EDICT/JMdict
Future Directions

Introduction

This page contains some thoughts I have been having about the future (if any) of the EDICT and JMdict project/files, and some possible courses it might take.

I welcome discussion, feedback, etc. It can be posted on the sci.lang.japan newsgroup or emailed to me here.

Wiktionary-oriented discussion is probably best carried out on the discussion page there.

Why Am I Raising This Topic?

Well, editing and coordinating these files has been largely a one-man-band. Me. Sure there has been a lot of advice from others, and certainly masses of input, but I have set the standards, done the updates, released new versions, etc. etc.

I think the files are useful enough and of a good enough quality that they deserve to live on, expand, be maintained, etc. They are the only freely-available and parseable source of Japanese-English lexical material.

I would like to see a future where:

1. the project is continuing and to a major extent self-sustaining, with the edit/update processes spread over a larger group of people;

2. it is not at all dependent on my continued involvement. I won't be around forever, and in fact after 14 years of EDICT I am seriously thinking about the rest of my life (of which there is not an awful lot left.)

3. it is not dependent on support from Monash University. My honorary appointment there will not continue for ever, and internal changes at Monash may well result in withdrawal of server support.

My Vision, Hopes, Whatever.

What I would like to see is something like this:

1. the underlying database from which the EDICT and JMdict files are generated migrates from its present form (a large text file on my PC) to an on-line database where it can be seen, edited, expanded, etc. by a community of users;

2. the edits to the database undergo some form of moderation/oversight, either prior to their commitment (e.g. a moderation panel) or a more passive after-the-event fixing of mistakes (more the Wiki model);

3. from the on-line database a regular and automatic extraction take place to generate the distributed forms of the file (currently EDICT and JMdict, but there may be other formats in the future.)

4. a more "open" copyright and usage licence arrangment. I am considering moving to a Creative Commons licence. The one I have in mind is the Attribution licence, which is very similar to the current one.

Options

Well, as I see it there are two main ways this vision could be achieved:

1. A Special System

   Outline. A server be developed around updating the EDICT/JMdict database.

   Pros:

   • could be made to match exactly the desired update model.
   • new releases of EDICT, JMdict, etc. would be relatively straightforward.
   • existing copyright, etc. arrangements continue.

   Cons:

   • a significant software development/debug effort needed.
   • maintenance, future location of the server become issues.

   Problems:

   • actually making it happen. I have been exploring PHP/MySql (all my WWW server work in the past has been in C and not used databases). I'm feeling a bit tired and old for yet-another major software development.
   • maybe some other software could be modified to do the task, e.g. Twiki.

2. Move Into An Established Environment, e.g. Wiktionary

   Outline. The entire dictionary database is uploaded into an established "wiki" environment, e.g. Wiktionary. Edits would happen in that environment, i.e. anyone could edit any entry.

   Pros:

   • a well-established environment, with excellent prospects for ongoing support and availability;
   • high and growing visibility, with the prospects of gathering a large(r) support community.

   Cons:

   • it may be very difficult to maintain a tight entry format. Wiktionary is more oriented to what you see on the screen and less to generation of an accurately marked-up data set.
   • the Wiktionary editing process is rather clumsy, and more oriented to free-form text than tightly defined fields.
   • Copyright. Wiktionary uses the GNU Free Documentation License, which is more oriented to things like software manuals. It has no provision for mandating attribution of source in followup applications of the files, for example.

   Problems:

   • actually getting such a move accepted within Wiktionary. Wiktionary is currently a set of monolingual subprojects. It currently has no formal structure for bilingual dictionaries.
   • getting the EDICT/JMdict data structure accepted. It is very much tailored to the issues associated with Japanese lexicography, and may not find acceptance in a a wider framework.
   • download of the files and conversion into the EDICT, JMdict, etc. forms may be a major task.