I am currently working on an ontology driven text mining project. We are searching for a way to integrate our developed ontology in KNiME in order to get access to the concepts described and henceforth be able to annotate the documents based on the ontology classes.
I would like to kindly ask for pointers on a suitable plugin that will allow us to integrate the ontology with KNiME or any sort of software that could enable us to work with the ontology within the KNiME environment.
the Network Mining plugin (http://tech.knime.org/network-mining) allows for the representation and handling of networks within KNIME. An ontology can be represented as a network or graph. Concepts could be represented as nodes in the network and their relations as edges. Labels can be assigned to edges and nodes as features.
There are already KNIME nodes that allow for the creation of networks, as well as filtering basic mining and visualizations. Have a look at: http://tech.knime.org/network-mining to learn more about the plugin.
I believe there is a special relation between the concepts within ontologies, which makes them interesting in searching: inheritance; though other relations -where transitive closure is interesting (for example "contained")- also require usually some kind of (usually computationally intensive when combined with features) inference. Are these tasks supported by the network mining nodes? (I see the Hierarchy Extractor node as a candidate, but I am not sure it is capable of.)
the network mining feature offers general network mining features. You could store different relation types e.g. part-of, contained, etc. using different edge partitions and uses this information later on e.g. by filtering certain partitions etc.. However specialized functions such as transitive closures etc. are not supported.
The Hierarchy Extractor node can be used to extract a hierarchical tree from a given property member set. An example could be a bag of words with the terms as property and the documents that posses a term as its members. The node would then arrange the terms in a hierarchical tree from the most general terms to the most specific once.