Separating Named Entities


ULIPOVÁ Barbora GRÁC Marek

Type Article in Proceedings
Conference Eighth Workshop on Recent Advances in Slavonic Natural Language Processing
MU Faculty or unit

Faculty of Arts

Field Linguistics
Keywords text corpus; mutual information; named entities
Description In this paper, we analyze the situation of long sequences of mostly capitalized words which look like a named entity but in fact they consist of several named entities. An example of such phenomena is hokejista (hockey player) New York Rangers Jaromír Jágr. Without splitting the sequence correctly, we will wrongly assume that the whole capitalized sequence is a name of the hockey player. To find out how the sequence should be split into the correct named entities, we tested several methods. These methods are based on the frequencies of the words they consist of and their n-grams. The method DIFF-2 proposed in this article obtained much better results than MI-score or logDice.
Related projects: