- 33 -
LEARNING ENGLISH WITH IT'S ENLSH
J.G.M. Jspes, G. Knselar, W.A.M. Kok
Depment of Educational Sciences, Utrecht University
Miling adress: rof. dr. Gellof Kanselar
Depment of ducatioal Sciences
Heidelbergln 2, 3584 CS Utrecht, oe Netherlnds
31.30.532352
31.30.534893, Fax:
Tel:
Emil:
[email protected]
ABSTRACT:
uis paper descibes a new Computer Assisted
Instruction program or oreign language (english)
teaching based on a communicative approach.
Due to the greater mobility and the closer ties in
the Europen community, oreign languages have
bcome an even more important subject in the
school cuiculum. Bsed on the results of a study in
1989 [3], a new Computer Assisted Instruction
program or oreign language teachng (English)
has been deve!oped. Main eaures of his program
re:he communicative approach, a 70.000 word
dictionay, multimedia nd a syntactic prser.
BACKGROUNDS IN THE
DEVELOPMENT OF IT'S ENGLISH
To a lrge extent oreign lnguage teaching is
directed towrds
communicative
goals.
Scio
cultural developments (mass media, holidays abroad
etc.), scientific developments in teaching
mehodology. the use of new media, linguistics nd
psycholinguistics have led to a more funciona1
approach to foreign lnguage teaching.
he traditional grammar-ranslation method, in
which vocabulary was Iened mostly y
memorizing L/LI lnguage pairs nd in which
grammar was taught as a system of rules, did not
directly lead to luent orl proficiency.
In developing the CAI progrme we chose the
communicative approach to foreign language
teaching. In this approach language is viewed as an
instrument of human behaviour. he behavioral
"siuation" is carefully selectd in teaching. oe
linguistic context, he siuation, spkers' rol es and
types of texts re ll taken into creul consideration
in the communicative approach. Linguistic unction
is seen as more important than Iinguistic om. oe
structuring of the leaning mateials is irst and
oremost detenined by communicative needs.
When designing he lening materials or the
CAI programme we used Neuner' s typology as a
stating point for he constuction of exercises. he
ree types of exercises in the programme, including
receptive, reproductive (cloze texts) and productive
(ree sentence input) exercises can e placed on this
continuum.
BASIC AIMS AND ASSUMPTIONS
he aim of the projct is to create a computer
assisted leaning enviroment for communicative
oreign language teaching [IJ. uis fom of oreign
language teaching is oiented towrds an active and
pssive command of a foreign language in everday
situations. lt avoids trnslation as the oly or the
most impont means of conveying meaning.
Vocabuly is acquired on he basis of context
speciic sentences and menings in the oreign
language, to ensure the development of a Janguage
sysem that is seprate rom the Iener's native
language. In addition to the syntactic and
mophologica1 characteristics of words, heir
pronunciation is of great impotance, as is sentence
intonation. he sound of a new word is a basic
element in assigning meaning to that word. In order
to efct this a reat number of words and sentences
were stored on CD-ROM. oe aim of
comicative language teaching is mainly
achieved by having students use the non-native
language productively. In order to achieve rndom
lnguage-input a sentence paser was developed or
the program. uis mkes use of he grammatica!
infomation contaned in the D-ROM diciony.
- 34 -
he combination of dictionay, phonetic word
representations and parser mkes it possible to
create diferent receptive and productive exercises
in this interactive lening environment.
he sotwre progm T'S ENGLISH is
primarily aimed at teaching vocabuly. However,
as an intelligent tutorial system are werking on
making it usable in the teaching of writing. In he
following sections he program itself will e
descibed.
A short history of lt's English
During a previous sudy [3] a program clled
"PESIGE" written in SD-PROLOG and running
on MSDOS ws created. This progrm allowed,
among ether things, students to enter their own text.
his limited dictiony (± 2000 words) proved toa
small. Sudent aften used words not contained in
this small dictionry.
To overcome he problem of the limited
dictionay we used a 70.000 word dictiony that
was made avilable to us. he size of this dictionay
however and the addition of sound capabilities to
the progam made it inevitable to use anoher
storage medium. At the time the CDROM-XA
stndard was the most adequate choice. his
resulted in a CDROM-XA disk which held bah the
dictionay daabases and a great number of
soundiles. This version of the program was tested
in our groups of students in secondy education
[7].
he MSDOS version of the progrm had o
major drawbacks: the limited aount of memory
available to he program and the limited capabiliies
of the 80*25 character screen. In order to remove
memoy limitations and improve on the user
inteface a new version of the program is eing
developed. îe new version uns under Microsot
Windows version 3.1 in ehanced mode. his
version was implemented in Prolog-2 or Windows,
with some extensions in TurboPascal for windows.
his ransition removed he memoy problem nd
improved the user-intelace.
not
The CDROM-XA standard or sound however is
suppoted by the windows multimedia
environment. Conversion of the soundfiles to he
standard multimedia omat of Windows (.WAV
files) is curently in progress.
