1. E-dictionaries and phonolexicographic needs of EFL users Włodzimierz Sobkowiak School of English Adam Mickiewicz University Poznań, Poland

2. Abstract
The phonetic aspect of (EFL) dictionaries is among the most seriously underrated and underdeveloped in (meta) lexicography. Pertinent bibliography is scant and even the best learner dictionaries are found wanting on a number of counts. This contribution is both a summary of my 13-year-long research into (pedagogical) phonolexicography (see bibliography at and a look ahead. I present the current state-of-the-art in phono-lexicography with particular attention paid to how the leading pedagogical EFL e‑dictionaries relate to the actual and potential phonolexicographic needs of their users, both students and teachers.

3. Presentation plan
Phonolexicography: the state of the art and the main issues
Phonolexicographic needs of EFL users
Phonetic representation
Phonetic access
Didactic aspects of phonolexicography
Phonolexicographic dreams
Conclusions
(40 slides in all)

4. Is pronunciation important?

5. Phonolexicography (1): the state of the art
Phonolexicography is almost completely ignored linguistically and lexicographically
"Dictionaries are less satisfactory in pronunciation than in spelling, meaning, or etymology. The record of the spoken language is difficult to acquire, difficult to transcribe accurately and unambiguously, difficult to represent understandably in a dictionary transcription, and in most cases of less interest to the user than other kinds of information" (Hulbert 1955, summary from Landau 1991:97).
"Today, when we no longer regard speech as a degraded form of writing, the pronunciation entry in dictionaries […] should be accorded much greater importance". "Unfortunately, the theory underlying the pronunciation component in a dictionary is too frequently difficult to discern" (Gimson 1973:115).

6. Phonolexicography (2): main issues
The choice of phonetic transcription (IPA? none?, respelling?) to be used in the microstructure of the entry.
Which text should be transcribed (head entries, run-ons, phrases)?
The choice of default accent (RP? LFC? GA?) and phonostylistic level to represent.
The scope and choice of phonolectal variants.
Ensuring consistency of phonetic representation across the entries.
Questions of phonetic access in both traditional and computer-mediated dictionaries.
Audio sound representation in electronic dictionaries (e.g. recordings vs. TTS).
The choice of fonts, colours and other technical issues of visual sound representation.

7. Phonolexicographic needs of EFL users (1): are there any? (1)
Some of the largest groups of dictionary users – translators, businessmen, secretaries, scientists – do not need pronunciation in their lexical resources.
In heavy dictionary use for professional purposes, where the word's pronunciation is irrelevant and can actually be completely ignored, phonetic transcription can be worse than superfluous – it can be obtrusive.
Notice in this context that there is no option in, e.g. MEDAL, to switch transcription off completely for screen display.

8. Phonolexicographic needs of EFL users (2): are there any? (2)
Foreign language learners also mainly use dictionaries for ad hoc translation rather than for learning.
All studies of monolingual (E)FL dictionary use, from Tomaszczyk 1979 to Lew 2004, show that search for meaning is the main rationale for dictionary use to most learners, most of the time.
But most lexicographic (questionnaire) research is about learners' wants, not needs, so care should be taken in interpreting the results.
In Lew 2004, college learners were asked how often they were looking in their dictionaries for information on how a word was pronounced:
Never
Rarely
Often
Most often
179
336
160 35

9. Phonolexicographic needs of EFL users (3): what are they?
Sobkowiak 1999 (university EFL students) :
Statement
Yes No ?
English dictionaries should show pronunciation in some way
601 24 20
I check pronunciation of some words when I read
547 95 3
I have recently checked pronunciation in an English dictionary
611 32 2
In listening comprehension tasks I look up some words I hear
458
170 17
Learners' dictionaries should address their common pronunciation problems
528 51 66
Separate pronunciation dictionaries are nonsense 90
519 36
In a multimedia computer dictionary, phonetic transcription is useless 33
372
240

10. Phonolexicographic needs of EFL users (4): learner friendliness
If lexicographers are aiming at global 'user-friendliness' of their definitions (e.g. defining vocabulary), they should certainly also make them phonetically friendly.
For example, EFL electronic dictionaries could dynamically adjust their definitions to the learner's needs and requirements, not only in terms of their lexical scope (defining vocabulary) and syntax, but also in terms of pronunciation (see de Schryver 2003 for this and other lexicographer's dreams).
For example, if thorough is among the phonetically hardest lexical items for EFL learners, why not use a substitute in definitions (e.g. complete) adjusted for pre-intermediate learners, where appropriate? Or at least why not reduce the definition incidence of thorough (it now stands at 22 in MEDAL)?

