semantic wikis

Wikis are so useful because they strike the right balance between making it easy for a purveyor of information to write down something new, or improve on something someone else wrote, while still allowing readers to view the content in a web-accessible, platform-neutral way. They don’t impose restrictions on the structure or format of the information being entered.

Yet some well-known wiki flavors, such as MediaWiki, do make use of some level of self-imposed structure. For example, the information in Wikipedia, the best known MediaWiki instance, is logically ordered into pages that each represent an encyclopedia entry. Each entry itself will typically have a logical structure in its writing style, as most good writing does. The structure contains sections and sub-sections which are hierarchical in nature. Editing of a page can be done at a page level, section level, or sub-section level, because the wiki has an informal notion of structure.

Like most web-based applications, their power lies in the ability to easily link to other pieces of information on the web. Standard wikis provide the same linking capability as other applications — the link has a target and a name associated with it.

Imposing some structure in the information clearly provides benefits. The text itself, though, is still in free form. While this is powerful, some additional categorization would be useful. Some wikis (including MediaWiki) provide the ability to tag entries into categories. These categories provide metadata about the entries, which can assist with filtering and searching of data. The categories can increase relevancy when searching.

Even simple categories are great, but entries could be improved further by adding some other types of semantic information to the text that makes up an entry. Relations are a way to add semantic information to links. In other words, you can tie together not just the link with its name, but also describe how they are related. Attributes are a way to add semantic information to a statement.

Semantic wikis introduce the use of categories, relations, and attributes into a regular wiki structure. Wiki entries written with these additional semantic capabilities allow computers to do powerful filtering and searching of data within the wiki.

Last night, I had the opportunity to have dinner with York Sure, who has been thinking a lot about the Semantic Web. He is at the University of Karlsruhe in Germany, where some of his grad students have been working on a project called Semantic MediaWiki. As its name suggests, the goal is to add semantic constructs to the existing MediaWiki software. They have a working implementation already, and are running an instance of their own software. The software is released under the GPL license, which is compatible with the MediaWiki software itself.

Unlike many other academic projects, they are serious about improving their prototype version and turning it into a fully scalable, robust piece of software. They are working to reach the point where they can prove that their semantic additions are ready for prime time in a production system. When this occurs, there is a good chance that they could be incorporated into the main MediaWiki codebase itself. Of course, convincing Wikipedia writers (and other MediaWiki instances) to then make use the semantic capabilities on a wide scale is the ultimate goal.

We had the opportunity to discuss some of the work being done by the HyperScope project as well as the goals of earlier projects such as PurpleWiki. Adding Purple Numbers to the Semantic MediaWiki would be a great way to augment the existing semantic linking capabilities of the wiki by adding higher resolution addressing. Having Node Identifier support in a wiki implementation that is focused on scalability and is supported by a small team of full-time programmers would provide the opportunity for eventual wide-scale usage of high resolution addressing.