DESCRPTION OF THE 'IT'S
ENGLISH' PROGRM.
A computer program hat connects with the above
basic aims and assumptions should enable the user
to reieve semntic deinitions and context
sentences in English, as well as synonms,
antonyms, pronunciation and the syntactic and
mophological properties of words. In addition such
a program should e capable of evaluating the
user's use of lnguage to enable him or her to app!y
the vocabulay in (re)poductive exercises. o
achieve all this the program consists of several
components:
l. An automatic sentence parser which checks
random Jnguage input or syntactic,
orthographic or inlectional etc. corectness.
2. A CD-ROM database, which contains
deiitions and context sentences in everyday
English, as well as the grammatical properties
and synonyms and ntonyms of over 70,000
words.
3. A CD-ROM database with the digitized
pronunciation of over 5000 words, 1200 context
sentences for the presentation of phonetic word
representations and English intonation.
4. An exercise generator, which generates
exercises or passive, active and reproductive
Janguage use.
5. A eedback-generator, which uses the domain
input of the parser to reveal and remedy lerner
erors in (re)productive exercises.
The Collins Cobuild English Language
Dicionry [2] was chosen as a basis or the lexica!
CD-ROM database. his dictiony meets the
following requirements. lt was compiled or 'non
naive lemers of English'. Each word is
accompnied by a semantic defition in simple
English, and a number of authentic, explanatory
citations. In addition. each word meaning is
accompanied by systematic and moderately detailed
gammatica! infomation, for use by the parser. his
direct connection etween grammatica} inonation
and he semantic section makes he meaning of
evey word hat is recognized by the parser instantly
avilable for reieval, as well as a number of
context sentences and ether relevant infomation
such as synonyms, ntonyms and superordinate
- 35 -
terms. One inherent restriction in the application is
that the parser's "power", that is, the number of
English structures that can be automatically parsed,
is limited by the grammatica! information available
in COBUILD. An important advantage of using
such an extensive dictionary is the availability of
information on nearly all English words. In
previous studies [3] we found that in productive
exercises with limited dictionaries (2000 words)
pupils often use words that are not in the selected
subset of English.
CD-ROM AS A STORAGE MEDIUM.
Until recently it had not been possible in computeraided education to devote much attention to foreignlanguage pronunciation. in view of the large
amounts of extemal memory involved. To a large
extent the arrival of CD-ROM has removed Ibis
restriction. In this project CD-ROM is used as a
storage medium bath for digitalized pronunciation
and the lexica! database based on Collins Cobuild
Dictionary. The choice of the Extended Architecture
(XA) standard for the CD-ROM in our project is
exclusively related to the audio-component.
The most important considerations in the choice
of the XA-standard for the audio-component were
the following:
1.
The sound-quality had to be optima!. In
particular the absence in speech of the higher
sound-frequencies (3500 - 11000 Hz) makes it
very difficult to distinguish the pronunciations
of certain words in English, whereas more or
less subtle sound differences in fact signa!
differences in meaning. Tuis is especially
important in speech from female speakers,
because !heir speech is naturally richer in higher
frequencies.
2.
In the acqms1t10n of a goed foreignlanguage pronunciation it is of importance for
learners that the phonetic representations of
words should be presented not only in isolation
but also in sentential contexts. The learner
should experience the pronunciation of words as
being connected with and influenced by !heir
contextual position as well as by the intonation
of the sentence.
PROGRAM EXERCISES.
The exercise generator.
The generator is capable of generating different
types of exercises on the basis of texts stored in the
database. Exercises are been developed for the
program which range from reading exercises to free
writing. These exercises follow a route from
receptive via reproductive to productive language
use, and may be completed in a teacher- or learnerdetermined sequence. Three types of exercises have
been developed so far in the making of the program.
These are prototypical of the type of application that
is possible with such programs. In developing these
exercises we have followed Neuner [4], who
distinguishes three main phases in an exercise
sequence that leads to communicative 1anguage use.
As a result of the newly-developed parser, adding
exercises in more elaborate versions of the program
is a relatively simple matter.
- 36 -
Figure 1: A reading exercise.
