CalicoSoftwareReport AntidoteGrammarChecker
description
Transcript of CalicoSoftwareReport AntidoteGrammarChecker
CALICO Software ReportCALICO Journal, Volume 16 Number 2, pp.
197-212Antidote 98 Reviewer: Jack Burston - Monash
UniversityProduct at a Glance
Product type: French Grammar/Spell Checker
Language-level: Intermediate to Advanced Level
Media format: One CD-ROM
Computer platforms: Windows 3.xx / 95 / 98 / NT 4 MAC-OS 7 /
8
Documentation: On-line Help & User's Manual (100 pages)
Introduction
Antidote is a relatively new entry into the French grammar/spell
checker arena. As with the well known Le Correcteur 101, it is a
product of French-speaking Quebec. The latest version of Le
Correcteur itself is extensively evaluated in this volume of the
CALICO Journal. The purpose of this software report is to put
Antidote through the same series of tests to see how well it
performs in comparison.. The producers of Antidote themselves are
very much aware that LeCorrecteur is the benchmark against which to
be measured and in their advertising claim superiority on tests
conducted by two magazines PC Mag (October 1997) and Protgez-vous
(February 1998). In addition, glowing reports are cited from four
other magazine reviews: PC Expert (October 1998), MacWorld France
(January 1998), SVM Science & Vie Micro (July-August 1997), and
Clic (February 1997), not to mention a half dozen testimonials from
satisfied customers. With such credentials, one might well ask why
Antidote is being reviewed yet again in the pages of CALICO. The
answer is twofold. Firstly, while they are very well placed to
assess the technological features of a product, popular computer
magazines rarely have the expertise needed to make qualified
judgements about highly specialized content areas such as those
embodied in a grammar checker. Secondly, all of the above reviews
and testimonials presuppose use by native speakers of French. How
well Antidote performs within the context of foreign language
students remains to be determined. General Description
Summary of features
Antidote is actually a suite of four interrelated tools consisting
of a grammar checker plus a dictionary, verb conjugator, and
reference grammar. The grammar checker is claimed to handle over
800 cases of agreement and to be sensitive to the occurrence of
homonyms (e.g. vert/verre), paronyms (e.g. minent/imminent),
anglicisms (e.g. email) and regionalisms (e.g. dbarbouillette). The
norm used to determine grammatical correctness is none other than
the venerable Le Bon Usage (M. Grevisse and A. Goosse). The on-line
dictionary is said to contain 100,000 entries incorporating 350,000
definitions drawn from the Dictionnaire de l'Acadmie franaise and
nearly 40 other sources. The verb conjugator boasts an inventory of
8000 verbs and 400,000 conjugated forms. The reference grammar,
also inspired by Le Bon Usage, is an in-house production which
treats over 250 topics. In addition, Antidote provides a built-in
link to a third party multimedia dictionary called Le Visuel. This
needs to be purchased separately and was not tested as part of this
review. Antidote can be used as a stand-alone text editor or
accessed directly within certain word processors. On the Windows
platform, this includes Word (Ver 6,7,97) and WordPerfect (ver
6.1). On the Macintosh, Word (Ver 6,98) and Claris Works (Ver 4,5)
are supported. Moreover, Antidote is compliant with Word Services
and be accessed from programs, such as Eudora Pro, which adhere to
this protocol. In stand-alone mode, Antidote can work with imported
documents in RTF or simple text (ASCII/AINSI) format. It is to be
noted that the user interface in Antidote is totally in French.
