Representing
information about words digitally
Digital Audio Archiving Workshop
Jane Simpson
Linguistics,
Table of Contents
Abstract
Introduction
Collecting and
interpreting data
Storage and display
Problems with digital dictionaries
Multimedia dictionaries: pictures
Multimedia dictionaries: sound
Conclusion
Acknowledgments
References
Figures
Figure
1: Wordlist with a geographic interface and the possibility of
interactive adding of words
Figure 2: Dictionaries with multiple
ways of looking up information
Figure
3: Multimedia dictionaries: the Kirrkirr interface for the
Warlpiri Dictionary
Figure
4: The Wagiman
online dictionary
Figure 5: Dictionaries with
pictures
Figure 6: Talking
dictionaries: dictionaries for the sight-impaired
Figure 7: Talking
dictionaries: an Albanian handheld dictionary
Figure 8: Talking
dictionaries: dictionaries with speech recognition
Figure 9: Talking
dictionaries: a Hmong talking dictionary
Figure
10: Talking dictionaries: Koi fish
encyclopaedia
Figure 11: Talking dictionary:
Celebrity Talking Dictionary of breast cancer terms
Abstract
The late 1960s saw the start of the "electronic-dictionary age" (de Schryver, 2003). The growth in the use of computers has
transformed all aspects of dictionary-making, from collecting data about word
meanings and uses, creating a set of dictionary entries, and displaying, using,
preserving and distributing these entries and the data on which they are
based. This paper discusses the transformations, and considers the ways
in which dictionaries for minority languages are leading or lagging in the
electronic-dictionary age. Illustrations are taken mostly from the uses
of digital sound in modern multimedia dictionaries.
Introduction
Dictionary makers have four main tasks: to collect data, to analyse and
organise it, to store it, and to display it for users. The advent of
computers has changed our understanding of all four tasks, and has made new
things possible in the areas of representing signs (spoken or signed), access
to information, and indicating the denotations of words. However, as Bird and
Simons (2003) note, new problems have arisen as a result of the new technology,
in particular the tensions between the desire for longevity of format, the
technical skills required to implement some of the new functions, and the small
market for dictionaries of endangered languages. Sharp divisions are
rising between what is possible for large commercial dictionaries of world
languages, what is possible for dictionaries of endangered languages, and what
is possible for one-off displays of word-lists. I illustrate these points
mostly by reference to 'talking' dictionaries. The causes of these
divisions lie not only in the relative lack of resources, but also in the
difference in demands for longevity for dictionaries. I shall show
how the new technology has affected the tasks of dictionary-making, before
turning to the rise of multimedia dictionaries.
Collecting and interpreting
data
The first task for a dictionary-maker is the collection of data. The main
data collected about words for dictionaries concerns information about sound,
meaning, use, history and relations to other words. Before computers
became widely available, the main source of words was from written material,
and to a much lesser extent transcription of spoken words. Thus
dictionaries tended to represent the words used in writing, and not words or
phrases used mainly when speaking. In the days before sound recording
made it possible to retrieve and verify spoken words, the fact that a word
appeared in written work, and was accepted by readers as a real word, was the
main guarantee of the authenticity of the word. But lexicographers still
rely on searching written transcriptions for new words, and then going from the
transcript to the audio file to check for the pronunciation, since there are as
yet no easy tools for searching streaming sound from, say, radio or television,
for new words, or even for collecting many examples of the pronunciation of the
same word.
Once a word has been identified as needing an entry, lexicographers interpret
the material surrounding the word in the light of as many examples as they can
collect and examine in the time available. The reliability of the interpretation
depends in part on how much material the lexicographer can assess in preparing
the entry. Gathering material has been made much easier by the arrival of
computers, the development of large-scale digital corpora, the growing amounts
of material on the internet, and the possibilities for cheap solicitation of
data on the internet, (for example the Australian Broadcasting Commission and
the Macquarie Dictionary's online collaboration at "Australian Word Map" (Figure 1)). This has meant that the number
of examples to examine has increased greatly, even for endangered languages
where the body of texts is usually small. What can be done with the
examples is limited by the lexicographers' time and ability to take in so many
examples. Techniques for automatic sorting and classification of the
information are needed (Fillmore and Atkins 1994), in order to reduce the
burden on the lexicographer. Some are implemented in large commercial
dictionaries. For example, lexicographers and
computational linguists from the Information Technology Research Institute (
On the basis of the collected material, lexicographers use their own judgment
to create lexical entries, including guides to the pronunciation of the
word. In the past, sound was particularly difficult to collect, because
English spelling does not serve as an unambiguous guide to pronunciation. If
the lexicographers had not heard the word, they could be led astray; thus the Australian National Dictionary (Ramson 1988) treats 'Nunga' (the name many South Australian Indigenous people
call themselves) as though it were the same as 'Nyungar'
(the name many southern Western Australian Indigenous people call
themselves). The lexicographers did not realise that the orthographic 'u'
in the first syllable represents different vowels in each case.
