Abstract

Named Entity Recognition (NER) for Myanmar Language is essential to Myanmarnatural language processing research work. In this work, NER for Myanmarlanguage is treated as a sequence tagging problem and the effectiveness of deepneural networks on NER for Myanmar language has been investigated. Experimentsare performed by applying deep neural network architectures on syllable levelMyanmar contexts. Very first manually annotated NER corpus for Myanmar languageis also constructed and proposed. In developing our in-house NER corpus,sentences from online news website and also sentences supported fromALT-Parallel-Corpus are also used. This ALT corpus is one part of the AsianLanguage Treebank (ALT) project under ASEAN IVO. This paper contributes thefirst evaluation of neural network models on NER task for Myanmar language. Theexperimental results show that those neural sequence models can producepromising results compared to the baseline CRF model. Among those neuralarchitectures, bidirectional LSTM network added CRF layer above gives thehighest F-score value. This work also aims to discover the effectiveness ofneural network approaches to Myanmar textual processing as well as to promotefurther researches on this understudied language.

 

Introduction (beta)

None

 

Conclusion (beta)

None