As a fellow computational lexicologist, I would like to second Robert's suggestions and add a few more, inside and outside the box.
Inside the box, we examine the content the lexicographers have provided us. Thus, all of Bob's suggestions, and there is probably more. In about 1995, in a chat with George Miller, he lamented that he could only include a relatively small number of relations (compared to previous work he had done). Perhaps we could look to the full battery of Mel'cuk's lexical functions, as well as still other unmined data (e.g., dates for encyclopedic entries of people).
Outside the box, we can attempt to incorporate recent efforts that are not usually included in standard lexicographic practice. For example, Patrick Hanks, in his efforts at corpus pattern analysis, is including ontological values for verb arguments, reflecting the patterns he sees. This moves us into the linguistic domain, capturing arguments as in VerbNet. And still further, we reach the frame semantics of FrameNet. FrameNets are mushrooming for non-English languages as well. Wouldn't it be nice to see, not only definitions of words in context, but also the frame elements that we should expect in examining a word. Atkins & Rundell have provided a glimmer of this in their guide to "practical lexicography."
Here's to a superdictionary of the future.
Ken
On 11/9/2012 10:55 AM, amsler@xxxxxxxxxxxxx wrote:
I feel I should add something to this discussion...
I'm a computational lexicologist. My interest has been in the use of
computers to study the contents of 'machine-readable dictionaries', a
term I coined in 1980 in my dissertation on The Structure of the
Merriam-Webster Pocket Dictionary. (That work, in turn, led to George
Miller producing WordNet).
Electronic dictionaries have only partially achieved their potential
because they have only expanded their access capabilities in fairly
minor ways despite an avalanche of new computational capabilities.
Fundamentally, electronic dictionaries "think" of themselves as print
dictionaries being offered via electronic access. This is a very
limiting vision.
The work I did on the analysis of dictionary definitions demonstrated
that there was an imperfect, yet intriguing, taxonomy of definition
texts and showed that the alphabetic organization of dictionary
entries was outmoded except under special circumstances. I.e., for
example, you had to know how to spell a word to look it up; you had to
know a word existed that dealt with the meaning you were trying to
express to know to how to look it up; and when you did look a word up
you were given a tiny view of the dictionary's contents that didn't
show you the other words whose definitions were related to the entry
you were examining in terms of taxonomic relatives. Sure, some
dictionaries did an excellent job of including information on synonyms
(Merriam-Webster's "synonym paragraphs" come to mind, for their
inclusion of defining differences in text explanations; but NONE gave
taxonomic or part/whole related headwords).
Electronic dictionaries offer new capabilities in terms of now
providing one-at-a-time retrieval of entries based on words within
definitions; provide for word game options such as finding anagrams of
words. Algorithmic techniques such as the SOUNDEX system allow finding
words based on their sounds instead of their spellings (something that
Google seems better at than electronic dictionaries).
But fundamentally, dictionaries as isolated islands of knowledge, are dying.
Wikipedia offers "disambiguation pages" that extend beyond what is in
any dictionary, print or electronic. They engage in post-modern
lexicography in which proper nouns ('named entities' in the
computational linguistic community's jargon) share the likelihood of
being what a user is interested in looking up instead of just lexical
headwords. I remember my shock at discovering that in the Brown Corpus
the word "TIME" most often referred to the name of a magazine and not
any of the senses in a dictionary....
Web search engines have implemented "definition" as a search box
keyword that retrieves multiple web site hits giving the definition of
terms. Some (duckduckgo.com) have even taken to assuming that a
definition is the fundamental information to retrieve for any isolated
keyword entered into a search box. The dictionary as a specifically
evoked search is intuitively determined from the query string.
I doubt users will for long want to go to one publisher's web site and
then learn their specific interface all just to look up one unknown
word to get one publisher's take on its meaning... unless they are
interested in a very specialized type of knowledge such as definitive
etymological knowledge or a very specialized form of display. The pace
of new vocabulary has made most print dictionary publisher's web sites
antiquated.
So, where does the future of lexicography lie. I believe it lies in
the development of new lexical knowledge resources, new ways to
display existing dictionary information and in connecting dictionary
information to other knowledge.