The
lexicographers' pursuit of reliability of data and validity of interpretation
of data is in part driven by the weight societies often place on dictionaries
as reference works. Before computers were used in dictionary-making, it was
hard to distribute primary data (sound, video, contemporary notes). The
lexicographer's job was to synthesise and analyse that material, and the
dictionary entry was a representation of that synthesis. In terms of information,
the dictionary came to be seen as semi-primary documentation - a way of making
primary data accessible through the filter of the lexicographer.
References to published sources of example sentences, as in the Australian National Dictionary (Ramson 1988) are references to primary documentation for
the existence of a word, and for its use with the meaning proposed by the
lexicographer. They act as a guarantee of the lexicographer's proposal.
The lexicographer's filter seemed invisible to the normal dictionary user, who
often assigned to dictionaries the role of authorities, arbiters of what a word
means, or what a reasonable person would understand by such and such a
word. Once dictionaries were viewed as authorities, they gained the
function of language standardisers. What a word
means, or how it should be pronounced, is determined by the dictionary.
Now that
digitisation has become widely available for text, sound and image from
recordings of speech events, it means that the primary data of field recordings
can be made widely available. This has led to the development of the
distinction between language documentation and language description (Himmelmann 1998). However, making material available
is not the same as making it accessible, and it is generally hard to find the
information one wants quickly in raw field notes. So, dictionaries still serve
as ways of making the primary documentation accessible. But more and more
they are seen as consisting of two parts, a way of linking to primary
documentation, and the lexicographers interpretation of that primary
documentation. Such primary documentation could consist of sound and text
collections, which can be linked to dictionaries (e.g. Baker and Kovacs 2003, Nathan 2000).
These types of links provide the user with the means to judge the authenticity
of the data. At the same time, computers have caused certain kinds of
dictionaries to maintain the language standardising function - the dictionaries
associated with spell-checkers help standardise the spelling of words, while
the thesauruses associated with some word processing programs reinforce the
idea that particular words are partial synonyms of each other.
Digital
linking between the primary documentation and the lexical entries requires
software, and the technical skills to use that software. Large commercial
dictionaries can afford to employ programmers and even software designers to
customise this software. Unfortunately for ordinary lexicographers, the
most widely available software that is easily used is commercial proprietary
software; for example FileMaker Pro allows the insertion of images and sound
files (and see Baker and Manning 1998). The ease of use and the existence
of documentation and a pool of other users make such products very attractive
to small groups attempting to provide dictionaries for endangered languages,
(or of other resources, such as image archives and catalogues). But
the seductiveness of the ease of installation, data entry and linking blinds
the users to the long-term storage problems of proprietary software as well as
to the restrictions on display - it is difficult for naive users to produce
attractive printed displays from FileMaker Pro. Ideally, dictionaries
will be prepared in open source standards such as HTML or XML, which will allow
multi-media linking, and have more chance of longevity (Bird and Simons
2003). At the moment, however, there are no widely available
lexicography software packages for multimedia dictionaries which implement
widely agreed-on standards for linking dictionaries.
Storage and Display
Once materials have been collected and analysed, the next tasks involve storage
and display. The two are treated as independent actions, although they do
interact. In the pre-computer days, the book was mostly treated as both storage
and display; the card indexes from which the dictionary was created were rarely
accessed, and often contained early versions of entries, rather than the final
copy-edited and proof-read entry. Thus the book was usually the master
copy. But the longevity and archiving problems were essentially solved by
distribution of books, and the presence of copies of books in different
libraries and archives. Once a book was printed and in libraries, the
lexicographer could relax, thinking that the work was now in principle
indefinitely available.
