specific CALL programs a learner might access to try out
recommended strategies.
References
1. Linton, F. (1997) Learning to Learn from an
ITS, in B. du Boulay & R. Mizoguchi (eds), Artificial Intelligence in Education, IOS Press, Amsterdam.
2. Broady, E. & Kenning, M-M. Eds. (1996) Promoting Learner Autonomy in University
Language Teaching, AFLS/CILT, London.
3. Oxford, R. (1994) Language Learning
Strategies: An Update, ERIC Clearinghouse
on Languages and Linguistics.
4. Fox, J. & Matthews, C. (1991) Learner
Strategies and Learner Needs in the Design of CALL Help Systems, Proceedings of EUROCALL, Helsinki,
127-132.
5. Kohonen, V. (1992) Experiential Language
Learning, in D. Nunan (ed), Collaborative
Language Learning and Teaching, Cambridge University Press, 14-39.
6. Rubin, J. (1975) What the 'Good Language
Learner' Can Teach Us, TESOL Quarterly
9(1).
7. Vann, R.J. & Abraham, R.G. (1990)
Strategies of Unsuccessful Language Learners, TESOL Quarterly 24(2), 177-198.
8. Stevick, E.W. (1989) Success with Foreign Languages: Seven who achieved it and what worked
for them, Prentice Hall, London.
9. Ehrman, M. & Oxford, R. (1990) Adult
Language Learning Styles and Strategies in an Intensive Training Setting, The Modern Language Journal 74(3),
311-327.
10. Gillette, B. (1987) Two Successful
Language Learners, in C. Faerch & G. Kasper (eds), Introspection in Second Language Research, Multilingual Matters,
Clevedon, Philadelphia.
11. Bull, S. (1997) Promoting Effective
Learning Strategy Use in CALL, Computer
Assisted Language Learning 10(1), 3-39.
12. Meskill, C. (1991) Language Learning
Strategies Advice: A Study on the Effects of On-Line Messaging, System 19(3), 277-287.
13. Rubin, J. (1996) Using Multimedia for
Learner Strategy Instruction, in R. Oxford (ed), Language Learning Strategies around the World: Cross-Cultural
Perspectives, Technical Report 13, Second Language Teaching and Curriculum
Center, University of Hawaii.
14. Beck, J., Stern, M. & Woolf, B.P.
(1997) Cooperative Student Models, in B. du Boulay & R. Mizoguchi (eds) Artificial Intelligence in Education,
IOS Press, Amsterdam, 127-134.
15. Bull, S. & Pain, H. (1995)
"Did I say what I think I said, and do you agree with me?":
Inspecting and Questioning the Student Model, in J. Greer (ed), Proceedings of World Conference on
Artificial Intelligence in Education, AACE, 501-508.
16. Dimitrova, V., Self, J. & Brna,
P. (1999) The Interactive Maintenance of Open Learner Models, in S.P. Lajoie
& M. Vivet (eds), Artificial
Intelligence in Education, IOS Press.
17. Kay, J. (1999) Learner Control,
submitted for publication.
18. Morales, R., Pain, H. & Conlon,
T. (1999) From Behaviour to Understandable Presentation of Learner Models: A
Case Study, Proceedings of Workshop on
Open, Interactive and Other Overt Approaches to Learner Modelling, AIED'99,
Le Mans, France, 15-24.
19. Specht, M., Weber, G. & Schöch,
V. (1997) ADI: Ein adaptiver Informations-und Lehragent im WWW, in R. Schäfer
& M. Bauer (eds) ABIS-97: 5
GI-Workshop, Adaptivität und Benutzermodellierung in interaktiven
Softwaresystemen, Universität des Saarlandes, 53-60.
20. Felder, R.M. (1993) Reaching the
Second Tier: Learning and Teaching Styles in College Science Education, Journal of College Science Teaching
23(5), 286-290.
21. Kolb, D. (1984) Experiential Learning: Experience as the Source of Learning and
Development, Prentice Hall, Englewood Cliffs NJ.
22. Myers, I.B. & McCaulley, M.H.
(1985) Manual: A Guide to the Development
and Use of the Myers-Briggs Type Indicator, Consulting Psychologists Press,
Palo Alto.
23. Jung, C.G. (1971) Psychological Types, Princeton
University Press.
24. Bull, S. & Ma, Y. (submitted)
Raising Learner Awareness of Language Learning Strategies in Situations of
Limited Resources, submitted for publication.
25. Oxford, R.L. (1990) Language Learning Strategies: what every
teacher should know, Heinle and Heinle Publishers, Boston MA.
26. Oxford, R.L. & Burry-Stock, J.A.
(1995) Assessing the Use of Language Learning Strategies Worldwide with the
ESL/EFL Version of the Strategy Inventory for Language Learning (SILL), System 23(1), 1-23.
27. Bull, S., Pain, H. & Brna, P.
(1993) Student Modelling in an Intelligent Computer Assisted Language Learning
System: Language Transfer and Learning Strategies, in T-W. Chan (ed), Proceedings of International Conference on
Computers in Education, Taiwan, 121-126.
28. Jones, C. (1998) Multimedia and
Vocabulary Learning: A Marriage Made in Heaven?, in K. Cameron (ed), Multimedia CALL: Theory and Practice,
Elm Bank Publications, Exeter.
29. Milton, J. (1998) Exploiting L1 and
Interlanguage Corpora in the Design of an Electronic Language Learning and
Production Environment, in S. Granger (ed), Learner
English on Computer, Addison Wesley Longman Limited, London, 186-198.
30. Gillespie, J.H. & Gray, B. (1992)
HyperCard and the Development of Translation and Vocabulary Skills, Computer Assisted Language Learning
5(1-2), 3-11.
31. Sawada, S., Higashigawa, L., Bandoh,
H. & Nakagawa, M. (1997) A CAI System for Overseas Students to Learn
Reading and Writing of Japanese Text, in B. du Boulay & R. Mizoguchi (eds) Artificial Intelligence in Education,
IOS Press, Amsterdam, 653-655.
32. Bull, S. (1995) Handling Native and
Non-Native Language Transfer in CALL: Theory and Practice, in R. Wakely, A.
Barker, D. Frier, P. Graves & Y. Suleiman (eds), Language Teaching and Learning in Higher Education: Issues and
Perspectives, CILT, London.
33. Catt, M. & Hirst, G. (1990), An
Intelligent CALI System for Grammatical Error Diagnosis, Computer Assisted Language Learning 3, 3-26.
34. Schuster, E. (1986) The Role of
Native Grammars in Correcting Errors in Second Language Learning, Computational Intelligence 2, 93-98.
35. Wang, Y. & Garigliano, R. (1995),
Empirical Studies and Intelligent Language Tutoring, Instructional Science 10, 225-240.
36. Chapelle, C. & Mizuno, S. (1989)
Students' Strategies with Learner-Controlled CALL, CALICO Journal, Dec 1989,
25-47.