When operated in stand-alone mode, Antidote displays a menu bar at
the top of the screen from which its various functions can be
accessed. In addition to the menu bar, all Antidote tools are
available though keyboard shortcuts.Fichier (File), dition (Edit),
Fentre (Window) and Aide (Help) menus provide standard text editing
facilities. The Outils menu makes available the grammar checker,
dictionary, verb conjugator, and reference grammar. When the
grammar checker is called up, the text editing window is reduced in
size and a separate analysis window is opened below it. By default,
the grammar checker automatically analyzes the whole sentence
within which the text cursor is located. It is also possible to
check sequences below the sentence level by highlighting the
desired elements, whereupon Antidote will limit its analysis to the
selection. When grammar checking, the Diagnostic menu may be used
to display an explanation of an error, a corresponding grammar
rule, the part of speech/grammatical function of any word or a full
sentence analysis. Access to the dictionary, verb conjugator, and
reference grammar is context sensitive. Whatever word the text
cursor is on (or whatever portion of text is highlighted) serves as
a default lookup. Automatic lookup works as well from within
Antidote's pop-up windows, thus providing hypertext links between
the various tools. Dictionary searches identify exact matches
first. This is followed by a list of phonetically similar words. As
would be expected, the dictionary allows for the addition of new
words and definitions. Optionally, Antidote offers the possibility
of doing analogical lexical searches, i.e. a list of all dictionary
entries containing the search item.
When integrated into a compatible word processor, Antidote adds five toolbar buttons to the screen, the first four correspond to its primary functions and the last a link to Le Visuel:
These may be kept as a separate toolbar menu or dragged over to the word processor toolbar. The link to Le Visuel may be deleted if it is not needed. Installation (and deinstallation) of the Antidote toolbar is accomplished by simply opening a special word processor file containing the necessary macros. As happens with its own stand-alone text editor, when the grammar checker is called up, the word processor window remains on screen so that both the original document and the analysis window are visible.
As can be observed, when Antidote is activated its menu bar
appears at the top of the screen above that of the word processor
and a window displaying the selected tool is placed at the bottom
of the screen (all windows can, of course be repositioned as
desired). In the case of the grammar checker, the analysis window
itself is split into two panels, a large one at the top shows the
corrected sentence and a small one at the bottom displays the
original text. Clicking on any underlined correction pops up an
explanation of the mistake. Clicking on the Corriger button accepts
the correction and replaces the corresponding text in the original
document. The first double arrow button at the right of the window
moves the cursor to the next underlined correction in the sentence.
The second double arrow button moves the cursor to the next
sentence to be checked. A button marked Taire allows specific error
alerts to be turned off (words affected are still underlined). The
Ajouter button is used to add unrecognized words to the personal
dictionary. At the far left of the correction window, just above
the bottom panel which displays the original text, is a small
triangle. Clicking on this gives the user a complete syntactic
analysis of the sentence being checked. Lastly, toolbar buttons are
included within the correction window to allow direct access to the
dictionary, verb conjugator and reference grammar.
The grammar checker in Antidote can be configured to operate with
various degrees of control over an impressive range of parameters.
Most notable are the multiple options affecting sensitivity to
stylistic constraints, regionalisms and anglicisms.
Documentation
Antidote comes with extensive on-line help and a well illustrated
printed User's Manual of over 100 pages. The latter is called
(somewhat tongue in cheek) a Posologie, the subtitle of which is
equally playful: Le remde tous vos mots. For the language
specialist, the user's manual even includes a five page
bibliography of its lexicographical and grammatical sources.
Evaluation
Technological features
From a technological point of view, Antidote certainly merits the
acclaim it has received. For this review it was tested on a
Pentium/166 running Windows 95 and a Mac 8500/120 running OS 8.1
and performed flawlessly on both platforms. Installation was a
breeze and, as has become standard practice on Windows software, an
equally effective Uninstall program is also included. Speed of
operation was lightning fast on both computers. Screen layout is
clear and uncluttered. Icon functions are transparent and the user
interface is consistent and highly intuitive. So much so, in fact,
that few of the system's pull-down menus are likely to see much
use. Likewise, on the Windows platform, right mouse-button menu
links largely eliminate the need for keyboard shortcuts. Compared
to Le Correcteur, Antidote offers both a simpler user interface and
tighter integration of its suite of tools, most notably in the
hypertext links it provides. So, too, Antidote can be accessed from
within an external word processor. Unlike Le Correcteur , there is
no need to close a document and switch out of the word processor in
order to run the proofing tools. Whereas Le Correcteur can directly
import word processor documents in the formats it supports (e.g.