Having a
book as the primary way of storing and displaying information has
limitations. In terms of representation of information, the requirement
that the information be representable on a page
places limits on how helpful the representation of a word is. The appearance
of, and information in, a dictionary entry are constrained by tradition as well
as by the confines of the printed page, the cost of production of books, and
the limits imposed by the need for books to be handled comfortably. For
sign language dictionaries these limitations are highly constraining, as
pictures and words do not convey adequately the movement of handshapes.
The limitations for sounds can also be severe. Most readers have trouble
with the International Phonetic Alphabet and the kinds of pronunciation guides
used in dictionaries (Fraser 1997). This is a significant problem for
endangered languages. Aboriginal participants in languages workshops have
told me that it is painful to learn to use a dictionary, learn a word from the dictionary,
and pronounce it to a speaker, only to have that speaker reject their
pronunciation or fail to understand what they have said. This leads them
to distrust dictionaries.
A related
problem concerns the difference between the sound denoted by a word and the
sound of a word as a sign, as in the following English example:
Tut /t
t/ tuts, tutting, tutted
1. Tut is used in writing to represent a
clicking sound that you make with your tongue to indicate disapproval,
annoyance and sympathy 2. If someone tuts, they make a clicking sound with their tongue to
indicate disapproval, annoyance or sympathy. He tutted and shook his head.
[Collins Cobuild English Dictionary.
The
dictionary entry does not make it clear that there are now two different
lexical items in our mental lexicons, the click /|/ or /!/
and the CVC sequence /tt/, which is also used in writing to represent the click sound.
They directly result from the difficulty of representing click sounds in the
English orthography. In writing we have developed a way of representing the
sound /|/, as "tut". But from
reading, we have started to pronounce "tut"
as it is spelled, /tt/. We use this as a verb or interjection to express disapproval,
/tt//tt/, as well as the click /|//|/. This brings up the need (raised in
Burke (1998) and Sobkowiak (2003)), to provide audio
files to represent the denotation of a word when it involves sound. In the
following Warlpiri dictionary entry, the
technical definition "retroflex affricate click" is of little use for
the average reader of a dictionary. Being able to hear the sound of the
young kangaroo would be a more effective way of letting them learn what kirrkirrmani denotes.
kirrkirr-ma-ni V. make click Definition: x produce clicking
sound, characteristic sound produced by young kangaroo
Kirrkirr-manulparla kurdu-pardu
marluku. The joey was going click, click, click
to the kangaroo.
Note: Actual sound is like that of retroflex affricate click. Warlpiri
Dictionary. (Laughren et al in prep.).
The
problem of describing the denotation of kirrkirrmani, and the lack of
clarity about the two pronunciations of "tut"
arise because of the difficulty of recreating the sound of a word from a
spelling or description of it. These problems can, and are, being solved
now, because sounds can now be made available digitally. Commercial
dictionaries such as the new Macmillan dictionaries (Rundell
2003) come with CD-ROMs that show both the pronunciation and the denotation of
words denoting sounds. For a listener/viewer to go from a dictionary
entry to an example of a word in a simple digital sound recording of a spoken
text, there must be a link, which could be a link to a particular sound file,
or to a time-coded section of a larger sound file, or it could be a link
mediated through searching through a transcript of a sound file. However,
doing this linking can be very time-consuming. The ordinary linguist or
lexicographer does not yet have robust tools for efficient linking of sound and
dictionary entry (and see Baker and Kovacs 2003).
Displaying
dictionaries digitally has a number of advantages which have been discussed
elsewhere (Corris et al 2000), including fast
searching of big dictionaries (compare using the twenty volume Oxford English Dictionary with the online version), different ways of
searching (for example, geographic interfaces (Figure 1),
bilingual interfaces and thesaurus interfaces (Austin and Nathan 1998, Figure 2)), and the ability to link entries
to each other. The comparative lack of constraints on space means that
pictures and sounds can be added in. Some of these ideas have been
implemented in Kirrkirr (Manning et al
2001), a multimedia interface for dictionaries in XML format, which has
been used for the Warlpiri Dictionary (Laughren
et al in prep.) (Figure 3), as well
as for Nahuatl. It was designed by Christopher
Manning and Kevin Jansz, and is maintained by
Manning.
