PoS-tagging and lemmatization with a deep recurrent neural network

Neural networks have been shown to successfully solve many natural language processing tasks previously tackled by rule-based and statistical approaches. We present a deep recurrent network with long short-term memory, identical to engines used in machine translation, to solve the problem of joint P...

Full description

Saved in:
Bibliographic Details
Main Author: Ugray Gábor
Corporate Author: Magyar Számítógépes Nyelvészeti Konferencia (15.) (2019) (Szeged)
Format: Book part
Published: 2019
Series:Magyar Számítógépes Nyelvészeti Konferencia 15
Kulcsszavak:Nyelvészet - számítógép alkalmazása
Online Access:http://acta.bibl.u-szeged.hu/59087
Description
Summary:Neural networks have been shown to successfully solve many natural language processing tasks previously tackled by rule-based and statistical approaches. We present a deep recurrent network with long short-term memory, identical to engines used in machine translation, to solve the problem of joint PoS-tagging and lemmatization in Hungarian and German. Our model achieves comparable or superior results to a state-of-the-art statistical PoS tagger. We are able to enhance the Hungarian model’s performance, as measured on a manually annotated sample unrelated to the initial training corpus, through an additional synthesized dataset.
Physical Description:215-224
ISBN:978-963-315-393-2