2 Replies Latest reply on Oct 21, 2015 4:56 PM by 3054700

Full-Text Search

3054700 Oct 21, 2015 12:03 PM

I am using the Oracle-specific orardf:textContains SPARQL FILTER function to perform full-text searches on the MDSYS.RDF_VALUE$ table.

Following the Spatial and Graph RDF Semantic Graph Developer's Guide 12c to create the index:

EXECUTE SEM_APIS.ADD_DATATYPE_INDEX('http://xmlns.oracle.com/rdf/text');

It works as expected but I would like to be able to perform "accent-insensitive" searches. I found that when creating the index with "CREATE INDEX", a custom lexer can be passed in parameters. Something like that:

BEGIN
CTX_DDL.create_preference( 'my_lexer', 'BASIC_LEXER');
CTX_DDL.set_attribute('my_lexer', 'BASE_LETTER', 'true');
CTX_DDL.set_attribute('my_lexer', 'OVERRIDE_BASE_LETTER', 'true');
END;
/

CREATE INDEX my_index on TABLE (COLUMN)
  indextype is ctxsys.context
  parameters ( 'LEXER my_lexer' );

I cannot figure out how to achieve the same result when creating the index with "SEM_APIS.ADD_DATATYPE_INDEX". Is it possible?

Thank you

1. Re: Full-Text Search

Matperry-Oracle Oct 21, 2015 12:48 AM (in response to 3054700)

Hi,

This customization is not supported through add_datatype_index, but you can directly create a text index on MDSYS.RDF_VALUE$ and use that with orardf:textContains in SPARQL queries. That is, you do not have to go through add_datatype_index to create the index. As long as VNAME_PREFIX is included in the set of columns that are indexed, then orardf:textContains should work.

For example:

SQL> create index mdsys.rdf_v$text_idx on mdsys.rdf_value$(vname_prefix) indextype is ctxsys.context;

Index created.

SQL> select s, o
from table(sem_match(
'SELECT ?s ?o
WHERE { ?s rdfs:label ?o FILTER(orardf:textContains(?o, "New York")) }'
,sem_models('lgd_historic')
,null,null,null,null
,' '));
2 3 4 5 6 7
S
----------------------------------------------------------------------------------------------------------------------------------------------------------------
O
----------------------------------------------------------------------------------------------------------------------------------------------------------------
http://linkedgeodata.org/triplify/node358960285
25th New York Volunteer Cavelry Monument

http://linkedgeodata.org/triplify/node879249295
New York Regiment

http://linkedgeodata.org/triplify/node358239660
New York State Monument

http://linkedgeodata.org/triplify/node1251645997
New York Life Insurance Company War Memorial

http://linkedgeodata.org/triplify/node1242148250
New York Vietnam Veterans Memorial

http://linkedgeodata.org/triplify/node2971826546
Glocke aus New York

http://linkedgeodata.org/triplify/node358960283
122nd New York Volunteer Mounment

http://linkedgeodata.org/triplify/node1040277511
New York State Vietnam War Memorial

http://linkedgeodata.org/triplify/node1251645998
New York Life Insurance Company Memorial

http://linkedgeodata.org/triplify/node427545363
New York Central Roundhouse

10 rows selected.

SQL>

Hope this helps,
- Matt
Like Show 1 Likes(1)

Actions
2. Re: Full-Text Search

3054700 Oct 21, 2015 4:56 PM (in response to Matperry-Oracle)

Thank you Matt it worked as suggested.

For the record, to make it work I created the lexer and the index with mdsys user. I also set the sync(on commit) parameter, I assume it is reasonable since the index created by SEM_APIS.ADD_DATATYPE_INDEX is created that way (impact on performance will be tested later though).

SQLPLUS 'mdsys/<MDSYS_PASSWORD>'
SQL> BEGIN
CTX_DDL.create_preference('ai_lexer', 'BASIC_LEXER');
CTX_DDL.set_attribute('ai_lexer', 'BASE_LETTER', 'true');
CTX_DDL.set_attribute('ai_lexer', 'OVERRIDE_BASE_LETTER', 'true');
END;
SQL> CREATE INDEX MDSYS.RDF_V$TEXT_IDX ON MDSYS.RDF_VALUE$(VNAME_PREFIX)
indextype is ctxsys.context
parameters ('lexer AI_LEXER sync (on commit)');
Like Show 1 Likes(1)

Actions

Go to original post

Join the world’s largest interactive community dedicated to Oracle technologies.