11. Phonolexicographic needs of EFL users (5): teacher's dictionary? (1)
Do teachers use pedagogical e-dictionaries?
Should they differ from learners’ dictionaries?
Google search (in website title) for "teacher's dictionary": 14 hits, among them:
O.U.H.Jung: Elsevier's foreign-language teacher's dictionary of acronyms and abbreviations.
Alan Corkish: Teacher's Dictionary: Essential Manual for Teachers, Students and Parents.

12. Phonolexicographic needs of EFL users (6): teacher's dictionary? (2)
Some expected features, from the phonolexicographic perspective:
powerful multicriterial phonetic access mechanisms
liberal use of phonetic transcriptions, both for representation and indexing
word-list generation mechanisms
dependable cut-and-paste mechanisms
indication of phonolectal variation, with...
advice on the preferred phonetic form
advice on likely phonetic problems with the given entry/definition/example
remedial drills and exercises, generated automatically from phonolexicographic context

13. Phonetic representation (1): transcription vs. others
Phonetic transcription offers categorization, systematization, accuracy anchor
Audio representation offers a receptive and productive model
Waveform representation offers multisensory feedback for practice, but is risky because...
"The visual comparison of the two sound waves, the model’s and the learner’s repetition [...] is at best inconsequential, and at worst thoroughly misleading and frustrating. The graphic representation of a waveform has a very complex relationship to its acoustic basis, and the latter to both its articulation and perception" (Sobkowiak 1997).

14. Phonetic representation (2): simplified transcription?
To foreign learners there are no self-pronouncing words: the case of my desultory (on analogy with compulsory).
Respelling is good for native users/learners, as it is heavily target-language-dependent.
"A respelling pronunciation system is fairly practical. No special characters or diacritics are used. No pronunciation guide must be relied upon. Examples are: accident (AK-si-dunht), diamond (DIE-muhnd), garage (gah-RAHZH, guh-RAHZH, GA-rahzh)" (from the web).
In e-dictionaries many types of transcription are in principle mutually convertible and selectable.

15. Phonetic representation (3): typography
Screen rendering: The whole of CD-ROM MEDAL is set in sans serif arial-like font [...]; the only exception being the non-Roman IPA symbols [...] which are serifed. Depending on the graphics card's selected screen resolution, both the shape and the size of the transcription field symbols may differ:
/kntrvsi/.
Pasting: the phonetic fragment of controversy entry copied into MS Word 2000:
/B  `k\197ntrEv3:si/.

16. Phonetic representation (4): choice of the model
Which accent? Which phonostylistic variant? Traditional monolectal solutions seem to work best for learners.
On the other hand, real phonetic variation caused by dialect, phonic style, lexical frequency, or indeed the so-called 'free variation' should not be suppressed by the overwhelming force of the 'monostylistic curse' (Bailey 1986:26).
Solution: customize display of phonetic content to the needs of the user (e.g. learner vs. teacher, different proficiency levels).

17. Phonetic representation (5): para-phonetic information, e.g. frequency
Should phonetic representations in dictionaries be made sensitive to frequency-motivated phonological rules? One example: "Vowels in frequently-used words reduce more often than in relatively rarely-used words" (Fidelholtz 1975:208).
Word frequency does matter in the subjective judgement of the word's phonetic difficulty by foreign learners. The more common the word the less phonetically difficult it seems, other variables kept constant (Sobkowiak, unpublished).

