Setup in January 2009 the Brede Wiki contains data from neuroscience, particularly from published neuroimaging peer-reviewed papers. Data is stored in simple MediaWiki templates and it can automatically be extracted and represented in an SQL format. Off-wiki Web-scripts can use the SQL database so items in the wiki can be queried efficiently, e.g., to find close brain activations to a given coordinate. Template content is non-nested and without wiki markup making extraction simple and complete.
for recording data or describing its content more semantically. Exceptions are NeuroLex and NeuroCommons.
We have gained some experience in processing the templates of Wikipedia
(more specifically the journal field of the cite journal template for scientific
citations) and for submitting the processed templates to statistical analyses [
Here I describe a system, the Brede Wiki (http://neuro.imm.dtu.dk/wiki/), that use MediaWiki and its template functionality to record neuroinformatics data. Associated scripts can extract the template content from the entire XML dump of the wiki and automatically construct SQL representation of the data. The SQL database enables more advanced queries than is possible with the present standard MediaWiki. 2
To avoid making data extraction unnecessarily complex and to match with the SQL language the application of MediaWiki templates has been kept simple: Present templates do not nest, have lower-case characters (except for first letter) and wiki markup is avoided within the template field values. The following is an example of an application of the paper template: {{Paper | author1 = Daniela Balslev | author2 = Finn Nielsen | author3 = Olaf B. Paulson | author4 = Ian Law | title = Right Temporoparietal Cortex Activation during
Visuo-proprioceptive Conflict | journal = Cerebral Cortex | volume = 15 | issue = 2 | pages = 166-169 | year = 2004 | pmid = 15238438 | doi = 10.1093/cercor/bhh119 | wobib = 128 }}
In the present wiki one page may contain multiple templates, e.g., a page for a specific neuroimaging paper can contain the paper template as well as multiple templates for brain coordinates: the Talairach coordinate template. Most template definitions format the template content by construction of an infobox as known from Wikipedia. What is somewhat different from Wikipedia is the extensive use of wiki links within the template definitions. For example, for the paper template wiki links to the authors and journal is constructed, with MediaWiki template definitions like [[{{{author1}}}]]. External links are created from the external database identifiers, such as DOI and the PubMed Identifier (PMID). For the templates which usually come in sets, such as Talairach coordinate, each application of a template defines a row in a table.
An example of a page within the Brede Wiki is displayed in Figure 1 on page 4 with the paper template as the upper right infobox and with three formatted applications of the Talairach coordinate template at the bottom. The x, y and z fields of the template can be combined in a query to external specialized coordinate search engines. These links are displayed in the right-most column.
As the templates are not nested relatively simple Perl regular expressions retrieve them with m/{{(.*?)}}/sg and subsequently extract the template name and its content with m/([a-z][a-z0-9]*(?:[ _][a-z0-9]+)*)\s*(\|.*)?/si Spaces are later substituted with underscores. Another regular expression in the same style extracts name-value pairs from each field. The extracted content is written to SQL tables, where the master table is presently defined for SQLite as CREATE TABLE brede(id INTEGER, pid INTEGER, title, tid INTEGER, template, field, fid INTEGER, value);
id is the row identifier, pid an identifier for the wiki page which title is also (redundantly) represented in the title column. tid and template are identifier and name of the template, while field is the field name, fid the field number and value the value, i.e., the actual content of the field. An insert with the author2 field from the above displayed paper template example may be issued as: INSERT INTO brede VALUES(387, 31, ’Right temporoparietal cortex activation during visuo-proprioceptive conflict’, 136, ’paper’, ’author’, 2, ’Finn Nielsen’);
Apart from the master table several other tables are built: One for each template. For the paper template the SQL definition for the corresponding table may look like the following: CREATE TABLE brede_paper(__tid, __pid, __title, _author1, _author2, _author3, _author4, _title, _journal, _volume, _issue, _pages, _year, _pmid, _doi, _wobib);
The final number of columns will depend on the number of different field
names discovered during reading of the XML dump. The table names are prefixed
with brede and the column names with an underscore to avoid clashes between
Brede Wiki names and SQL reserved words. A specialized search engine use the
constructed SQL database when searching for nearby Talairach coordinates to
a query coordinate [
At the time of writing 43 papers were represented in the Brede Wiki, 31 which potentially contain Talairach coordinates. In comparison the Brede and BrainMap databases have presently 186 and 1711 papers, respectively. Apart from templates for papers and Talairach coordinates, the Brede Wiki has also templates for, e.g., brain regions, researchers, subject groups and brain volume results. 3
There are both advantages and disadvantages with the Brede Wiki compared to our previous system. Some of the advantages are online versioning, immediate access to entered data, incremental addition of data, the possibility for free format text descriptions as well as discussion pages. Furthermore the data structure is extensible, i.e., it is relatively easy for editors to add new ‘columns’ (fields).
The wiki does not solve the data entry problem per se. The data in the Brede
Wiki has so far for the most part been entered manually in the raw wiki text. A
small Matlab script can convert results from SPM so they appear in the Brede
Wiki template format. Our small fielded wiki with form entry for personality
genetics association studies [
The present consensus on notability in Wikipedia restricts the information that a Wikipedian can enter: If it is not notable enough another Wikipedian might delete it. Ordinary researchers and individual research articles do not have sufficient notability such that individual pages can be constructed for each of them. When DPpedia relies on Wikipedia then this combination does not suffice for our purpose.
Compared to true semantic wikis the Brede Wiki cannot perform ‘in-wiki’ semantic queries. However, an off-wiki Web-script using the SQLite database can search for Talairach foci extracted from the Talairach coordinate templates, and the wiki and the Web-script link between each others. In the wiki such links are automatically constructed by the template. Similar Web-scripts may be setup that utilize other parts of the extracted data in more complex ways.
In our small fielded wiki of personality genetics we can perform on-the-fly
meta-analysis and data plotting. The vision is that the Brede Wiki can constitute
the basis for large-scale Web-based meta-analyses similar to that of the
(nonwiki) AlzGene database [
The Brede Wiki is one of the first steps in neuroinformatics with Web 2.0 and with a high degree of structured content. It shows the possibility to output structured MediaWiki content to an SQL database.