In
principle, the display of digital dictionaries can be refreshed and reproduced
more easily than paper dictionaries; that is, new information can be imported.
This is undoubtedly the case for large commercial dictionaries, with
in-house programmers. But for lexicographers working on endangered
languages, the time and labour spent in going from stored material to displayed
material means that second editions of such dictionaries are rare. Another
serious side-effect is that often last-minute changes are made on the displayed
version of the dictionary, which then becomes the master, rather than the
stored version. This is a particular problem when books are printed
from databases. Since the stored version usually is more computationally
tractable than the displayed version, this results in a loss of usefulness
long-term.
Problems with
digital dictionaries
As the preceding discussion has shown, producing a substantial digital dictionary
which makes sensible use of the possibilities of digitisation (linking and
multimedia) requires more computer skills than the "ordinary working
linguist" (Lawler and Dry 1998) or lexicographer generally has. And,
whereas when a paper dictionary is produced, it can be shelved in a library and
left for hundreds of years, anything other than a plain text digital dictionary
requires constant attention to ensure that it still works on the latest
computers. This is not something that the ordinary lexicographer or
linguist is trained to be conscious of. If they are conscious of it, then
this tends to make them nervous about embarking on a project whose longevity is
not guaranteed. The same problems arise with scholarly editions of literary or
historical works; scholars are wary of spending several years of their lives
preparing a work which may not be useable in ten years' time (Berrie 2001).
Distributing
dictionaries on the web adds further complications, the
internet allows user interaction, comments and feedback on words, as discussed
earlier. In first world countries and in cities, access can be fast and
cheap. However, in the remote areas and third world countries where most
of the world's endangered languages are spoken, access to computers, download
charges and phone charges can make online dictionaries inaccessible (and see
Bird and Simons 2003). The internet offers widespread distribution, which
helps short-term preservation ("lots of copies keep stuff safe" LOCKSS). But, in the view of many
Indigenous language speakers, this advantage is offset by the loss of
intellectual property in their languages, once representations of them are
freely available on the web. When wordlists of Indigenous languages are
on the web, then people will collect those words and use them for purposes the
original depositors did not dream of - electronic poetry for instance.
Words from the lists will be used for names of houses, farms, boats,
businesses, dogs and babies. For example, in 2003 the word lardili-yan was
used on a sculpture in an exhibition in
"The Kamilaroi/ Gamilaraay
language belongs to the Kamilaroi people and to Kamilaroi country, northern
But in
general, the idea that languages belong to the speakers is a new idea to many
users of the Internet, who, as speakers of world languages, are accustomed to
the idea that a language is available for anyone to learn and use, and that
material on the web is there to be used in whatever way the user feels like.
Multimedia dictionaries:
pictures
I turn now to specific concerns in building multimedia dictionaries. More
and more dictionaries are making use of multimedia (Macmillan, as already
mentioned, but also Oxford Advanced
Learners Dictionary on CD-ROM (Hornby and Wehmeier 2000), and the American Heritage
Dictionary). Pictures have been the first to come, drawing on the
old tradition of illustrated dictionaries. They are useful in improving
the representation of signs (whether spoken or sign language signs), access to
the information, and evocation of the denotation of words. Some electronic
picture dictionaries are already available for Australian languages (Hamilton 1996-8) (Figure 5).
It is certainly
true that for many words, their denotation can be more quickly grasped by
looking at a picture. Words for natural kinds, tools, weather, and
topographic terms are fairly readily picturable (of
course many other words are not so easily picturable).
However, my experience with collecting pictures for the Warlpiri
dictionary is that it is much more time-consuming than expected. While
there are many images available, ideally these will be tested with speakers to
ensure that the speakers recognise what the pictures are supposed to represent,
and that they are happy with the pictures being used to illustrate the word
concerned (a point made by Joyce Hudson in the pre-computer days). Many
pictures of plants, birds and animals, for example, do not give a indication of size. And many pictures of plants and
animals do not give the habitat (creek bed, sandhill,
rocky hill), although for Indigenous people of
Pictures
are mostly used to indicate the denotation of a word. But they can be used
for access - clicking on a picture to learn what the word for it is. For
sign languages, pictures can also be used to represent the sign. However,
video is generally more useful for this, since it allows the representation of
movement and change of handshape (e.g.
