README about ispell eo dictionary

About this File
About the Package
The Ispell utility program
Esperanto encodings
Simple Install (Unix or Cygwin)
Open Office
Customized Build
Command line

About this File

This is a short htmlized version of the readme file from the ispelleo.tar.bz2 distribution tarball.

Name: ./readme
Content: Information about Esperanto dictionary for the Ispell speller
Created: 1997-08-30 by Sergio Pokrovskij <sergio.pokrovskij(ĉe)>
Version 3.7

Copyright © 1997, 1998, 2003, 2006, 2008 by Sergio Pokrovskij

This dictionary package is available on the terms of GNU General Public License (Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA).

About the Package

Here is an Esperanto dictionary, compiled by Sergio Pokrovskij for the version 3 of ispell.

The Ispell utility program

You will first need to get ispell 3.0+; it is available from here; make sure that in its local.h the NO8BIT thing is commented out and MASKBITS = 64. To see the options ispell has been compiled with please run

$ ispell -vv

Esperanto encodings

Several encodings used with Esperanto text are supported:

  1. The very best is Unicode or its subset, which contains the esperantic letters, like the Microsoft's WGL4; you can use it with xterm or Emacs under Unix, or with UniRed under Windows.

    Unicode is available in UTF-8 encoding, which is preferred for the Unices.

  2. The second best choice used to be the Latin-3 encoding (ISO-8859-3); it is obsolete by now.
  3. For the sake of the ASCII-impaired (and ANSI-impaired), there are two surrogates:


Simple Install (Unix or Cygwin)

1. Unpack ispelleo.tar.bz2.

2. Go to the root directory ispell-eo (where this readme resides).

3. Say

  $ make first

(or simply make) in order to check your ispell program.

Examine the output, e.g. do you have the permissions to write the hash file(s) at the install phase?

4. If everything is OK, say

  $ make eo

(to built the strict dictionary), or

  $ make esperanto

(to built a permissive dictionary), or

  $ make all

(to built both).

You'll get a few warnings of from buildhash, like this one:

eo.aff line 218: Flag must be alphabetic

Just ignore them.

5. Type

  $ make install

to copy the hash file(s) to where ispell expects them to be (probably you already have your american.hash there; normally you shall need the root rights to make install).

After that you can call

  $ ispell -d eo filename

Open Office

You'll need emacs to produce the dictionary for myspell which works with Open Office. Customize the word provision as described in “Customized Build” (except the buildhash step); and then say

  $ make OO

That should produce the files eo_l3.aff and eo_l3.dic in the work/ subdirectory.

Customized Build

In order to enable selective construction of dictionaries, some entries in the source dictionary ./src/vortoj.l3 are marked with keywords indicating the special field they belong to:

#arhx archaic words, like ĥina (= ĉina) or malkompreni (= miskompreni)
#bot a rare botanic word
#Eujo vocabulary of the Esperanto Movement (of Esperantujo)
#etn countries and ethnography
#komp some computer-science terminology according to the Komputada Leksikono
#mav redundant words, which are used by some esperantists, though they are less precise and unnecessarily complicate the language; e.g. olda (maljuna or malnova), mava (= malbona)
#pers given names and names of important personalities (e.g. Petro, Zamenhof ...)
#pok the words specific to my idiolect
#rar rare words which may coincide with a misspelling of a more frequent word; e.g. ajuna, komanditi, liona
#var variant which I do not use but which is frequent enough (e.g. kemio, tekniko opposed to ĥemio and teĥniko)

You can

  $ grep '#mav' ./src/vortoj.l3 | less
in order to see if you feel like me about them; you can either remove all of them from the target dictionary, or remove the #mav mark from those you do use and like; the default setting in the ./Makefile is
short_list      =  komp,etn,Eujo,pers,mll
eo_list         =  $(short_list),drv
esperanto_list  =  $(short_list),arhx,mav,rar
Unless included in the custom list (like eo_list), a marked word is considered as a special one and is excluded from the build; the above custom lists thus specify “positive criteria”. But some words have several marks (e.g. a word may be “ethnic” and “obsolete” or dangerously close to a misspelling of a frequent word). Thus a “negative” filtration is available via the $(sen) macro; you can specify it in src/Makefile, or in the command-line:
  $ make eo sen="arhx,rar"

In this way some otherwise eligible words from the komp or etn or mll categories shall be discarded, if they are also obsolete or rare.


Command line

You can use ispell in a stand-alone mode; type

  $ ispell -d eo filename

or you may prefer to customize your emacs; e.g. copy ./emacs/ispell-ini.el from this distribution into your site-lisp (or somewhere else on your emacs load-path), and put this into your .emacs

(load "ispell-ini.el")

In order to get a list of all misspelled or unknown words from a text in the Latin-3 encoding you could say (in Linux):

   export LC_ALL=eo_XX.ISO-8859-3
   ispell -d esperanto -T .l3 -l < FILENAME | sort -u

Some more comments are in legu-min.l3 (in esperanto).


This probably is no longer required for emacs-21 or newer.

I enclose the emacs/ispell.el file, which is a modified version of ispell.el which comes with the Emacs-20 distribution. There are two modifications:

  1. the variable ispell-dictionary-alist-3 is modified to use the latin-3 encoding (it is latin-1 in the distribution);
  2. the coordinates of the correction for the word being checked is done in a more thorough fashion. The original version fails to account for the fact that ispell returns multibyte representation, while the internal Emacs representation for the UTF-8 takes one unit per character. As a result, the "misalignment error" occurs, when the word is close to the end of line, and the correction end extends beyond the end of line. With my patch the esperantic texts in UTF-8 are spellchecked normally, unless a foreign character occurs in the buffer; but that is a different bug, which is present for the unibyte encodings as well.