• First, set up psql. A simple set up is discussed in LexDbSimplePsql (the details which will work for single-user and multi-user mode are in LexDbPsqlInitialize)

• Then, initialize a database, populate it with the lexicon entries from a lexicon.tdl file and tell the LKB to use this when you load a grammar. Step-by-step instructions follow. These instructions presuppose that you already have suitable mappings from the paths in the typed feature structures to database fields in a .dfn file. The easiest approach is to use an existing .dfn file as a model. More details are given in LexDbInitialize.

The instructions as given below are suitable if you want to use the ERG with a Lex DB in single user mode.

[a version of these instructions may be found in lkb/lexdb/README.single-user]

$ export MyLexDB=erg
$ createdb $MyLexDB

[The database is initialized using the SQL script file 'su-init.sql' which may be edited first to specify the Lex DB fields you want. See LexDbFieldMappings for further information. ]

$ psql $MyLexDB
=> \i su-init.sql

[now load the field mappings...]

=> \copy dfn from lexdb.dfn
=> \q

[NOW load your grammar, as normal, in the LKB, then open a connection to the currently empty LexDB...]

LKB(): (initialize-lexdb :type :single-user :dbname "erg")

[and export your lexicon to dump files for the database. The files will go into your LKB tmp directory. ]

LKB(): (export-lexicon)

[import the dump file into the database...] [NOTE: encoding of lexicon/dump files must be UTF-8]

$ psql $MyLexDB
=> \copy lex from /home/bmw20/tmp/LinGO.rev

[Now set the parameters in the LKB appropriately, via the menu]

LKB->Options->Set Options->*LEXDB-PARAMS* = ((:dbname "erg") (:type :single-user))

[Edit the LKB script file to specify that the database should be used, if necessary. The ERG script is already set up to do this, but see LexDbInitialize if you are working with a grammar which has not previously been used with the Lex DB. ]

[NOW reload grammar in LKB, and you should see a messages similar to the following: ]

...
(LexDB) connected to LexDB erg2@localhost:5432 as database user bmw20
(LexDB) total 'lex' entries available: 23055
...

[now we initialize the table of orthography keys, which allows the Lex DB to be used for parsing ...]

LKB(): (new-lex-key-table *lexicon*)

[and the database SEMI, for indices for generation (although it will still be necessary to index the lexicon before it is possible to generate) ...]

LKB(): (new-semi *lexicon*)

THAT'S IT

If anything goes wrong, and you need to start over, you can clear the database by:

$ dropdb $MyLexDB

Description:

• - lexicon entries should be made available in 'lex' (which can be a table or a view) - the mapping from database fields into AVM path-values in specified in 'dfn' - a table 'lex_key' is generated by '(new-lex-key-table *lexicon*)' and is used for lexicon lookup - a number of tables (and a view) 'semi_*' are generated by '(new-semi *lexicon*)' and store the database object SEMI - you are responsible for ensuring that there exists an appropriate index on the 'name' field of 'lex' (or things will run slow...) - you are free to define 'lex' as a view (eg. you could store a large number of lexicon entries in another table 'all_entries', create a table 'active'to specify which entry 'name's you want in view, then define 'lex' as:
  • CREATE VIEW lex AS active JOIN all_entries USING name;