EFL Wordstation

0. Introduction
There is an amazing variety of machine-readable
English dictionaries (MRDs) now available in
terms of required hardware, software platform,
design, size and purpose. Yet, there are relatively
few MRDs which were custom-made for learners
of English as a foreign language. The Collins
Cobuild and Longman Interactive readily come to
mind, of course, and it is obvious that from the
point of view of EFL they are much more useful
than, say, the second edition of the OED on CDROM. Even these excellent resources, however,
share the two possibly most damaging weaknesses
of all EFL-oriented MRDs: L1-insensitivity and
access-inflexibility.
In this paper I will describe an English-Polish
MRD now in preparation in the School of English,
Adam Mickiewicz University, Poznan, Poland.
This MRD avoids the flaws mentioned above, and
thanks to its radically innovative design deserves
the proud name of Wordstation.
1. What is available?
Monolingual MRDs, just like other, technologically more traditional EFL resources are made for
the “generic” learner, i.e. the learner of an indefinite L1 background. By this token, they cannot
take into account those factors whose exclusion is
perceived by most EFL teachers and methodologists (including the undersigned) as inadmissible,
for example the notorious L1-to-L2 transfer on all
levels of language structure. Even the existing
bilingual English-Polish MRDs, of which there
are a few, are only nominally L-1 sensitive. Most
of them are hastily prepared word lists, with practically no phraseology, grammatical or usage advice and thesaurus functions, not to mention such
niceties as phonetic transcription1
or sense subcategorization. Usually no morphological normalization is built in, so that the EFL user must first
reduce children to child him/herself, only to find
the following entry (in one English-Polish MRD):
child node, child process, child task, duly translated into Polish computerese.
Another deficiency in English-Polish MRDs is
their inflexibility. While some monolingual English MRDs allow wildcard and definition (or
reverse) searches, this is a feature conspicuously
absent from Polish EFL MRDs. More specifically,
but very importantly in this context, among all the
off-the-shelf English MRDs known to me only the
OED2 provides for (poorly implemented) phonetic access, whereby users can search for words
through their pronunciation, an indispensable feature if learners are to be convinced that the foreign
language which they are acquiring is mostly spoken. Thus, the only type of dictionary access offered in most existing MRDs remains the thoroughly
traditional spelling lookup, which is a rather obvious waste of the available technology.
2. Multi-access in the EFL workstation
Following the recommendations of the more farsighted computer lexicographers (e.g. Calzolari
1989 or Knowles 1990), we have decided to offer
the user of the Wordstation the maximum number
of access paths to our 60-thousand-wordform English-Polish MRD. Thus, the user will not only be
able to look up a word as s/he would do in a
hard-copy dictionary (and see it on screen in comparable graphical layout), but also formulate arbitrarily elaborate queries yielding carefully filtered
word-lists, which can then be used for practice,
drilling, material-writing, testing, or simply serendipitous browsing.
Starting with the lowest level of language structure: phonetic access (see Sobkowiak 1994 a,b,c).
All pronunciation-related information built into
the dictionary will be readily available: (a) the
phonetic transcription of the British and American
pronunciation of each headword and wordform,
(b) segmental length in letters and phonemes, (c)
accentual pattern (primary and secondary stress),
(d) syllabic length, boundaries and C/V structure,
(e) some EFL-wise important distinctive features
like voicing in obstruents or vowel tenseness.
All this phonetic information will be made available to the user from a simple and easily customizable menu so that queries are possible of the type:
“What English words have two syllables with a
syllabic nasal at the end?”, or “Give me all trisyllables with primary stress on the penult”, or “What
words retain unstressed vowels or diphthongs in
unstressed syllables?”. On the other hand, users
will be able to reduce the phonetic display at will
should they not currently be interested in the pronunciation of words looked up. This reduction can
be quite radical, to the point of avoiding any
phonetic information whatsoever. This principle
applies equally to other types of lexical information in this MRD.
Second: frequency access. Word frequency has
only recently been recognized as an important
lexical datum worthy of inclusion in a learner’s
dictionary (see the recent line of corpus-based
learner’s dictionaries and resources published by
Collins and Longman (The former based on the
COBUILD corpus, the latter on the Bank of English corpus). In our Wordstation both written
(printed) and spoken word frequency will be provided, so that the learner will be able to (using
simple syntax and menus) formulate queries like
the following: “Which common English words
(say, among the first thousand in rank) are relatively 2
more frequent in writing than in speech?”.
Exact frequency data will be available if required
(unlike in the Collins and Longman dictionaries),
so that material-developers and EFL researchers
may use these on top of such global frequency
codes as rare.