Word, WordPerfect AmiPro;Claris Works), Antidote only works with
RTF and simple text files when used in stand-alone mode. This,
however, is not likely to cause much inconvenience since, with the
exception of AmiPro, Antidote integrates well with most popular
word processors. In fact, users of these word processors would have
little reason to access Antidote in stand-alone mode. Stand-alone
operation, however, can be important when using a grammar checker
in an academic environment where RTF and simple text files provide
a lowest common denominator for the submission of student work. In
view of the formatting problems with imported simple text files
reported in the accompanying review of Le Correcteur, special
testing was undertaken to check for similar difficulties in
Antidote. Le Correcteur (as does Antidote) interprets hard returns
as sentence delimiters and these have to be removed for correction
to operate properly. The main problem with Le Correcteur is that,
even when hard returns have been deleted, the program sometimes
continues to stop its analysis at their previous location in the
text. To check how well Antidote handled imported simple texts, a
document was created in various Windows and Mac versions of Word
and WordPerfect and saved as simple text. These were then imported
directly into Antidote, were hard returns were removed. While no
problems were encountered with Antidote itself, incompatibilities
did occur with the encoding of accented characters when importing
texts across Windows and Mac platforms. In order for accented
characters to display properly, simple texts intended for
correction on a Windows platform need to be saved in a form
compatible with that environment and the same applies, of course,
in the other direction for texts intended for correction on a Mac
platform. The character code conversion itself is a simple process
and the addition to Antidote of an inbuilt Windows/Mac accent
converter would be a most welcome enhancement for institutional
users.
While on the topic of accented characters, it is worthwhile
pointing out a shortcoming which Antidote shares with Le
Correcteur: the inability to conveniently access diacritics from
the keyboard. Being intended for native French speakers, it is
assumed that the program will be used with a French keyboard
installed. On a Mac this is not problematic since users with QWERTY
keyboards can access accented letters through a standard set of key
combinations. In the absence of such an interface on the Windows
platform, users must resort to entering AINSI code or cutting and
pasting from a character map to obtain accented letters. When
Antidote is accessed from within a text editor, the typing of
accented letters is handled by whatever means is made available by
the word processor; but only for text entered in the word
processing window itself. Once the user begins working with the
Antidote dictionary or conjugation tools, word processor key
combinations cease to be operative. Word searches are sensitive to
the presence of diacritics and, though the program will list
phonetically similar items, typing in words without their required
accents can require the user to wade through a dozen or more
entries to locate what is actually being sought. This annoyance
turns to real frustration if Antidote is used in stand-alone mode,
as there is no direct text editing support for the typing of
diacritics. Granted, one can blame Windows for not providing
anything better than charmap.exe to handle accented characters and
hope some future version will remedy this weakness. In the
meantime, internal Antidote support for the typing of diacritics on
a QWERTY keyboard would be much appreciated.
Compared to Le Correcteur, Antidote offers fewer auxiliary
facilities. It cannot do fully automatic corrections or lexical
scans. The lack of automatic correction, i.e. one which analyzes an
entire text and makes changes without interactive confirmation, in
reality is not much of a loss since the accuracy of such correction
leaves much to be desired. With typical student compositions, Le
Correcteur allows over two-thirds of all errors to go undetected.
Of those it does spot, only about 25% are actually corrected, the
remainder are simply tagged as requiring attention (Mogilevski
1998). The absence of lexical scanning, however, is to be regretted
in Antidote. Lexical scanning is a very simple procedure involving
the creation of a list of all items in a text which cannot be found
in the dictionary (or verb conjugator). As a pedagogical device it
is very useful in focusing student attention on the simplest, and
alas most frequent, kinds of mistakes: verbal misconjugations,
irregular nominal forms, spelling, accents, and sundry typos.
Because of the high frequency of errors in student compositions,
compared to that of a native French speaker, grammar checkers
frequently can do no better than to provide a partial analysis of a
sentence. Eliminating the kinds of mistakes picked up by a lexical
scan before a grammatical analysis is attempted can greatly improve
the ability of a grammar checker to complete its task and, thus to
better identify and correct any remaining errors. For student usage
at least, Antidote really needs to add lexical scanning to its tool
box.