Multimedia dictionaries:
sound
Sound has made its way into digital dictionaries. On
Talking
dictionaries for the sight impaired are an obvious use (Figure 6) - providing a
medium for them to find more information. Language learners are an obvious
market for talking dictionaries, because pronunciation is hard to replicate
from transcriptions. Since there are many people who want to learn
languages, the market both for talking versions of the big English dictionaries, and for the handheld talking dictionaries is
potentially very large, and was the goal of the seven advertisements
found. But, surprisingly, among the handheld dictionaries were some for
languages which are rarely the targets of language learners- Albanian (Figure 7) being one such (perhaps the
foreign aid workers and reconstruction workers provide a commercially viable
market?). Most of these dictionaries provide the sounds of the words, but
are accessed by typing. Some however offer voice recognition, 'say a
phrase in language X, and receive a spoken translation in language Y' (Figure 8).
There
are also a number of non-commercial bilingual dictionaries providing the sounds
of words, for example a Hmong talking
dictionary (Figure 9).
Special interest groups also sometimes put up talking dictionaries, especially
where the words are unusual, such as a dictionary of koi
fish (Figure 10). These
usually only contain a small number of words, reflecting the large amount of
time presently needed to add digital sound to dictionaries.
It is
quite noticeable however, that these dictionaries use sound mostly for
representing the sign, and occasionally for accessing the information.
They rarely use sound for indicating the denotation of the word. This is
clearly an area for further development. For example, users would find it
helpful to have the definitions of words denoting sounds or sound related
actions extended with prototypical examples, e.g. for English screech,
husky, sigh, opera or Warlpiri kirrkirr-mani (clicking), yawulyu (women's
ceremony). Definitions are still essential - I don't think one could tell
just from listening what can be called a 'screech' and
what a 'scream'. Definitions of birds and animals could be enhanced by
recordings of their characteristic sounds -
particularly important for creatures like frogs which may be more often heard
than seen. This would also be useful for creatures whose name is
onomatopoeic - hearing the call of a crested bellbird would help the user
understand why the Warumungu call that bird, karnparnpalala.
There are
also many decisions to be made in collecting sound for dictionary, which
usually need to be made in conjunction with the speakers of the language, and
can be quite time-consuming. These include decisions such as:
• Whose voice to
use? This is an important issue in Indigenous communities where language
ownership is a living concern. The words should be pronounced by
someone who has the authority to do so. This is also the case more
generally, if dictionaries are treated as major representations of the
language, or as carrying some kind of authority, or some other message. For
example, the front page of the "Celebrity Talking
Dictionary of breast cancer terms", (Figure 11), has as a selling point:
"You'll be able to hear how the terms are pronounced and what they mean in
the voices of the fabulous celebrities - from the media, sports, music, and
medical worlds -- who have contributed to the Dictionary."
Perhaps
the compilers are right. The vocabulary of breast cancer is intimidating, and
hearing a term defined by someone who is not a specialist in the area, but who
is known to the user, (and who perhaps the user admires) will probably make the
definition easier to take in.
• How many
speakers? Ideally, more than one speaker would be recorded; this allows users to
abstract away from speaker peculiarities. It also gives them some choice, so
that if they find it hard to listen to one voice, they can listen to a
different voice.
• What variation
should be represented? The answer depends on what the users and the
community want the dictionary to represent - what is acceptable variation, and
what is considered unacceptable, ungrammatical, substandard, uneducated
etc. It is an important question to discuss, because the interests of the
language documenter may clash with those of language standardisers,
people who want their language represented by a standard form.
• Problems of
collecting the sound (quality and content) A decision has to
be made if the text is to be read or spoken. Natural spoken texts have the
advantage of authenticity, but may result in unusable material if the speech is
fast. If the text is to be read, there are often problems finding a
literate speaker as opposed to a fluent speaker, let alone a literate speaker
with authority. This speaker then has to be put in a good situation
for recording (free of noise and interruptions). The speaker has also to
agree to give up a large amount of time for the recording. There are also
problems of content. The dictionary cannot be completed before the
speaker records the words and sentences, because the speaker will inevitably
make changes and have disagreements. In effect the speaker is acting as
an editor for the dictionary; any words or sentences they object to will have
to be modified in the accompanying print dictionary.
Making
the decisions mentioned above can take months of negotiation and planning.
