Title :
Ladder tagger — Splitting decision space to boost tagging quality
Author :
Paradowski, Mariusz ; Radziszewski, Adam
Author_Institution :
Inst. of Inf., Wroclaw Univ. of Technol., Wroclaw, Poland
Abstract :
This paper describes a part of speech tagger. The tagger is based on a set of probability mixture models. Each mixture model is responsible for tagging of a specific class of words, sharing similar context properties. Probability mixture models contain 25 various mixture components. The tagger is tested on Polish language and compared to other available taggers.
Keywords :
mixture models; natural language processing; probability; Polish language; decision space splitting; ladder tagger; probability mixture models; speech tagger; tagging quality; Context; Indexes; Mathematical model; Smoothing methods; Tagging; Training; Training data;
Conference_Titel :
Computer Science and Information Systems (FedCSIS), 2014 Federated Conference on
Conference_Location :
Warsaw