Another of Le Correcteur's useful tools lacking in Antidote is an
on-line thesaurus. This is offset somewhat by the ability of
Antidote to do analogical lexical searches. Likewise, in terms of
the number of lexical entries, the internal dictionary included
with Antidote rivals Le Correcteur's add-on Petit Robert
Electronique facility. Also, the reference grammar in Antidote,
which has no counterpart in Le Correcteur, offers clear and concise
context sensitive explanations. Grammar correction
Whatever its technological niceties and auxiliary features, what
really matters above all else in a grammar checker is its ability
to successfully detect and rectify mistakes. In so doing, it must
not only accurately identify the nature of an error but must also
avoid falsely finding mistakes were none exist. This is a tall task
when dealing with native speakers and an even greater challenge
when dealing with the written work of language learners. In
evaluating the effectiveness of a grammar checker, it is important
to recognize from the outset the current limits of the state of the
art. As powerful as modern desktop computers have become, it still
remains beyond their ability to support grammar checkers capable of
semantic analysis. Neither Le Correcteur nor Antidote has any
understanding of the meaning of the sentences it analyzes, as both
make clear in their user's guide. One important consequence of this
is that semantically based mistakes are largely immune to
detection. For instance, no grammar checker today is capable of
judging the appropriateness of pass-compos versus imparfait tense
usage. Similarly, they are insensitive to errors in anaphoric
reference such as Depuis 1890 le monde fait des films. Certaines
sont merveilleuses. Likewise, misuse of determiners inevitably goes
undetected. For example, both Le Correcteur and Antidote see
nothing wrong with sentences like J'ai cass ma jambe. Surprisingly
perhaps, the semantic blind spot also affects preposition usage.
Stereotypical student errors like J'ai rencontr Pierre sur le train
trigger no alerts. The reason for all this, of course, is that
semantic deviance is a matter of contextual congruence. Given the
right context, semantically anomalous sentences can become
perfectly acceptable utterances. Demonstration files
Notwithstanding semantic limitations, quite a wide range of
morphosyntactic mistakes are amenable to computer identification
and remediation. To demonstrate this, both Le Correcteur and
Antidote supply with their program a file of sentences illustrating
the kinds of errors they can correct. Naturally, each grammar
checker scores 100% with its own demonstration materials. As a
first step in assessing the relative effectiveness of Le Correcteur
and Antidote, each was tested against the demonstration file of its
competitor. In Table 1 below, corrected errors include mistakes
that were either replaced by correct forms or identified in such a
way to make correction straightforward (e.g. clicking on the
dictionary to find an unrecognizable word, making subject/verb
agreement in accordance with the grammatical analysis given).
Flagged errors were those which, though marked, required the user
to independently decide what corrective action to take.
Misidentified errors were those in which the source of the mistake
was wrongly attributed. Since these were misleading, they were not
counted as successful detections. False corrections were those
which were wrongly diagnosed as errors.Table 1 Demonstration File
Analysis
N/Err Detected MisidentifiedUndetected False
Corrected Flagged Total
Correcteur183 64% 24% 88% 1% 11% 1
Antidote82 73% 15% 88% 4% 8% 0
Out of a total of 183 errors in the sentences of the Antidote
demo file, Le Correcteur successfully detected 163, i.e. 88%. Of
these, 64% were directly corrected and 24% were flagged. The
Antidote demo file also contained 15 regionalisms and slang
expressions which Le Correcteur either passed over in silence or
detected without being able to identify. In addition, it also made
one false correction and incorrectly analyzed the nature of two
errors (1%). On the other hand, Antidote made one questionable
correction in its own example sentences, insisting that the
feminine of professeur is professeure. While many hitherto
masculine-only forms are in the process of acquiring feminine
counterparts, the latter is a regionalism not yet in common usage.