Third, as much of the morphosyntactic information will be accessible for active searching as is
practically implementable. Derivatives and inflectional forms will of course be listed, but properly
linked to their headwords, and the user can decide
how s/he wants to view them. Other morpho-syntactic information encoded and potentially exploited in lexical searches will be (unsurprisingly): (a)
part-of-speech tagging with subcategorization
where appropriate and useful (verb transitivity,
tense-form and 3rd ps sg, noun plurality, adjective
grading, zero-derivation, etc.), (b) compound flag
to allow easy retrieval of compounded forms,
which must be differentiated from hyphenated
strings, fore-stressed polysyllabic nouns and multi-word sequences with separating spaces, (c) important idiosyncrasies in the morphological and
syntactic behaviour of words, for example pluralia tanta or the fact that some transitive verbs (e.g.
lack, survive, thank, deserve, undergo, etc., cf
Kjellmer 1992:341) are practically never passivized in natural English, both areas extremely problematic for EFL learners.
Thus, in addition to phonetic searches or in combination with them the user will be able to formulate queries like: “Which verbs are not regularly
inflected for past tense from their base form?”
(most of the so-called ‘irregular verbs’), or “List
all nominal binomial compounds with a deverbal
first term” (e.g. looker-on). The query language
and user interface will allow this level of sophistication without the requirement of the elaborate
metalanguage or grammatical jargon, as in the
examples above.
Fourth: phraseology, idiom, style. Currently no
comprehensive contextual exemplification is envisaged as part of the Wordstation. In a bilingual
dictionary this function can to some extent be
fulfilled by translation. Yet, typical English collocations will normally be listed for two categories
of words: (a) those in some respect irregular, idiosyncratic or collocationally restricted (e.g. the
noun abandon, practically untranslatable into Polish when outside the phrase with (reckless) abandon), and (b) most of the so-called ’function’
words, especially articles, prepositions and some
pronouns. Additionally, acronyms and abbreviations will be expanded, again with idiosyncrasies
marked (e.g. AAA is Amateur Athletic Association
in GB, but Automobile Association of America in
the US).
This access mode will give the EFL user some idea
of the collocability of the most troublesome
words, another area which is commonly regarded
as one of the most difficult in EFL instruction. The
realization that a given word or class of words do
not exhibit collocational idiosyncrasy is an important EFL datum in its own right.
Fifth: meanings. Three avenues of semantic access will be implemented: (1) through Polish
translation, which may be thought of as a direct
analogue of definition search mentioned above,
(2) through about 60 semantic field labels, like
tools, army, animals, space, bedroom, colours,
added to nouns with high concreteness ratings, and
(3) through a thesaurus facility containing common synonyms and antonyms of the word, where
relevant. Using all this information the user will
be able to formulate queries like: “Which synonyms of aberration have a medical meaning
(semantic field medicine)?”. Answer: abnormality, delusion, dementia, derangement, hallucination, insanity, mania, psychosis.
3. L1-sensitivity in the EFL Wordstation
Let us start the discussion of EFL MRD L1-sensitivity with errors. It is very useful, but by no means
sufficient, in an EFL MRD to append — as the
Longman Interactive does — a note of ’common
error’ committed by ’generic’ learners of English
in connection with a given lexical entry. There
may well be a common denominator for EFL
errors made by learners of varying language backgrounds, but this is a relatively restricted area that
needs to be complemented by a careful treatment
of L1-specific errors.
In our Wordstation Polish interference in the Polish-English interlanguage (or Polglish, for short)
is fully taken into account. First, on the level of
pronunciation, we have designed a simplified Polglish phonetic transcription (see Sobkowiak, in
press) whereby beginning-level users can enter
English words as they hear them (phonetic access)
and the way they would spell them in Polish, rather
than in the somewhat cryptic IPA transcription or
its derivate. For example, entering szol (with the
slashed Polish l) would correctly retrieve show
[SÎu], with all its associated lexical information.
By using Polglish transcription in this manner we
are making the dictionary user-friendly and L1-
sensitive on two counts: (a) to Polish spelling, and
(b) to Polish pronunciation errors.
244
This transcription is bound to be inaccurate phonetically and massively ambiguous, of course.
These deficiences are mitigated by the fact that it
will only be used heuristically, i.e. to help active
searches,and not representationally: proper IPA
representation will appear on screen (if desired by
the user). Fuzzy search programming (as in spellcheckers) will take care of excessive or zero hits.
Another way in which L1-sensitivity is manifest
in the phonetic stratum of the dictionary is the
phonetic difficulty index. The index contains two
fields: the first is a numerical tag carried by each
wordform which indicates its approximate difficulty for the Polish learner; the latter is a code of
the actual phonetic difficulty/ies present in the
wordform. The index has been derived through a
combination of algorithmic and manual tagging
and is a result of years of EFL teachers’ experience.
