Hungarian named entity recognition with a maximum entropy approach

In the analysis of natural language text a key step is named entity recognition, finding all complex noun phrases that denote persons, organizations, locations, and other entities designated by a name. In this paper we introduce the hunner open source language-independent named entity recognition system, and present results for Hungarian. When the input to hunner is already morphologically analyzed, we apply the system together with the hunpos morphological disambiguator, but hunner is also capable of working on raw (morphologically unanalyzed) text.