18. Phonetic representation (6): phonological consistency
"To the familiar appeals for transcriptional and stylistic consistency of dictionary phonetic representations I would like to add mine - for what, in need of a better term, I call phonological consistency. This consistency will normally result from conformity of representations with the established phonological rules of the language" (Sobkowiak 1997:98)
Electronic dictionaries for learners give them, as well as the lexicographers, the unprecedented power of pattern-spotting, rule-discovery and exception (and lexicographic error) identification.
For example, in MEDAL I discovered twenty headwords starting with /ænti-/ (e.g.: antibiotic, anticlerical, anticyclone, antidepressant, antifreeze, etc.) and seven with /ænt-/ (antibody, anticlockwise, antidote, antigen, Antipodean, the Antipodes, antithetical).

19. Phonetic representation (7): transcription vs. audio
The main problems with the audio component of many dictionaries are: (a) the list-reading effects (e.g. contrastive stress, sustained intonation contours), and (b) the occasional mismatch between recording and transcription.
I listened to all British recordings in MEDAL within the <a>-letter range, and found about seventy such cases in the few thousand entries, for example: academician is transcribed as /kdmn/ but pronounced as /kdmn/ in MEDAL. Or consider the non-coalesced recording of aperture /pt/.

20. Phonetic representation (8): scope (1)
Can/should phonetic representation extend further than the headword?
"The phonological behaviour of words in context and its representation deserves equal attention" (Magay 1979:103)
Some of the hardest pronunciation problems arise for learners at the suprasegmental level of dictionary definitions and have to do with sandhi phenomena (inter-word juncture), rhythm, stress and intonation.
Phonetic analysis of EFL dictionary definitions (Sobkowiak, in press) shows that there are systematic ways in which dictionary makers can control them more tightly than has so far been the case, for example by phonetically sensitive DV word selection or by paying more attention to juncture phenomena.

21. Phonetic representation (9): scope (2)
Why not include definitions and examples in the scope of phonetic representation?
Automatic transcription from text is feasible: /If ju melt 'sVmwVn ju rI'l&ks &z D4 h5ld ju kl5s w4/.
TTS can now produce human-like speech. Listen to Crystal (ATT on-line demo):

22. Phonetic access (1): transcription functions
The representational function of transcription differs from its indexical (query) function.
In traditional dictionaries the latter did not exist.
In electronic dictionaries both functions must be smoothly integrated.
Advanced users (teachers and students) need the latter function more than beginners.
The question of grapho-phonemic biuniqueness for data data entry, for example: the search for /*/ words yields: aitch, approach, arch, attach, etc., if affricates are coded bisegmentally.

23. Phonetic access (2): Phonetic Access Dictionary (PAD)
In a PAD all phonetic infomation inherent in the lexicographic content is searchable through a user-friendly interface. Multi-criterial Boolean query mechanisms are built in. Lists of words meeting the desired criteria are generated.
Examples: Dictionary X, MEDAL, CEPD, PAD – e.g. querying for aspirated plosives or vocalic hiatus .

24. Phonetic access (3): Dictionary X (2003) interface - no access

25. Phonetic access (4): MEDAL interface

26. Phonetic access (5): CEPD interface

27. Phonetic access (6): PAD interface

28. Didactic aspects of phonolexicography (1): e-dictionaries as teaching/learning resources?
Brazil (1987:161): "It cannot realistically be seen as part of the dictionary's function to teach the sound system".
The traditional idiographic perspective of lexicography is not well suited to the needs of the foreign language learner, with his/her craving for some foothold of rules in the quicksand of idiosyncrasies.
Which EFL resource other than properly coded e-dictionary could (semi-)automatically generate exercises spun around those English animal names which are relatively common in colloquial English, but relatively difficult (grapho-)phonetically to Polish (foreign?) learners: calf, lamb, sow, bison, donkey, giraffe, leopard, monkey, reindeer?

29. Didactic aspects of phonolexicography (2): techniques (1)
The only type of pronunciation exercise currently used in EFL e-dictionaries is "repeat after me", cf. an example from MEDAL:

30. Didactic aspects of phonolexicography (3): techniques (2)
Pronunciation practice could easily be combined with flashcards. Why is it that the only elements of an entry's microstructure used in flashcards are its headword and definition? Why not let the learner "type mystery word" in response to its phonetic transcription or record-ing (dictation), or both, as well as to its definition?
Why not flash the headword and ask the learner "Do you remember how to pronounce this word?", as well as "Do you remember what this word means?", the only option now built in?