The difficulty index can be used in a number of
ways in the actual word searches and queries.
First, it will caution the user as to the high phonetic
difficulty of the word currently displayed. Second,
the index can be used in direct queries of the type:
“Which words of this or that semantic/morphological category are particularly difficult phonetically?”, or “Give me the phonetically difficult words
of the first 1000 (spoken) frequency rank”. Third,
because the index contains information about the
exact nature of the difficulty involved, it will allow
the user to investigate it directly through listing
words with this same difficulty present, for example: “If radio, which I am now having on screen is
pronunciation-wise difficult for Poles because
they tend to reduce the second-syllable vowel to a
glide /j/ and the whole word to a bisyllable, give
me more words with this phonetic problem in
them”. An exemplary answer to such a query
(listed in the order of frequency): ratio, enthusiasm, appreciate, studio, embryo, associate, abbreviate, negotiate, cardiac, video, dissociate, kiosk,
mediate, deviate, humiliate, pistachio, stereo.
To account for L1-motivated errors originating on
higher levels of linguistic structure explicit notes
will be taken in the body of the given lexical record
of common morphosyntactic and semantic difficulties associated with it, such as (a) translational
pseudoequivalents (’false friends’; E(actual)
P(aktualny) ’topical’), (b) homonymy and homophony, (c) homography, (d) all other common
Polish errors and problems. The treatment of the
last category will be informed — apart from the
anecdotal, experiential and intuitive data — by the
analysis of the corpus of English essays written by
Polish students and collected in our School as part
of the International Corpus of Learner English
(ICLE) Project coordinated from Louvain, Belgium (cf. Granger 1993). The Polish component of
the corpus, which is currently being collected,
edited, tagged and parsed, will contain about
200,000 words of argumentative and expository
writing at an advanced academic level of English
proficiency.
Finally, all the available lexical information contained in the Wordstation will be dynamically
used as a database for the CALL-like vocabulary
exercise component which we plan to build into
the package. The scope and detail of data as presented above will make it possible to design a
variety of exercise types, including multiple choice (“Which of these objects is not a musical instrument: banjo, bass, baton, cello?”), matching
(“Connect these Polish words with their American
English equivalents: ...”), classification (“What
category do these objects belong to: leek, carrot,
potato, peas?”), identification (“What is the English word for maly domek mysliwski?”), elicitation (“Write five words beginning with the prefix
con-”), ordering (“Order the following words from
the most to the least common: ...”), etc.
We believe that with this CALL facility our MRD
will transgress the limits of the multi-access L1-
sensitive dictionary, and will become a fully fledged EFL Wordstation.
Notes
1 The CD-ROM technology is quite recent in
Poland so there are thus far no dedicated multimedia speaking dictionaries of English.
Some such facilities appear as parts of certain
advanced CALL packages.
2 Relatively is an important word here, of course, as in absolute terms the predominance of
speech over writing is such that (in general
English) no written word is ever more frequent
than its spoken equivalent.
Bibliography
Calzolari, N. 1989. “Computer-aided lexicography: dictionaries and word data bases”. In I.S.
Batori, W. Lenders and W. Putschke (eds).
1989.Computational linguistics. Berlin: Walter de Gruyter. 510-519.
Granger, S.1993.“The international corpus of learner English”. In J. Aarts, P.de Haan & N.
Oostdijk (eds).1993. English language corpora: design, analysis and exploitation. [ICAME
13 proceedings].Amsterdam: Rodopi. 57-71.
Kjellmer, G. 1992. “Grammatical or nativelike?”.
In G. Leitner (ed.). 1992. New directions in
English language corpora. Berlin: Mouton de
Gruyter. 329-344.
Knowles, F.E.1990.“The computer in lexicography”. In Hausmann et al. (eds). 1990. Wörterbücher. Dictionaries. Dictionnaires. Berlin:
Walter de Gruyter. 1645-1672.
Sobkowiak, W. 1994a.“Phonetic-access dictionaries in TEFL: from vision to project”. Nordlyd
21: 33-41.
245
Sobkowiak, W. 1994b. “Phonetic-access dictionaries with L1-based simplified transcription”.
Poster presented at the 6th EURALEX Congress, Amsterdam 30.8.-3.9. 1994.
Sobkowiak, W. 1994c. “Beyond the year 2000:
phonetic access dictionaries (with frequency
information) in EFL”. System 22.4: 509-23.
Sobkowiak, W. (in press). “Radically simplified
phonetic transcription for Polglish speakers|”.
In S.Puppel & R.Hickey(eds). Festschrift for
Professor Jacek Fisiak on His 60th birthday.
Berlin: Mouton