This node assigns a named entity tag to each term of a document. It is applicable for English, German and Spanish texts. The built-in tagger models are models created by the Stanford NLP group:
http://nlp.stanford.edu/software/ .
You can use the StanfordNLP NE Learner to create your own model based on untagged documents and a dictionary and forward the model to the second input port of this node. If there is no input model, the "use model from input port" option will be deactivated. The other way around, if there is a model at the input port and the optionis activated, the StanfordNLP model selection will be disabled.

Note: The provided tagger models vary in memory consumption and processing speed. Especially the distsim models have an increased runtime, but mostly a better performance as well. There are also models without distributional similarity features. For the usage of these models it is recommended to run KNIME with at least 2GB of heap space. To increase the head space, change the -Xmx setting in the knime.ini file.