?,;':,; ••
=
flle
IT'S Fnglish
.Exerclses
Yldeo
M1111
(',nhmld · read
s1ory
セ@
Windows
D:IENGLISH\EXERCISE\JJ.STO
[8J Sound
fZIMeaning
l8J Conte><t
Thls Isa new story.
lts short but enough to demonstrate the mos,
44@11®1
features.
l8l Synonym
fZI Antonym
!8l Superord.
l8lwordforms
Something that is essential is considered to be
extremety important or e>d:remely necessary
Deategory
l8l Paragraph
fora particular sltuatlon or activity.
e.g.
lt Is essentlal to set yourtargets reanstlcalty _
e.g.
e.g.
e.g.
Land Is essentlal for food and for werk ...
ft is absolutety essential thatyou continue ...
Qualifications are not essential. but some
previous experience is useful __ _
A degree ls by no means an essentlal
e.g.
Joumallstlc qualtflcatlon.
Synonym
A.
vltal
Reading texts (receptive use).
A fragment of a story is selected by the student or
teacher from the text database. Learners are free to
retrieve a definition of each text-word in English, or
the context sentence that accompanies each word.
As soon as a text has appeared on the screen, the
user can click on any word on the screen to search
for the word in the lexica! database. Definitions and
contexts are retrieved and presented. If the selected
word occurs m the sound-database that
soundfragment will be presented to the student. The
first meaning paragraph from the dictionary will be
shown. Appropriate synonyms, antonyms and
superordinate terms will be displayed. The user can
select the type(s) of information by enabling options
in the box in the upperleft corner of the screen The
NEXT and PREVIOUS buttons enable the user to
browse through the available information (see figure
1). It will also be possible for the teacher to mark a
word in advance to indicate that the leamer should
call up information about that word later. In this
process word-specific feedback can be provided by
the teacher.
B.
Texts with blanks (reproductive use)
In the creation of texts with blanks one or more
properties of a word, all of which are recorded in
the parser's property-list, may serve as a criterion
for the omission of that word from the exercise, or
for its inclusion in it. For example, texts could be
created without any prepositions or adjectives. The
decision to leave out all nouns, verbs, prepositions
or adjectives can be taken by the teacher or the
ieamer. Sentence constituents such as verb phrases
can also be left out. This type of filling-in exercise
is clearly superior to the cloze-texts of current
prevalence, in which only every n-th word may be
omitted. Since the database can be put into
operation in the analysis of the leamers' answers,
feedback can be provided which is aimed at various
kinds of errors. This makes IT' S ENGLISH
different from traditional programs, in which every
erroneous answer, plus the accompanying feedback,
- 37 -
Figure 2: A gapfiling exercise
セ@
IT'S fnqhsh
Bie
fxerclses
,Çobulld
farser ~deo
Ml(J,:
Coln11ld
fill g;,ip~ m story
Windows
A _ _ ol 11orniilhing
Is II putlcuhir
part of lt
or ch11r11dcrlstlc
th11t lt has , whlch you notlce
,
bcaiuac lt 11ccms lmportzmt or lnt!:rcsUng
-
Thlslsanew_.
lts short but enough to demonstrate the most essentlal
has to be anticipated and dealt with. The type of
feedback that is provided by IT' S ENGLISH for
filling-in exercises can be of the following kind: the
typed-in answer is synonymous with a superordinate
concept; the word class is correct but the actual
word is not; information about possible 'typing
errors' which is based on an a]gorithm; information
about conjugation and other grammatica!
characteristics. It is also possible to provide help by
giving a context sentence provided by the dictionary
from which the target word has also been omitted
(see figure 2)
C.
Random text-input (productive use).
This type of exercise allows the learner to type in a
story. The parser checks each sentence for
grammatica] correctness. Parsing of natura!
language is quite a complicated process. The
current version of the parser knows over 400 rules
to connect words to form grammatical sentenceconstituents. The current version of the parser is
able to analyse some quite complicated sentences.
Most parsers are build around a relatively small
dictionary where the information is tailored to the
needs of the parser. Our parser is completely
dependend on the information from the Cobuild
dictionary. One consequence is the programs
inability to obtain gender information from the
dictionary.
For a parser to be able to pinpoint an error in a
sentence we will have to adapt the parsing
mechanism to generale a hypothesis for possible
errors. Tuis will require extensive modifications to
the parsing mechanism and research into the most
common errors of students. When an error occurs a
feedback message will be generated and the learner
is given an opportunity to correct the error. The
final story is stored in the text database, and can
thus serve as a basis for new exercises. Tuis enables
leamers, among other things. to create fill-in
exercises for each other. All the help options based
on the dictionary that are available for the receptive
exercises are similarly available for this type of
productive exercise. The structure of a sentence can
be made visible in tree-form on the screen. This
possibility of random language input has one
drawback: it makes great demands on the speed and
capacity of the computer. The dictionary generates
all possible grammatica! uses of each word in a
sentence. Tuis gives rise to a combinatorial
explosion of possible ambiguities. Tuis slows down
the parsing process.We are looking into ways of
optimizing the parsing process to cape with this
problem. However, in our opinion the program is
uniquely valuable at present precisely because of its
- 38 -
Figure 3: A parsed sentence
-
.
