Re: [edict-jmdict] RE: parsing submission data

["Jim Breen" <jimbreen@*********>]

G'day all,

Back home now, and a huge amount of catching-up to do...

On 12/07/07, Stuart McGraw <smcg4191@frii.com> wrote:

> Hope you {are having / have had} a very enjoyable trip, Jim.

Great. Wait till you see the pictures (about 1500 of them - my wife has
a new camera). I managed to survive two weeks without even looking at email.

>  On June 20, 2007, Jim Breen wrote:

>  > - for setting up the tagging on kana, we could have something like
>  > [rinf=ik,restr=KANJI] or [rinf=ik,restr="KANJI1,KANJI2"]
>
>  In other places in the language (e.g. note="...", lsrc="...", etc.),
>  quoted strings are treated as single lexical tokens.  It seemed
>  more consistent to maintain that behaviour consistently thoughout
>  the language.  It is also simpler to do it that way because if not,
>  then either one has to add yet another state to the lexer (i.e.
>  make it behave differently when parsing a quoted string following
>  a "restr=" or "xref=" tag than in other places), or do 2 levels of
>  parsing: use a second parser to reparse the quoted string when
>  it occurs in restr or xref value.
>
>  > -I wonder about having semicolons inside the [....] structures
>
>  Without the quotes (which were awkward as I mentioned above)
>  commas were a problem because in "adv,restr=str1,str2,adj" for
>  example, is "str2" part of the restr list, or an independent tag
>  like "adj"?  One could look at "str2" and disambiguate depending
>  on whether it contained japanese characters (since all restr and
>  xref values will have japanese text values and none of the other
>  tags do) but using something other than a comma seemed simpler:
>  "id,restr=str1;str2,adj" is not ambiguous.  Semicolons seemed best
>  for this since that what is used to separate the kanji and kana items
>  in the kanji and reading sections so there is some degree of
>  consistency in that choice.

Yes, that's quite clear. Semicolons make a lot of sense there.

>  > - I guess there could be an implied [1] at the start of the senses box. An empty
>  > box could even be preloaded with it.
>
>  Already done :-)  Before I left, I changed the database on Arakawa
>  to add an Edit button to each displayed entry.  The Edit and New Entry
>  (http://www.edrdg.org/~smg/cgi-bin/edform.pl) pages were changed
>  to use the input language parser and I did what you suggest -- the
>  Sense box is preloaded with "[1][n]".
>
>  The Help links on those pages don't work yet but the language is the
>  same as described in the original test page:
>    http://www.edrdg.org/~smg/cgi-bin/jbparser.pl
>  and the pages work -- you can add and edit entries using them (although
>  I didn't have time to give them any real testing).  I'll start working on this
>  again in a few days (just got home yesterday and need a few days to
>  unwind :-)

I know the feeling. I need to get my Cyrcadian rhythm sorted out too. Waking at
3am is not great.

Jim
--
Jim Breen
Honorary Senior Research Fellow
Clayton School of Information Technology,
Monash University, VIC 3800, Australia
http://www.csse.monash.edu.au/~jwb/