Semantic wiki

A semantic wiki is a wiki that has an underlying model of the knowledge described in its pages. Regular, or syntactic, wikis have structured text and untyped hyperlinks. Semantic wikis, on the other hand, provide the ability to capture or identify information about the data within pages, and the relationships between pages, in ways that can be queried or exported like a database.

Semantic wikis were first proposed in the early 2000s, and began to be implemented seriously around 2005. As of 2010, the best-known semantic wiki software may be Semantic MediaWiki, while the best-known standalone semantic wiki may be Freebase.

Key characteristics

Formal notation

The knowledge model found in a semantic wiki is typically available in a formal language, so that machines can process it into an entity-relationship or relational database.

The formal notation may be included in the pages themselves by users, as in Semantic MediaWiki; or it may be derived from the pages or the page names or the means of linking. For instance, using a specific alternative page name might indicate a specific type of link was intended. This is especially common in wikis devoted to code projects. It should be easy to examine and fix, e.g. to identify problems in parsing and conventions introduced by newer users.

In either case, providing information through a formal notation allows machines to calculate new facts (e.g. relations between pages) from the facts represented in the knowledge model.

Semantic Web compatibility

The technologies developed by the Semantic Web community provide one basis for formal reasoning about the knowledge model that is developed by importing this data. However, there are also a wide array of technologies that work on ERD or relational data.

Example

Imagine a semantic wiki devoted to food. The page for an apple would contain, in addition to standard text information, some machine-readable or at least machine-intuitable semantic data. The most basic kind of data would be that an apple is a kind of fruit - what's known as an inheritance relationship. The wiki would thus be able to automatically generate a list of fruits, simply by listing all pages that are tagged as being of type "fruit." Further semantic tags in the "apple" page could indicate other data about apples, including their possible colors and sizes, nutritional information and serving suggestions, and so on. These tags could be derived from the text but with some chance of error - accordingly they should be presented alongside that data to be easily corrected.

If the wiki exports all this data in RDF or a similar format, it can then be queried in a similar way to a database - so that an external user or site could, for instance, request a list of all fruits that are red and can be baked in a pie.

Use in knowledge management

Where wikis replace older CMS or knowledge management tools, semantic wikis try to serve similar functions: to allow users to make their internal knowledge more explicit and more formal, so that the information in a wiki can be searched in better ways than just with keywords, offering queries similar to structural databases.

Some systems are aimed at personal knowledge management, some more at knowledge management for communities. The amount of formalisation and the way the semantic information is made explicit vary. Existing systems range from primarily content-oriented (like Semantic MediaWiki) where semantics are entered by creating annotated hyperlinks, via approaches mixing content and semantics in plain text, via content-oriented with a strong formal background (like KiWi), to systems where the formal knowledge is the primary interest (like Metaweb), where semantics are entered into explicit fields for that purpose.

Also, semantic wiki systems differ in the level of ontology support they offer. While most systems can export their data as RDF, some even support various levels of ontology reasoning.

History

In the 1980s, before the Web began, there were several technologies to process typed links between collectively-maintained hypertext pages, such as NoteCards, KMS and gIBIS. Extensive research was published on these tools by the collaboration software, computer-mediated communication, hypertext, and computer supported cooperative work communities.

The first known usage of the term "Semantic Wiki" was a Usenet posting by Andy Dingley in January 2001. Its first known appearance in a technical paper was in a 2003 paper by Austrian researcher Leo Sauermann.

Many of the existing semantic wiki applications were started in the mid-2000s, including Semantic MediaWiki (2005), Freebase (2005) and OntoWiki (2006).

June 2006 saw the first meeting dedicated to semantic wikis, "SemWiki2006", co-located with the European Semantic Web Conference in Montenegro.

The site DBpedia, launched in 2007, though not a semantic wiki, publishes structured data from Wikipedia in RDF form, which enables semantic querying of Wikipedia's data.

In March 2008, Wikia, the world's largest wiki farm, made the use of Semantic MediaWiki available for all their wikis, thus allowing all the wikis they hosted to function as semantic wikis.

Semantic wiki software

Category: Semantic Wiki

• ArtificialMemory - Personal Semantic Wiki and Argumentation System
• Freebase (database) – web database with semantic wiki like properties
• Metaweb - the software that powers Freebase
• OntoWiki
• PhpWiki
• Semantic MediaWiki - an extension to MediaWiki that turns it into a semantic wiki
• TikiWiki CMS/Groupware integrates Semantic links as a core feature
• Wagn - Ruby on Rails-based

Challenges

Classification

Although adding link-types to a wiki is straightforward, the number of link types can often be quite large. The Cyc system has over 15,000 different types of links. In order to create the right type of link a set of questions is often used to create the correct link type. Rules can also be added to check that the destination page is appropriate for that link type. For example a link of "capital_of" might only be appropriate when linking a city to a region or a country.

