AFNER - Named Entity Recognition Logo


Afner is a package for named entity recognition. It is written in C++. A combination of approaches is used in order to find named entities. A list of named entities is created and each method adds the found named entities to the list.

Firstly, regular expressions are used to find simple case named entities such as simple dates, times, speeds, etc. Secondly, parts of text matching listed named entities are found. The regular expression and list matches are then used in a maximum entropy based classifier. Features relating to individual tokens (including list and regular expression matches) as well as contextual features are used.

The lists used are of person names, organisation names, locations, and miscellaneous entities. The lists were obtained from the CNTS language independent named entity recognition specification of the on the 2003 CoNLL shared task.