The demonstration file for Le Correcteur contained only 82 errors,
of which 72 , i.e. 88% were successfully detected by Antidote. Of
these, 73% were directly corrected and 15% flagged. No false
corrections were made, but the nature of 3 errors (4%) was
misidentified. While Antidote missed 3 lexical items which Le
Correcteur tagged as problematic, it also issued 6 alerts of its
own about possible vocabulary problems unnoted by Le Correcteur. In
addition, Antidote turned up 5 minor punctuation errors (missing
commas) which Le Correcteur passed over in silence. Thus all things
considered, though the margin of difference is not very great, the
larger number of direct corrections, absence of false corrections,
additional lexical alerts and punctuation corrections tilt the
balance in favor of Antidote. Native Speaker Texts
In reviewing Le Correcteur, a test was undertaken to determine how
well it could perform with an error-free French literary text: "Le
Capitaine Fracasse" by Thophile Gautier. The result of the
correction of the first seven sentences (271 words) was one partial
analysis, eight possibly misleading comments and thirteen false
corrections. Response time was also very slow, Le Correcteur taking
for example over 6 minutes to analyse the following sentence:
Deux tours rondes, coiffes de toits en teignoir, flanquaient les
angles d'un btiment, sur la faade duquel deux rainures profondment
entailles trahissaient l'existence primitive d'un pont-levis rduit
l'tat de sincure par le nivelage du foss, et donnaient au manoir un
aspect fodal, avec leurs chauguettes en poivrire et leurs
girouettes queue d'aronde. When subjected to the same test,
Antidote produced three partial analyses, made one possibly
misleading comment and eleven false corrections. Unlike Le
Correcteur, Antidote offers no control over the amount of time
allowed to perform an analysis; but then none is needed as its
operation is virtually instantaneous. The longest any sentence took
to analyze was four seconds. Table 2 Literary Text Analysis
Partial Analyses Dubious Comments False Corrections Maximum Time per Sentence
Correcteur1 8 13 6 minutes
Antidote3 1 11 4 seconds
A further test was undertaken to compare the performance of Le
Correcteur and Antidote on more discursive native French prose. The
text chosen was a newspaper article (1481 words) Justice pour la
justice by Robert Badinter (Le Monde, 8 February 1997). A
representative sentence is given below:
Depuis des mois, les autorits franaises avaient fait voeu de
silence avec, il est vrai, quelques justes raisons: toute
dclaration manant de Paris dclenche, quelle qu'elle soit, les vieux
rflexes de rejet de l'ingrence colonialiste et la France sert de
repoussoir d'un extrme l'autre de l'chiquier politique algrien.
Since few users are likely to accept waiting more than a minute to
analyse a single sentence, however complex, maximum correction time
was set to a more realistic 60 seconds for Le Correcteur. The
results of this test are summarized in Table 3.Table 3 Prose Text
Analysis
Partial Analyses Dubious Comments False Corrections
Correcteur12 6 6
Antidote7 3 8
As can be observed, when constrained to a 60 second analysis, Le
Correcteur left 12 sentences partially analyzed, made 6 dubious
comments and 6 false corrections. In comparison Antidote, operating
in 4 seconds or less, left 7 sentences partially analysed, made 3
dubious comments and 8 false corrections. Though making slightly
more false corrections, Antidote's superior ability to completely
analyse complex sentences with fewer dubious comments and
unrivalled speed give it a clear advantage over Le Correcteur when
dealing with native French texts. Student compositions
Being a proofing tool intended for use by native speakers of
French, Antidote, like Le Correcteur, is sensitive to a range of
subtleties unlikely to be found in the essays of language learners,
e.g.
Mes amies se sont faites prendre par la pluie.
Les enfants ont tout chamboul, tels une tornade. On the other hand,
in dealing with student compositions, grammar checkers are
confronted with such a quantity of anomalies that it becomes
difficult to determine the syntactic structure needed to correct
even basic mistakes. Moreover, very often the nature of student
errors defies all attempts at analysis. The following example
illustrates the kind of sentences which are all too frequently
encountered:
Il fallait que nous avons promis notre mre de ne pas se sparer.