Argument

Most early wiki-like technologies, specifically gIBIS and NoteCards, had direct support for online deliberation, often including argumentation frameworks such as logic trees or more elaborated issue/position/argument structures that were flexible enough to support meeting agendas and support actual online meetings.

In these uses, the link types are deliberately limited to simplify presentation and also to avoid anyone gaining advantage in the debate or meeting by knowing the types better (often thought to be a primary reason why ordinary users strongly resist typed links). For instance, a link to an assertion that "contradicts" another or which "supports" it.

The challenge is to create new adversarial process designs that deal with the power imbalances and rapid pace of change in online forums, and support these with new tools. The open politics theory, for instance, developed some of these for use in politics.

Rights and restrictions

Another range of problems arises from the larger scale and open nature of a wiki. In the paper "Re-use rights and their relation to customization", the NeOn project deals with questions about customization or adaptation especially in "those cases when someone wants to re-use work of the other people." Questions include:

• "How much detail can one re-use?"
• "Can the level of detail be customized as a particular view of essentially the same ontology?"

Physical location and GPS data are cited as examples where "satellite companies want to differentiate between e.g. the paying customers and free services. The former may be able to express their location in the terminology of 'streets', 'floors', 'rooms'; whereas the latter would see only more general views – 'city', 'country', or 'unavailable'."

User rights

Another focus is on issues of authentication (‘is the user X who she claims she is’), of encryption ('how to prevent information misuse'), and, to a smaller extent, of authorization (‘can user X do a particular action’). The NeOn paper states that "in practice the vocabulary of rights is much richer (e.g. fine-grained, data-, not action-specific authorization levels)." A "knowledge owner (our GPS user) may differ from the knowledge provider (our GPS satellite), and each of them may want to restrict how third parties experience shared knowledge." Thus different "views" may need to be provided for different types of users.

Common features

Semantic wikis vary in their degree of formalization. Semantics may be either included in, or placed separately from, the wiki markup. Users may be supported when adding this content, using forms or autocompletion, or more complex proposal generation or consistency checks. The representation language may be wiki syntax, a standard language like RDF or OWL, or some database directly populated by the tool that withdraws the semantics from the raw data. Separate versioning support or correction editing for the formalized content may also be provided. Provenance support for the formalized content, that is, tagging the author of the data separately from the data itself, varies.

What data can get formalized also varies. One may be able to specify types for pages, categories or paragraphs or sentences (the latter features were more common in pre-web systems). Links are usually also typed. The source, property and target may be determined by some defaults, e.g. in Semantic MediaWiki the source is always the current page.

Reflexivity also varies. More reflexive user interfaces provide strong ontology support from within the wiki, and allow it to be loaded, saved, created and changed.

Some wikis inherit their ontology entirely from a pre-existing strong ontology like Cyc or SKOS, while, on the other extreme, in other semantic wikis the entire ontology is generated by users.

Conventional, non-semantic wikis typically still have ways for users to express data and metadata, typically by tagging, categorizing and using namespaces. In semantic wikis, these features still typically exist, but integrated these with other semantic declarations, and sometimes with their use restricted.

Some semantic wikis provide reasoning support, using a variety of engines. Such reasoning may require that all instance data comply with the ontologies.

Most semantic wikis have simple querying support (such as searching for all triples with a certain subject, predicate, object), but the degree of advanced query support varies; some semantic wikis provide querying in standard languages like SPARQL, while others instead provide a custom language. User interface support to construct these also varies. Visualization of the links especially may be supported.

Many semantic wikis can display the relationships between pages, or other data such as dates, geographical coordinates and number values, in various formats, such as graphs, tables, charts, calendars and maps.

See also

• Microformats
• Ontology
• RDF, RDFS, OWL, SPARQL
• Business Intelligence 2.0 (BI 2.0)

External links

• Semantic wiki article at SemanticWeb.org
• Semantic wiki projects - contains a list of active, defunct and proposed semantic wiki applications
• SemWiki.org – homepage of the semantic wiki community and workshop series
• SemanticWiki mini-series - homepage of the virtual mini-series jointly organized by FZI Karlsruhe, Mayo Clinic, Ontolog, RPI Tetherless World Constellation and Salzburg Research, Austria between Sep-2008 and Mar-2009.