LiLFeS / will

• First, install LiLFeS. Get the latest version of LiLFeS and untar the archive. Type ./configure; make; make install and you will get the executable and manuals installed. See LiLFeS Users Manual Section 2. Installation (in Japanese) for details.

• Next, install the graphical user interface will (You can install it in another machine). Java 2 is required to be installed. Get the latest version of will and untar the archive. You can run will by executing will.bat (in case of Windows) or will.sh (in case of Unix).

• Finally, let's play with a toy grammar and test whether LiLFeS works well.
  Download --> sample-grammar19990722.tar.gz (43KB)
  
  1. Expand the archive in some directory. Let me install one in ~/sample.

     > mkdir sample
     > cd sample
     > tar zxvf sample-grammar19990722.tar.gz

  2. Set the environment variable "LILFES_PATH". You can specify the path to the LiLFeS modules installed by separating each by colon (:). In this case,

     csh> setenv LILFES_PATH ~/sample

     
     Note) Since /usr/local/share/lilfes (depends on the installation options of LiLFeS) is the default path for searching modules, you don't have to set "LILFES_PATH" when you installed the grammar in this directory.
  3. Run sample_unary.perl, and LiLFeS starts to load the sample grammar.

     > cd HPSG/sample
     > perl sample_unary.perl
     lilfes -l sample_unary -l HPSG/nvparser/all -l will
     LiLFeS 1.082           by Takaki Makino and Tsujii lab., Tokyo Univ.
     Initialize lildbm.cpp
     Initialization file loaded.
     yes
     >

  4. Start willfirst, and let LiLFeS connect to will.

     > ?- will("host name or IP address where will is running").

     If the connection succeeds, will prints the welcome message in a status bar.
  5. Let's input "he runs" in the Input column (The sample grammar supports the words such as "he", "Mary", "runs", "loves", "the", "dog's", so you can test the parser by inputting sentences with these words. If the parsing succeeds, will prints the results. See the parsing analysis by clicking the "Feature Structure" button and the "Parse Tree" button.

English Grammar (RenTAL version)

Download
LTAG grammar converter --> rental_package20010109.tar.gz (7.5KB)
Converted grammar --> grammar.tar.gz (2.1Mb)

The converted grammar was obtained by converting the large-scale LTAG English grammar, The XTAG English Grammar (Jan. 2000, pre-release version) developed by XTAG group in University of Pennsylvania.

1. Get the archive and expand it. Let me expand it in ~/rental/.

   > mkdir ~/rental
   > cd ~/rental
   > tar zxvf rental20010109.tar.gz
   > tar zxvf grammar.tar.gz
   > ls -F
   Artistic  COPYRIGHT  HPSG/  grammar/

   Artistic shows the license of RenTAL (Artistic License), and COPYRIGHT for the converted grammar (GPL).
2. Set the environment variable "LILFES_PATH". You can specify the paths to LiLFeS modules by separating them by colon (:). In this case,

   csh> setenv LILFES_PATH ~/rental

   
   Note) Since /usr/local/share/lilfes (depends on the installation options of LiLFeS) is the default path for searching modules, you don't have to set "LILFES_PATH" when you installed the grammar in this directory.
3. Execute ~/rental/HPSG/rental/rental.prl in ~/rental/grammar/, and LiLFeS start to load the English grammar.

   > cd ~/rental/grammar/
   > perl ~/rental/HPSG/rental/rental.prl
   lilfes -l HPSG/rental/naive -l grammar_pref -l base -l spec -l lex -l
    templates_define -l toplabels -l templists -l HPSG/rental/interface
    -l HPSG/nvparser/all  -l HPSG/rental/xihpsg
   LiLFeS 1.082           by Takaki Makino and Tsujii lab., Tokyo Univ.
   Initialize lildbm.cpp
   Initialization file loaded.
     SUCCESS: open_lexicon
   yes
   >

   Now, start will, and type

   > ?- will("host name or IP address where will is running").

   then the LiLFeS is connected to will. You can parse a sentence by putting an English sentence in the Input column, can see feature structures and parse trees of parsing results. We can terminate LiLFeS by Ctrl-D or evaluating a predicate 'halt/1' (:- halt(0).).

