Full-text search

From semantic-mediawiki.org
Full-text search
Full-text search support for properties which data types use strings of characters or text to store their database tables . e.g. datatype "Page"Holds names of wiki pages, and displays them as a link, datatype "Text"Holds text of arbitrary length, datatype "Code"Holds technical, pre-formatted texts (similar to datatype Text) or datatype "URL"Holds URIs, URNs and URLs, etc.
Keywords
Table of Contents

Semantic MediaWiki 2.5.0Released on 14 March 2017 and compatible with MW 1.23.0 - 1.29.x. adds an experimental support for accessing the full-text capabilities of the relational databases (SQL back-end) for properties which data types use strings of characters or text to store their database tables, e.g. datatype "Page"Holds names of wiki pages, and displays them as a link, datatype "Text"Holds text of arbitrary length, datatype "Code"Holds technical, pre-formatted texts (similar to datatype Text) or datatype "URL"Holds URIs, URNs and URLs, etc. These datatypes use either CHAR, VARCHAR, or TEXT to store their data in the database tables.

This feature is not enabled by default since this feature is still considered experimental. In may be enabled for the wiki with configuration parameter $smwgEnabledFulltextSearchSets whether full-text search support for properties may be used.
  • Support was added for MySQL/MariaDB1 and SQLite2 while PostgreSQL34 is currently not supported.
  • Only SMWSQLStore3 is supported since the SPARQLStore would require the native support of full-text search capabilities by the triple-store.

Requirements[edit]

Features and limitations[edit]

General notes
  • The FT_SEARCH table aggregates search content for datatypes storing their data as BLOB and URI values against an index search is being executed, e.g. datatype "Page"Holds names of wiki pages, and displays them as a link, datatype "Text"Holds text of arbitrary length, datatype "Code"Holds technical, pre-formatted texts (similar to datatype Text) or datatype "URL"Holds URIs, URNs and URLs, etc.
  • Supported operations rely on the relational backend database (MySQL, MariaDB and SQLite)
  • For MySQL and MariaDB databases, IN BOOLEAN MODE is used as default search mode
  • Relevance and scores are not used for any sorting purpose, e.g. as in best match
  • TextSanitizer relies on the "onoi/tesa" library5 to help with the sanitization of text or string elements to provide some text manipulation support as well as a possibility to use language detection if enabled. This library is pre-installed for use by Semantic MediaWiki.
  • Custom stopwords are only applied by the "onoi/tesa" library5 in case the language detection is enabled but MySQL/MariaDB provide their own standard list6 which are enabled by default
  • Starting with Semantic MediaWiki 3.0.0Released on 11 October 2018 and compatible with MW 1.27.0 - 1.31.x.:
    • If the SMW_FIELDT_CHAR_NOCASE option to configuration parameter $smwgFieldTypeFeaturesSets relational database specific field type features is enabled the full-text search only comes into effect for selections using the comparators ~ and !~.7
    • API-module "smwtask"Allows to invoke and execute internal Semantic MediaWiki tasks is used instead of a socket connection via a special page to invoke extra "work" after an update has been completed as part of an independent transaction.8 See also configuration parameter $smwgPostEditUpdateSets how many jobs should be executed as part of a post-edit event.
Notes on Chinese, Japanese, and Korean support (CJK)
  • General CJK support is a challenging endeavour due to text elements to be broken into corresponding tokens that are not separate by spaces
  • The "onoi/tesa" library5 provides some simple Tokenizer's which does not require language detection and will try to provide rudimentary CJK search out-of-the box. This however requires ICU 54+ which is still not being used by MediaWiki as of version 1.29-alpha.
  • Mroonga is a MySQL storage engine and said to be a CJK-ready fulltext search, column store
  • MySQL comes with an optional ngram Full-Text Parser and MeCab Full-Text Parser Plugin.
  • According to this issue, MariadDB is missing those parser plug-ins

Configuration[edit]

Changes to any of the above settings, requires to re-run maintenance script "rebuildFulltextSearchTable.php"Allows to rebuild the full text search data table.

Usage and instructions[edit]

for users
  • Searching contains some examples and descriptions about the available search syntax
for system administrators
  • Indexing describes some methods on how to manually create and update the index table
for developers
  • Technical notes provides some information on the technical implementation, fine-tuning, and performance





References

  1. a b c  Semantic MediaWiki: GitHub pull request gh:smw:1481
  2. a b  Semantic MediaWiki: GitHub pull request gh:smw:1801
  3. ^  Semantic MediaWiki: GitHub pull request gh:smw:1956
  4. ^  PostgreSQL is not supported due to a different index schema (e.g. to_tsvector, to_tsquery) but users interested to make it available are encouraged to have a look at the "MySQLValueMatchConditionBuilder" on how to create a PostgreSQL specific implementation.
  5. a b c  "onoi/tesa" - A small library to help with the sanitization of text or string elements.
  6. ^  https://dev.mysql.com/doc/refman/5.6/en/fulltext-stopwords.html and https://mariadb.com/kb/en/mariadb/stopwords/
  7. ^  Semantic MediaWiki: GitHub issue comment gh:smw:2499:307624826
  8. ^  Semantic MediaWiki: GitHub pull request gh:smw:3318