DanNet

DanNet is a WordNet which is a lexico-semantic network graph that shows how senses of a language relate to others through named relations. One can also think of a WordNet as a kind of machine-readable dictionary.

On wordnet.dk/dannet you can not only browse, but also download the newest versions of multiple data sets.

Unlike a normal dictionary, the definitions of words aren't central; instead, relations to other words are the important part. For example, in DanNet you can see that a Swiss willow is a kind of bush, that a gazebo is located in a garden, that "fiberdrys" is for eating, and that cakes are typically produced by baking and usually made from flour and sugar.

On andreord.nors.ku.dk you can generate the above illustration based on an older version of DanNet. In the future, you will be able to find similar illustrations based on the newest data at wordnet.dk/dannet.

All in all, the wordnet currently contains around 70 thousand concepts kept together by approximately 500 thousand relations.

 

 

 

 

 

 

 

 

 

 

You can download the DanNet resource in two separate formats. By downloading a data set, you accept the CC BY-SA 4.0 license:

  • RDF data set (the primary resource)
    • A single TTL file that should now be considered the main resource. 
  • CSV data set (an alternative format)
    • This ZIP file contains several CSV files as well as CSVW metadata, which collectively can be used to recreate most of the contents of the RDF resource.

Several other RDF data sets can also be downloaded at  wordnet.dk/dannet. These alternative data sets can be used to extend the primary RDF resource with other useful data.

When referencing the resource, please use this article: Bolette S. Pedersen, Sanni Nimb, Jørg Asmussen, Nicolai H. Sørensen, Lars Trap-Jensen og Henrik Lorentzen. DanNet – the challenge of compiling a WordNet for Danish by reusing a monolingual dictionary (pdf). Lang Resources & Evaluation (2009) 43:269–299.

Get linguistic specifications

Please note that these specifications are based on an older version of DanNet. However, they should still allow for understanding the contents of the DanNet resource; just be aware that the names of certain relations/attributes may have changed.