This is a demo of the CRF-based name parser. Below are some examples of rare name forms. Overwhelmingly, most name forms are much simpler. Subsequent post-processing can be applied to these results to utilize additional dictionaries or rules. Click some samples below or type in the input box to try you own name tests. The immediate goal of this parser is to facilitate entity resolution. Subsequent improvement could yield presentation quality results for even very rare name forms.

This parser was created using my implementation of a boosted-tree CRF classifier based on methods developed by Thomas Dietterich described in the paper Training Conditional Random Fields via Gradient Tree Boosting. The code can be found here.