Automated Word Stress Detection in Russian

P. 31–35.

Пономарева М. А., Milintsevich K., Artemova E., Starostin A.

Abstract In this study we address the problem of automated word stress detection in Russian using character level models and no partspeech-taggers. We use a simple bidirectional  RNN with LSTM nodes and achieve the accuracy of 90% or higher. We experiment with two  training datasets and show that using the data from an annotated corpus is much more  efficient than using a dictionary, since it allows us to take into account word frequencies and  the morphological context of the word.

Keywords: нейронные сети, word stress, лексическое ударение, recurrent neural network

Publication based on the results of: Explanation-oriented Methods of Data Analysis for Semantically Rich Data and Their Applications (2017)

In book

Proceedings of the First Workshop on Subword and Character Level Models in NLP

Stroudsburg, PA: Association for Computational Linguistics, 2017.