For example, recording fewer than eight hundred words and sentences for the Warumungu Illustrated Dictionary has taken more
than five months (Samantha Disbray p.c. December
2003). Similarly, a large amount of time
is involved in digitising sound to archival quality, selecting the relevant
examples of words, and linking the dictionary entries to the
sounds. Baker and Kovacs (2003) show how an entry can be
efficiently and automatically linked to occurrences of words in pairings of
transcribed speech and digital sound, instead of the time-consuming and often
inadequate splicing of words from digital recordings and manual linking to the
dictionary entries. This looks as though it will be very useful,
depending on the factors mentioned above. Of course, the usefulness
does depend on how clear the examples in natural speech are, whether all the
words required appear in good quality digital sound-transcription pairings, and
whether the language owners are happy to have the speaker(s) in the texts represent
them.
It is
clear that multimedia dictionaries can in principle provide a much better
representation of the sign, whether as a record of the sound of a word, or as a
record of the movement and facial expression associated with a sign language
sign. The potential for access by voice recognition or by keystroke also
improves access to information for everyone, including the sight impaired and
the hearing impaired. Sound and pictures, together with the definition,
can also provide a much better evocation of the denotation of picturable words or sound words. But multimedia
dictionaries bring to a head the problems mentioned earlier. Two problems
that need to be addressed now are, firstly, that there is currently no easy way
for an ordinary lexicographer to make a sophisticated multimedia digital
dictionary which conforms to open source standards, without the help of a
programmer. And secondly there is no easy way to revise and update such a
dictionary. The problems of longevity and archiving are becoming increasingly
severe. Now, when standards change, the dictionary maintainer has not
just one thing to update, but a package of text, sound, picture and video and
the links between them to update. Since standards, software and hardware
for archiving text, video, picture and sound material may all change independently, maintaining useable copies of digital
dictionaries in archives is now a big undertaking.
Large
commercial lexicographers can afford this work, because they expect immediate
commercial returns from book sales, they expect continuing demand which will
fund new editions, and they know that for major languages the work of making dictionaries
will continue. Longevity of format is not generally a major concern
for them. People doing pilot projects with small amounts of data, such as
the Koi Fish
Encyclopaedia probably are not interested in longevity, and so can
experiment with different kinds of software. People with access to
computational linguists can engage in on-off projects which may produce a good
result for a particular language. Most lexicographers of endangered
languages have few resources, since the number of people who wish to learn
endangered languages is usually too small to justify the time and cost involved
in making a a good digital
dictionary. But they must consider carefully the longevity of their
work. Theirs may be the final documentation of a language, and so must be
stored in the best and most enduring format they can afford. They do not have
access to computational linguists and they do not have time to spend learning
software packages whose longterm archival survival is
doubtful. When they go into partnership with information technology
professionals, this frequently results in interfaces for the dictionary at hand
(often in proprietary formats), rather than interfaces which can be easily
installed and customised by a lexicographer for another language. The
problem is exacerbated by funding bodies and Government departments which want
to produce glossy CD-ROM products to show that they are doing something about
endangered languages. This has resulted, and is resulting, in expensive
CD-ROMs with lots of glitzy features that contain little language material, and
are not customisable for other languages. They are infertile plants, and they are the recipient of most funding
support. There has been no effort put into thinking about what users
will do with these CD-ROMs. And yet the effort and expense involved in
putting together a CD-ROM for endangered languages can only be justified if the
CD-ROM will have lots of users who will spend a lot of time looking at it or
doing things on it. Thus a CD-ROM of a big dictionary is justified, because
people will use it time and again as a reference. A CD-ROM for teaching
phonics will be justified if each year a new class of students will use it more
than a couple of times. And so on.
In conclusion, time, labour, lack of accepted standards for dictionary
software, lack of technical skills on the part of the lexicographer, and the
need for longevity of format are major reasons why there are so few talking
dictionaries of endangered languages. My 'lexicographic dream' (de Schryver 2003) is of fertile
plants. A fertile plant would be shell dictionary software which is easy to
use, and can be used for many languages. 'Easy to use' means that
language workers can enter in data easily, browse
it, update it, and access it (that is, not stored on a server accessible only
by a decaying phone link). It should allow users to print out hard copies of
different views of the data and different subsets of data. A key
requirement of the software is that it should meet archival and portability
standards (Bird and Simons 2003). This would make it easy to translate
when new sound, text and picture standards emerge.