Inevitably, such deviant forms produce incomplete analyses, and
with only a partial result to go by, comprehensive and accurate
error identification and remediation is simply not within the power
of any native-speaker grammar checker to provide. In the absence of
algorithms specifically modelled on student interlanguage1, it
needs to be realized that the application of programs such as Le
Correcteur and Antidote to the writing of language learners is
largely restricted to the detection and correction of purely
morphosyntactic errors: noun gender, adjective agreement, verbal
conjugation, spelling and the like. Notwithstanding, as most French
teachers are painfully aware, this is an area much in need of
remediation in student essays (cf Mogilevski and Burston2).
When put to the test against 40 second-year advanced level student
essays, containing some 1262 morphosyntactic mistakes, Le
Correcteur managed to successfully detect an impressive 91% of the
error types it is capable of recognizing. In order to assess how
well Antidote performs with student compositions, it was tested
against the same database3. The comparative results are summarized
in Table 4.Table 4 Student Compositions
N/Err Detected MisidentifiedUndetected False
Corrected Flagged Total
Correcteur1262 74% 17% 91% 7% 2% 49
Antidote1262 77% 5% 82% 11% 7% 28
As with the demonstration files, Antidote again outperforms Le
Correcteur in terms of the number of errors actually corrected: 77%
compared to 74%. However, this advantage is offset by far fewer
non-corrected error detections: only 5% compared to 17%. Moreover,
Le Correcteur with 91%, compared to 82% for Antidote, demonstrates
a substantially higher total detection rate. It is to be noted,
however, that this achievement comes at the expense of many more
false corrections, 49 compared to 28. Le Correcteur lets pass fewer
mistakes undetected, 2% compared to 7%. Likewise, it misidentifies
the source of errors less frequently (7%) than does Antidote (11%).
Based on over-all results, Le Correcteur would have to be judged
the better performer with student compositions.
Antidote would have fared better with student writing were it not
for its misidentification of some fairly obvious spelling errors
(e.g. premiere, videttes, emplois) and its failure to recognize
spelling mistakes in sentence initial words, which in every case
were incorrectly attributed to unknown proper nouns. Also when
operating in grammar correction mode, Antidote, unlike Le
Correcteur, does not offer lexical alternatives to misspelled
words. It either automatically makes a correction or informs the
user that it doesn't recognize the word. By being more cautious and
flagging such errors with a list of possible substitutes, Le
Correcteur relies on the user to make the ultimate decision and
thereby avoids lexical miscorrections. While automatic correction
of spelling errors may be advantageous for native speakers,
language learners are better served by Le Correcteur's approach. As
noted above, this is exactly what Antidote does when a word is
queried in dictionary lookup mode. Extending the presentation of
lexical alternatives to grammar correction mode is an improvement
which Antidote would do well to adopt.
The under-utilisation of Antidote's lexical resources to handle
misspellings in grammar correction mode contrasts markedly with the
program's remarkable attention to vocabulary usage. Although Le
Correcteur flags a fair number of lexical difficulties, Antidote is
in a class by itself when it comes to alerting the user to the
possible misuse of words: homonyms, paronyms, regionalisms, and
especially anglicisms. It issued a total of 236 vocabulary alerts
in the student compositions. Table 5 illustrates the most common
and useful of these:Table 5: Vocabulary Usage
actuellement/rellement franais/Franais, anglais/Anglais opportunit/occasion
anniversaire/fteimmigrant/migrant original/originel
apport/amener/emmener nombre/chiffrepeut-tre/peut tre
ces/sesprogramme/mission tche/ tache
When giving vocabulary warnings, Antidote provides a short
definition of the terms in question. Should this not prove
sufficient, a full dictionary lookup is only a mouse click away.
