Woogle4MediaWiki: An Extension to Improve Search and to Guide Contributions

Woogle4MediaWiki An Extension to Improve Search and to Guide Contributions

Hans-Jörg Happel, FZI Karlsruhe, GermanyDenny Vrandečić, Karlsruhe Institute of Technology, Germany

2010-08-05 @ Wikimania 2011; Haifa, Israel

Basic facts

• Woogle is an extension for MediaWiki

• Goal: Guide and motivate contributions – Show people what is „missing“ in the Wiki– Especially in smaller/enterprise Wikis

• Sub-Goal: improve search in MediaWiki

Woogle4MediaWiki - Wikimania 2011; Haifa, Israel

Contribution in Wikis

• Wikis are great in enabling contributions– Everybody may contribute

• But: Wikis are poor in guiding & motivating contributions– „Why should I contribute?“– „Where should I put it?“– „What can I contribute anyway?“


What to contribute to a Wiki?

• „Red links“– Great innovation: „Dead“ links that help to express a

demand rather than a (404) error– But: „Red links“ do not tell how much information is sought– What to fill first, if there are plenty of red links?

Special:WantedPages is useless on most Wikimedia Wikis


Woogle Feature: Red Link Popups (I)

Plain MediaWiki Red Link

• Normal display

• Mouse-over display

Woogle Red Link

• Normal display

• Mouse-over display


Woogle Feature: Red Link Popups (II)

• Lesson learned– Many red links denote synonyms

(abbreviations, misspellings) of existing pages – at least in smaller Wikis

• New feature/extension– „Redirector“ – create a one-click #REDIRECT

based on suggestions


Under development

Help appreciated!Under development

Help appreciated!

Woogle Feature: Search (I)

• Woogle replaces MediaWiki built-in search or can act as an additional/parallel search engine (user-configurable!)

• Small features– Search suggestions based on fulltext– Allows to index Wikitext or rendered text– Allows to index certain files (e.g., PDF)– Show previous search attempts based on query log– Social search (Wiki-page per query, social ranking)

• Two setups– WoogleNative: Native PHP; based on Zend Lucene Search

• For small Wikis – does not scale that good for Wikis >100.000 pages– WoogleRemote: Connect to Java-based search backend

• Scales well• Allows to search multiple Wikis and many other systems (e.g., SVN, JIRA,

Bugzilla, …) at once



http://localhost/wiki/Woogle:Woogle4MediaWiki

Each query can be a Wiki page

• Allows for• Wiki-Text on top of results

(„Social Did-You-Mean“)• Discussion page for query• Watches based on keywords

• Shows basic statistics about query• „How many people searched for that?“

Each query can be a Wiki page

• Allows for• Wiki-Text on top of results

(„Social Did-You-Mean“)• Discussion page for query• Watches based on keywords

• Shows basic statistics about query• „How many people searched for that?“

Woogle Feature: Search (III)

• Social search– Add, remove and re-rank query results– Add comments


Under development

Help appreciated!Under development

Help appreciated!

Woogle Feature: Even more…

• Query log statistics– Popular queries– Popular results

• Preferences might be defined globally or by each individual user (e.g., Woogle can be enabled by individual users)

• Instrumentation framework– Logs certain kinds of user interactions– Experimental groups with randomized user assignment– Customize system and override configuration for groups– Informed consent (user acceptance dialog) Useful for UI research and A/B testing


Further related tools to guide contributions


MediaWiki

SemanticMediaWiki

MediaWiki extension External tool (via API)

Intopedia(intopedia.org)

Intopedia(intopedia.org)

Extension:Woogle4Media

Wiki

Extension:Woogle4Media

Wiki

Extension:SemanticNeed

Extension:SemanticNeed

Summary

• Woogle is an extension to improve search & content contribution in Wikis– Info & Download

http://www.mediawiki.org/wiki/Extension:Woogle4MediaWiki– Online-Demo

http://amazonas.fzi.de/wooglenative/

• What‘s in for Wikimedia projects?– May help bootstrapping smaller Wikis (small

language Wikipedias Wikimedia projects)– Woogle allows to search across multiple

MediaWikis at once


BACKUP SLIDES


Demand guidance• Woogle uses queries and further metadata to qualify desired

content• Users are informed how strong the information need is, and in

which context it arises


Two flavours of Woogle

• Native PHP (using Zend Lucene Search)

• Can handle Wikis up to 50.000 pages (Wikipedia size) and serves normal queries in >1 sec

• Accesses Java (Lucene) backend via Webservices

• Supports searching additional sources (e.g. Ticket systems, SVN, other Wiki engines)


Collaborative description of the information need

Notifications concerningsearch activities

Discussion page

Directly create new knowledge from the search dialog

Search has a URI and can be linked from within the Wiki(one page per search term)

Search across different data sources/types (WoogleRemote)

Additional clues with meta-information about a need

http://localhost/wiki/Woogle:Woogle4MediaWiki


Statistics


Woogle Configuration


Implementation Trivia• Search UI

– Completely intercepts MW search process – indenpendent of MW search interfaces

• Red Links– Realized via JavaScript/jQuery (parses red links, retrieves needs from

backend batch-wise) to avoid patching MW

• Very basic implementation of access rights (via MediaWiki groups)

• Instrumentation addon– Allows to re-configure most system features based on user groups– Provides evaluation participation dialog– Logs various user behaviour data


Woogle4MediaWiki: An Extension to Improve Search and to Guide Contributions

Technology

woogle4mediawiki extension

search dialog search

wikis wikis

israel http

search logs

woogle feature

zend lucene search

search content contribution