Acknowledgments
This paper draws on ideas developed from working with
Miriam Corris, Christopher Manning and Susan Poetsch, as well as from conversations over the years with
Peter Austin, Brett Baker, Jenny Green, Robert Hoogenraad,
Doug Marmion, David Nash, David Nathan and Nick Thieberger, and with many other people trying to build
dictionaries for languages they are working on. It also benefited from
discussion at PARADISEC's Digital Audio Archiving
Workshop and from suggestions made by an anonymous reviewer and by Steven
Bird.
ABC Online. 2002. Australian
Word Map. http://www.abc.net.au/wordmap. Viewed on
American Heritage Dictionary. 2000. Fourth
Edition.
Anonymous. n. d. Celebrity Talking Dictionary of breast cancer terms. http://www.breastcancer.org/dictionary/welcome.php Viewed on
Anonymous. n.d.
LOCKSS. http://lockss.stanford.edu/. Viewed on
Austin, Peter, and Nathan, David. 1998. Kamilaroi/Gamilaraay Web Dictionary. http://coombs.anu.edu.au/WWWVLPages/AborigPages/LANG/GAMDICT/GAMDICT.HTM. Viewed on
Baker, Brett, and Christopher Manning. 1998. "A
dictionary database template for Australian Languages." Paper presented at
AUSTRALEX.
Baker, Brett, and Michael Kovacs. 2003. "The AudioText Linking Tool." In DIGITAL AUDIO ARCHIVING
WORKSHOP: Researchers, communities, institutions and sound recordings:
Research implications and feasibility of distributed digital archives for ethnographic sound material in
Berrie, Phillip William. 2001. "Are electronic editions
inherently obsolete?" Paper presented at Computing Arts 2001 Conference,
Bird, Steven, and Gary Simons. 2003. "Seven
dimensions of portability for language documentation and description." Language 79(3): 557-582
Also http://www.language-archives.org/documents/portability.pdf.
Viewed
Burke,
Sean Michael. 1998. The design of
online lexicons. Master's thesis,
http://www.speech.cs.cmu.edu/~sburke/ma. Viewed on
Collins
COBUILD. 2003. Advanced Learner's English
Dictionary. With CD-ROM. Fourth
Edition.
Corris, Miriam, Manning,
Christopher, Poetsch, Susan, and Simpson, Jane. 2000. "Bilingual
dictionaries for Australian Aboriginal languages: user studies on the place of
paper and electronic dictionaries." In Proceedings
of the ninth EURALEX International Congress, EURALEX 2000, eds. Ulrich Heid, Stefan Evert, Egbert Lehmann and Christian
Rohrer, 169-181.
de Schryver, Gilles-Maurice. 2003. "Lexicographers'
dreams in the electronic dictionary age." International Journal of Lexicography 16:143-199.
Fillmore, Charles J., and Atkins, B.T.S. 1994. "Starting where the dictionaries stop: the challenge of corpus
lexicography." In
Computational approaches to the lexicon, eds. B.T. Sue Atkins and Antonio Zampolli, 349-393.
Fraser,
Helen. 1997. "Dictionary pronunciation guides for English." International Journal of Lexicography
10:181-208.
Hamilton,
Philip. 1996-1998. Uw Oykangand and Uw Olkola Multimedia Dictionary. http://www.geocities.com/Athens/Delphi/2970/. Viewed on
Himmelmann, Nikolaus P. (1998). "Documentary and descriptive linguistics." Linguistics 36: 161-195
Hornby, A. S. and Sally Wehmeier. 2000.
Johnston,
Trevor. 1998, c1997. Signs of Australia on CD-ROM [a dictionary of Auslan
(Australian Sign Language)]. North Rocks, N.S.W.: Royal Institute for
Deaf and Blind Children.
Kilgarriff,
Adam, and Michael Rundell. 2002. "Lexical
Profiling Software and its lexicographic applications :
a case study." In Proceedings
of the Tenth EURALEX International Congress, EURALEX 2002, Copenhagen,
Denmark, August 13-17, 2002, eds. Anna Braasch
and Claus Povlsen, 807-818.
ftp://ftp.itri.bton.ac.uk/reports/ITRI-02-18.pdf
Viewed on
Laughren, Mary, Hale,
Kenneth, and Hoogenraad, Robert. in prep. Warlpiri dictionary. Electronic datafiles. 2000 version.