While vocabulary items like those above are extremely relevant to
language learners, some warnings are much less so: e.g.
ton/taon/thon; anglais/anglet; jeune/jene;matin/mtin. Fortunately,
Antidote allows its sensitivity to lexical items to be controlled
in a number of ways, so it is possible to fine tune this feature or
turn it off completely. Furthermore, used as a pedagogical device,
lexical alerts can provide an interesting and contextually
sensitive means of vocabulary acquisition. Conclusion
French speakers are much better served by grammar checkers than are
their counterparts in other languages. Thanks especially to the
Quebecois interest in the field, francophones - be they native
speakers, teachers, or foreign language learners - have at their
disposition proofreading tools which are state-of-the-art. Le
Correcteur and Antidote are both excellent products, as
demonstrated by their performance on the very demanding tests to
which they have be subjected here. However, in assessing the
relative merits of these programs, it is clear that Antidote has
now become the new benchmark against which to measure French
grammar checkers. Its user interface is a model of simplicity and
efficiency. In terms of operating speed, it leaves all competition
in the dust. When used with native-speaker texts, it is more
successful than Le Correcteur in analyzing complex prose. With all
texts, it consistently makes more automatic corrections; though in
doing so it is also more prone than Le Correcteur to misidentify
the source of errors. When used to correct student compositions, Le
Correcteur has the advantage of detecting a substantially greater
number of morphosyntactic errors than does Antidote. In part this
follows from the fact that Le Correcteur misidentifies the source
of errors less frequently. On the other hand, many of Antidote's
misidentifications would be transparent to a student, and easily
rectifiable through use of its on-line dictionary. Notwithstanding,
this area remains a weak spot in need of attention. False
correction, too, is matter of particular concern when these
programs are used with language learners, and in this regard Le
Correcteur is more problematic than Antidote. Given the competition
between these two products, and the obvious efforts which have gone
into their development to date, one would expect to see needed
improvements in forthcoming versions.Summary
Scaled rating (1 low-5 high)
Implementation possibilities: 4
Pedagogical features (relative to evaluation parameters): 4
Use of computer capabilities: 4
Ease of use (student / teacher): 5
Over-all evaluation: 4
Value for money: 5Technical Details
Developer/distributor: Druide Informatique
Address: 5515, chemin de la Cte-Saint-Luc
Montral, Qubec H3X 2C6
Canada
Phone: +1-514-484-4998
Fax: +1-514-484-7709
Email: [email protected] (sales: [email protected])
WWW: www.druide.com Hardware requirements
PC: 8036+
RAM: 8 Mb (12 Mb recommended)
Windows 3.xx / 95 / 98 / Win NT 4 OS 7 /8
MAC: 68020 / Power-PC
RAM: 4 Mb (available memory)
Hard disk space: 15 Mb
Price
Individual copy: $130 CAN
Site license: 25 stations $1200 / 50 stations $2200 / 100 stations
$4000 CAN Reviewer Information
Jack Burston is Senior Lecturer in French, Department of French
Studies, Monash University. The author teaches a post-graduate
course in CALL and supervises M.A. and Ph.D. candidates in this
area. He is particularly interested in authoring languages and
systems for courseware development and is currently Software Review
Editor of the CALICO Journal. Reviewer Contact
Department of French Studies
Monash University
Clayton Victoria
Australia 3168 Phone: (613) 9905-2222
Fax: (613) 9905-2137
Email: [email protected] 1 It is interesting to note
that the producers of both Antidote and Le Correcteur are presently
working on a version of their product for English speaking learners
of French. 2 "Morphosyntactic Accuracy in the Written Compositions
of Advanced University Level Students of French", Australian Review
of Applied Linguistics, to appear. 3 I would like to thank Eugene
Mogilevski for providing me with the database of student essays
used in his evaluation of Le Correcteur.
References
"Druide informatique Antidote 98 -- Un correcteur brillant", PC
Expert, no 77, October 1998, p. 82."Test. Logiciels de correction:
qui la faute?", Protgez-vous, February 1998, p.18-21."Les 50 coups
de coeur de la rdaction -- Antidote", Macworld France, no 15,
January 1998, p. 74."Outils d'aide la rdaction", PC Mag, no 114,
October 1997, p. 82-88."Antidote 1.1.2 -- Un correcteur efficace et
rapide", SVM Science & Vie Micro, no 151, July-August 1997, p.
111."Le Petit Druide -- Une premire mondiale : un dictionnaire
lectronique descriptif et normatif pour la francophonie!", Clic, no
14, February 1997, p. 7-8.