31. Didactic aspects of phonolexicography (4): techniques (3)
But one could, of course, go much further, e.g. into data-driven learning. For the beginner: "What do these words have in common, as far as pronunciation goes: dough, go, know, sew, toe?". For the advanced: aphrodisiac, Chianti, cordiality, piano, react?" (a selection from [*iæ*] headwords, containing a particularly troblesome case of vocalic hiatus).
It is easy to transform these exercises into binary- or multiple-choice format, matching, selection or minimal pairs. It is possible to combine them with audio and/or transcription, as well as part-of-speech information, lexical frequency tagging, dialectal and stylistic stratification, illustrations, etc.

32. Didactic aspects of phonolexicography (5): L1-sensitivity
Both theoretical and applied phonolapsology must be L1-sensitive because phonetic interference from L1 is the amplest source of L2 pronunciation errors on most levels of proficiency.
Contemporary EFL dictionaries are not L1-sensitive, or only superficially so (e.g. false-friends lists), mostly due to the overwhelming commercial factors (costs of localization).
Example: how do EFL dictionaries on the Polish market take account of this Polish graphophonemic rule: <c>  /ts/? Listen to Martyna's (3 years of EFL) romantic /ro'mantits/: ... in a romantic way.

33. Didactic aspects of phonolexicography (6): the undiscovered potential of definitions
Listen to Martyna reading the MEDAL definition of ‘melt’ - if you melt into or against someone you relax as they hold you close in a romantic way
With phonetically treated definitions, some Martyna's phonetic problems could be dealt with by reference to a phonetic concordance (phoncordance) focused on the relevant difficulties, as follows:

34. Didactic aspects of phonolexicography (7): MEDAL definitions with linking /r/
exactly: in every way or every detail
intense: very great or extreme
severe/ly: very strict or extreme
to have one foot in the grave: to be very old or ill and likely to die soon

35. Didactic aspects of phonolexicography (8): MEDAL definitions with hard sandhi consonant clusters: /s_/ and /z_/
dripping: fat and juice that is produced by meat when it is cooked
mitigating circumstances: facts that help to explain a crime or mistake and make it seem less bad
payphone: a telephone in a public place that you pay to use
unvarnished: expressed in a very direct way that gives the true facts

36. Didactic aspects of phonolexicography (9): MEDAL definitions with sandhi /d+j/ coalescence
minicab: a car that is used as a taxi. You must call for it by telephone and you cannot stop it in the street
snore: a sound you make when you breathe noisily while you sleep
freestyle: a race in which you can use any style or method you want to
fulfil: to do what you have said you will do

37. More (phono)lexicographic dreams (1): phonetically treated defintions
I dream of dictionaries which would redress the anti-phonetic bias of current lexicography.
In such dictionaries not only the phonetic representation of the headword would be carefully thought over, but the entire entry would likely receive phonolexicographic attention.
For example, definitions would be designed and written according to some phonetic guidelines, just like they are written according to strict syntactic and stylistic guidelines today.
In consequence, such definitions would be easier to read, both as meta-text and text, and the incidental learning of vocabulary would get a boost.
Properly annotated, they could be used as a (phonetic) electronic corpus resource in simple look-up as well as in a variety of sophisticated queries informing word-list generation, test preparation, materials design, etc.

38. More (phono)lexicographic dreams (2): articulatory animation
The so-far lexicographically unimple-mented type of potentially pedagogic-ally useful phonetic representation is articulatory animation. Realistic avatars of the Baldi kind could be used, with speech animated in real time. Different perspectives, zooms, tempos and levels of transparency could be used. The animation can be co-ordinated with a human recording or with TTS output. The technology is now mature for use in e-dictionaries.

39. Instead of conclusions
"Indicating pronunciation is often under-estimated by the critics of dictionaries as being a derivative business" (Magay 1979:99)
"Suppose the user wishes to see all entries for three-syllable nouns which describe movable solid objects, whose second syllable has a schwa as peak, and whose third syllable has a coda that is a voiced stop" (Alshawi, Boguraev & Carter 1989:60) .

40. Thank you!