Lawler,
John M. , and Dry, Helen Aristar
eds. 1998. Using computers in linguistics : a practical guide.
Manning,
Christopher D. , Jansz,
Kevin, and Indurkhya, Nitin.
2001. "Kirrkirr: Software for browsing and visual exploration of a
structured Warlpiri dictionary." Literary and Linguistic Computing 16:123-139.
See also http://www-nlp.stanford.edu/kirrkirr/. Viewed on
Nathan,
David. 2000. "The spoken Karaim CD: sound, text,
lexicon and "active morphology" for language learning multimedia."
In Studies on Turkish and Turkic
Languages (Proceedings of the Ninth International Conference on Turkish
Linguistics), eds. Asli Goksel
and Celia Kerslake.
Ramson, William S. ed.
1988. The Australian National
Dictionary: a dictionary of Australianisms on historical principles.
Rundell, Michael (Editor-in-chief). 2002. The Macmillan English dictionary
for advanced learners.
http://www.macmillandictionary.com/about.htm. Viewed on
Rundell, Michael (Editor-in-chief). 2003. The Macmillan essential dictionary.
Sobkowiak, Wlodzimierz. 2003.
"Review article: Pronunciation in Macmillan English Dictionary for
Advanced Learners on CD-ROM." International
Journal of Lexicography 16, no. 4: 423-441.
Wilson, Stephen, Harvey, Mark, Dalpbalngali,
Lulu Martin, Emorrotjba, Helen Liddy,
Benbo, Paddy Huddlestone, Gumbirtbirtda, Clara McMahon, and Gapbuya,
Lenny Liddy. 2001. The Wagiman online dictionary. http://www.aiatsis.gov.au/lbry/dig_prgm/e_access/digital/a339234/dict/dict.html. Viewed on
Figure 1: Wordlist
with a geographic interface and the possibility of interactive adding of words:
screenshot of ABC Online. 2002. Australian
Word Map. http://www.abc.net.au/wordmap. Viewed on
Figure 2: Dictionaries with multiple ways of looking up
information: screenshot of Austin, Peter, and Nathan, David. 1998. Kamilaroi/Gamilaraay
Web Dictionary. http://coombs.anu.edu.au/WWWVLPages/AborigPages/LANG/GAMDICT/GAMDICT.HTM. Viewed on 109/12/03. Return to text
Figure 3: Multimedia dictionaries: the Kirrkirr interface for the Warlpiri Dictionary. Return to
text
Figure 4: The Wagiman online dictionary: screenshot of: Wilson
et al 2001. The Wagiman online
dictionary. http://www.aiatsis.gov.au/lbry/dig_prgm/e_access/digital/a339234/dict/dict.html. Viewed on
Figure 5: Dictionaries with pictures: Screen shot of
Hamilton, Philip. 1996-1998. Uw Oykangand and Uw Olkola Multimedia Dictionary. http://www.geocities.com/Athens/Delphi/2970/. Viewed on
Figure 6: Talking dictionaries: dictionaries for the
sight-impaired: screen shot of http://www.talkingsoftware.gothere.uk.com/.
Viewed on
Figure 7: Talking dictionaries: Albanian handheld dictionary:
screenshot of http://www.mindconnection.com/Merchant2/merchant.mvc?Screen=PROD&Store_Code=M&Product_Code=LANG-EAl600T&Category_Code=LANG-ALBANIAN Viewed on
Figure 8: Talking dictionaries: dictionaries with
speech recognition: screenshot of http://www.universal-translator.net/. Viewed on
Figure 9: Talking dictionaries: A Hmong
talking dictionary: screenshot of http://ww2.saturn.stpaul.k12.mn.us/Hmong/dictionary/hmongeng/newmenu.html Viewed on
Figure 10: Talking dictionaries: Koi
fish encyclopaedia: screenshot of http://www.koi.com/encyclo/hmuji.html Viewed on
Figure 11: Talking dictionary: Celebrity Talking
Dictionary of breast cancer terms: screenshot of http://www.breastcancer.org/dictionary/welcome.php
Viewed
on