For example. What would the dictionary look like if Google search
handled dictionary lookup? You'd have best match for strings of
keywords to a dictionary entry. You'd have sponsored links displayed
atop the free search hits. Sponsored links aren't all bad; it depends
on their relevance. If, for example, sponsored links went to the
titles of books related to the word or meaning being looked up, this
could be a good thing. I have often wondered how many of the headwords
in a dictionary have books with that title or books whose content is
about one of their sense definitions. If entries linked to government
publications or public service information or news stories for words
currently in the news it could be a good thing as well.
Of course, the problem here is that Wikipedia and Google and Amazon
already exist and they are all too eager to take the leap toward
incorporating dictionary information into their search results.
What isn't yet done may well be done by web-based companies. However,
in some ways dictionaries excel in what they do.
(1) Compaction of information. The dictionary entry may be the most
complex bit of typography ever devised. It involved more fonts and
formatting clues than any other type of text I've encountered. This
hasn't been well exploited by dictionaries in their electronic
interfaces. For example, if one could do arbitrary string search
through a dictionary's entries one could find similar entries to an
existing entry just based on the syntax of highly compact strings. No
need to detail what one is looking for, find me more entries that
contain: "n 1 cap:" (headwords whose 1st sense is a capitalized word)
or "<professor ~" (words that appear in example sentences following
'professor' as 'emeritus') or "`path-thik\" (the last part of the
pronunciation of "homeopathic" used as a query for find words that end
in similar pronounciations).
Note that in all these cases these are very incomplete strings taken
from actual definition entries being used as queries in a very simple
string search algorithm and not a highly structured search query that
required weeks and months of programming of an interface to allow
users to ask such questions about dictionary content. It's a "find me
more entries that contain this" query. And it works because of the
rigorous highly complex syntax print dictionaries have developed over
decades of evolutionary advances.
(2) defining formulae. Dictionaries employ similar defining styles
across entries with related content. Yet, they don't allow the user
convenient access to those defining formulae so they could retrieve
definitions based on their use. In part, I suspect this resulted from
handing separate lexicographers the task of defining all the entries
of certain groups of words such as animals, occupations, vehicles,
etc. It might be useful to be able to see the definitions that were
written for a given defining formula. Defining formulae are more
complex than can be retrieved by string searches since they employ
natural language that allows arbitrary numbers of adjectives and
and/or combinations to use the same formula. The underlying formula
would have to be identified to link together all definition texts that
used it.
(3) beyond one-at-a-time retrieval of dictionary entries. The
information science community has long used techniques such as
keyword-in-context to display search results as a concordance.
Electronic dictionaries have a very annoying habit of assuming that
readers want to read entries retrieved as one-at-a-time formatted
entries shown as they would appear in the printed book. NO, not all of
us do and many of us can read a KWIC listing more efficiently to see
what is going on across all entries that will match a query's results
displayed together, one result per line, formatted for horizontal
alignment of their shared text.
So... what to do. Either get busy dying or get busy living as the saying goes.
Dictionary publishers need to start figuring out how to live on the
web as a participant of their environment or figure out how to offer
their polished content in ways that don't currently exist. It isn't
quite a matter of whether it's a book, an online interface, or a
wireless interface, it's what it displays that is useful. It's a
matter of either having lexical knowlege that nobody else has or
displaying lexical knowledge in ways that are so convenient that other
means of access are less attractive.
There... now I've managed to offend as many people as possible...
Dr. Robert A. Amsler
Computational Lexicologist
Vienna, Virginia
------------------------------------
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/lexicographylist/
<*> Your email settings:
Individual Email | Traditional
<*> To change settings online go to:
http://groups.yahoo.com/group/lexicographylist/join
(Yahoo! ID required)
<*> To change settings via email:
lexicographylist-digest@xxxxxxxxxxxxxxx
lexicographylist-fullfeatured@xxxxxxxxxxxxxxx
<*> To unsubscribe from this group, send an email to:
lexicographylist-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
-- Ken Litkowski TEL.: 301-482-0237 CL Research EMAIL: ken@xxxxxxxxx 9208 Gue Road Home Page: http://www.clres.com Damascus, MD 20872-1025 USA Blog: http://www.clres.com/blog * To post to the AFRILEX Mailing List, send an e-mail to afrilex@xxxxxxxxxxxxx * To unsubscribe from the AFRILEX Mailing List, go to //www.freelists.org/list/afrilex, enter your e-mail address in the "User Options" box, choose the Unsubscribe option in the "Choose an action" dropdown list, and press "Go!"