English Grammar (XHPSG)

See the XHPSG Home Page for details of the XHPSG system. (We recommend you to use the English Grammar (RenTAL version) because the development of XHPSG has already been stopped.)

1. Download and expand the archive. Let me install it in ~/xhpsg.

   > mkdir ~/xhpsg
   > cd ~/xhpsg
   > tar zxvf xhpsg20000111.tar.gz
   > ls -F
   Artistic  HPSG/

   Artistic is the license agreement of the XHPSG system.
2. Next, make lexicon. You should confirm that LiLFeS & Perl 5 are put in the command search paths.

   > cd HPSG/xhpsg/lexicon/
   > make

   This instruction will take a long time. Don't mind that the program will print large amount of error messages.
3. Set the environment variable "LILFES_PATH". You can specify the paths to LiLFeS modules by separating them by colon (:). In this case,

   csh> setenv LILFES_PATH ~/xhpsg

   
   Note) Since /usr/local/share/lilfes (depends on the installation options of LiLFeS) is the default path for searching modules, you don't have to set "LILFES_PATH" when you installed the grammar in this directory.
4. Execute ~/xhpsg/HPSG/xhpsg/XHPSG, and LiLFeS starts to load the English grammar.

   > cd ~/xhpsg/HPSG/xhpsg/
   > ./XHPSG
   lilfes -l preproc/naive -l grammar/interface -l HPSG/nvparser/all
    -l util/parsetree -l util/utility -l HPSG/parser/parsetest -l will/xixhpsg
   LiLFeS 1.082           by Takaki Makino and Tsujii lab., Tokyo Univ.
   Initialize lildbm.cpp
   Initialization file loaded.
     SUCCESS: open_lexicon
   yes
   >

   Now, start will and type

   > ?- will("host name or IP address where will is running").

   then LiLFeS is connected to will. You can parse a sentence by putting an English sentence into the Input column, and see feature structures and parse trees of parsing results. You can terminate LiLFeS by Ctrl-D or evaluating the predicate 'halt/1'(:- halt(0).).
5. You will get the Japanese manual by executing `make' in ~/xhpsg/HPSG/xhpsg/doc/manual/.

Japanese Grammar (SLUNG)

Now preparing.

Packages and modules

Packages

• Sample English Grammar ( sample-grammar19990722.tar.gz (43kb))
  A toy grammar with a several word lexicon. For details, see the above documents.
• LTAG Grammar Converter RenTAL ( rental_package20010109.tar.gz (75kb))
  A grammar converter from LTAG to HPSG. For details, see the above documents.
• English Grammar XHPSG ( xhpsg20000111.tar.gz (10Mb))
  A large-scale English grammar. For details, see the above documents.

• parser ( parser-module.tar.gz (17kb))
  Common interface for HPSG parsers running on LiLFeS, and tools using the common interface.
• nvparser ( nvparser-module.tar.gz (8kb))
  CHP: An HPSG parser running on LiLFeS. (parser module is required to use this module)
• tnt ( tnt-module.tar.gz (16kb))
  TNT parser: An efficient HPSG parser with CFG filtering running on LiLFeS. This module includes a parser, and a compiler from HPSG to CFG. (parser module is required to use this module)
• sample ( sample-module19990722.tar.gz (5kb))
  An English toy grammar. (nvparser or tnt module is required to use this module)
• rental ( rental20010109.tar.gz (37kb))
  A grammar converter RenTAL including preprocessors. (nvparser or tnt module is required to use this module)
• grammar ( grammar.tar.gz (2.1Mb))
  An English grammar converted by RenTAL. (rental module is required to use this module)
• xhpsg ( xhpsg-module20000111.tar.gz (10Mb))
  An English grammar XHPSG. (nvparser or tnt module is required to use this module)

• LiLFeS: An efficient inference engine for processing typed feature structure logic
• RenTAL: LTAG grammar converter
• XHPSG: An English Grammar
• SLUNG: A Japanese Grammar

• Tsujii Laboratory