WikiTrust Online Mode

Display text trust, and analyze edits in real time

The WikiTrust online mode is what you need if you want to add a trust tab to your wiki, and make information on text trust available to all visitors. The tool can be installed as a MediaWiki extension; please see the README files in the tool itself for detailed installation instructions.

The tool can be deployed either on a brand-new, empty wiki, or on an existing wiki containing articles and revisions.

If installed on a new, empty wiki, the tool will analyze edits as they are made by users, and assign reputation to authors, and trust to the text. The trust of any article (including past revisions) can be viewed simply by clicking on a trust tab added by the WikiTrust extension.

If installed on an existing wiki, WikiTrust will first analyze the existing wiki content. This make take some time, depending on the size of the wiki; the task is performed in the background, so normal wiki editing can proceed while WikiTrust catches up with the past of the wiki. On our PC, with raid disks, WikiTrust running on a single CPU processes about 40 revisions / second. The code is completely parallel, so that many instances of WikiTrust are run automatically, with load-management to avoid database overload; thus, if your installation has many apache/php servers, the analysis will speed up proportionally. WikiTrust performs the analysis in a fault-tolerant, self-healing way: if for any reason the processes are killed or interrupted (for example, if the database connection is lost), the analysis will simply restart from where it left. Equally, if holes in the analysis are discovered (for example, because revisions are artificially inserted in the past), WikiTrust will realize that some analysis is missing, and it will analyze the overlooked revisions.

WikiTrust stores author reputation, text trust, and text origin information in additional database tables, that do not interfere with the database tables normally used by MediaWiki.

The author reputation system and the text trust system are resistant to attack, including Sybil attacks. Please see our published work for more details on the algorithms we use.

Comments