セ@
IT'S I nqhsh
IOIM
Cohrnlcl
P..trsc frpe
sc>nlPllCP
-
flle
noun_ph,..,,,.,
,.,
dctorminer
capacity for processing random input, with
graphical feedback about the grammaticality of
input-sentences (see figure 3)
In the teaching of wntmg various aspects of
writing practice can be dealt with, such as planning
and organization, fonnulation and revision. The
program offers help in fonnulation at sentence
level. A panicularly powerful aid is the ability to
call up context sentences for a particular word in
order to examine its use in frequently occuring
contexts.
Learner registration.
The pro gram stores information conceming the
words that each learner has studied up to a given
moment, the help that has been requested and the
answers to the exercises that have been given. These
data can be scrutinized by the teacher.
Exercise developrnent.
The system applications allow learners as well as
teachers to enter random texts for the development
of exercises. For example, a teacher could in
principle enter a text for later use by the learner to
develop exercises. There are two entry-checks. The
first one tests whether the word with the desired
grammatica} properties is present in the computer
lexicon. The second check tests whether the parser
is capable of parsing the input-sentence and
whether the sentence is grammatica!. Bath checks
are required for the creation of exercises.
Stories can also be entered after the parser has
been cancelled. In such cases the only type of
operation to be carried out is the connection of each
word to its first dictionary meaning. The advantages
of tb.is option are that the input-process is faster and
that texts of any degree of difficulty can be entered.
Tuis kind of text can only be used for receptive
exercises.
The generative potentia! of the parser and the
great amount of information in the lexica! database
make it possible for the progam to adjust quickly to
different teaching methods by allowing for the
addition of method-specific texts. The menu-driven
environment and the extensive context-sensitive
help-options make the program very user-friendly
for bath teachers and leamers. The unambiguous
screen layout makes it relatively simple to enter
- 39 -
commands. It should be bome in mind, however,
that the program's full potential can only be
realized with suitable machines (i.e. computers with
80486 processors, 50 Mhz, 8 Mb RAM, hard disk,
Windows 3.1 and a soundcard).
5
Markham, P. Effects of contexrual versus
definitional
computer-assisted
vocabulary
instruction on immediate an dlong-terrn
vocabulary retention of advanced ESL srudents.
Educational Psychology, 9(2), 121-126, 1989.
Multimedia
6
Gein, J. van de, The sense of sentences. A
srudy into the effects of grammar instruction
upon junior writing. Proefschrift, 17 mei 1991 te
Utrecht
The Windows environment supports a wide variety
of ways to present information. Text an graphics are
the most common types used in applications.
7
With the addition of a soundcard sound can be
easily recorded and played back. This way listening
comprehension tests could be created.
A relatively new addition is the ability to display
digital video in a windows on the screen. The
performance in frame-rate and the size of the
playback-windows is currently quite limited.
Advances in image compression will soon overcome
these problems. Use of digital video in language
teaching opens other powerfull ways of creating
interesting exercises.
References
Kanselaar, G. & T. van Maanen, Vreemdetalenonderwijs met computers in het voortgezet
onderwijs: verkenning en inventarisatie van een
onderzoeksgebied. In: Cluster V onderwijsonderzoek: twee studies van het Instituut voor
het onderzoek van her onderwijs. PSOI-reeks
nr.31, 1987, p. 3-154.
2
Sinclair, J. (Ed.), Col/ins COBU!W
English Language Dictionary. Collins, London
and G!asgow, 1987.
3
Kanselaar, G., M. Giezeman, J. Jaspers, L.
Koster, J. van der Veen, H. Wichmann, J.
Zuidema.
Computerondersteund
vreemde
talenonderwijs (eindverslag van SVO-project
7108). Utrecht: ISOR, Universiteit te Utrecht,
1989.
4
Neuner, G., M. Krüger & U. Grewer,
Übungstypologie
zum
kommunika.ti.ven
Deutschunterricht, Berlin: Langenscheidt KG,
1981.
Kanselaar,G., Kok, WAM, Jaspers, J.G.M.
It' s English: Engels leren met computer en
CDROM. ISOR-rapportnumrner 91.07, ISOR /
Universiteit Utrecht, 1991.