Skip to main content
2021 29th Telecommunications Forum, TELFOR 2021 - Proceedings202129th Telecommunications Forum, TELFOR 2021; Virtual, Belgrade; Serbia; 23 November 2021 through 24 November 2021; Category numberCFP2198P-CDR; Code 176031

Recurrent Neural Networks and Morphological Features in Language Modeling for Serbian(Conference Paper)

  • Pakoci, E.T.,
  • Popovic, B.Z.
  Save all to author list
  • aAlfaNum Speech Technologies, Novi Sad, 21000, Serbia
  • bUniversity of Novi Sad, Faculty of Technical Sciences, Department For Power, Electronic and Telecommunication Engineering, Novi Sad, 21000, Serbia
  • cDepartment For Music Production and Sound Design, Academy of Arts, Alfa BK University, Nemanjina 28, Belgrade, 11000, Serbia
  • dComputer Programming Agency Code85 Odzaci, eleznicka 51, Odzaci, 25250, Serbia

Abstract

This paper describes the current state-of-the-art language model for the Serbian language, and also a specific way of dealing with one of the issues that is present in Serbian automatic speech recognition systems. This issues arises from the fact that Serbian is a highly inflective language, which leads to recognition mistakes where the basic form of the word is guessed correctly, but the word ending is not. The mistaken word ending can indicate, for example, a wrong word case, grammatical gender or grammatical number. A way to solve this problem was to use additional lexical and morphological word features as neural network input while assessing the likelihoods of predicted words. The experiments have shown significant reduction in word error rates when employing certain combinations of these additional features. © 2021 IEEE.

Author keywords

Automatic speech recognitionDeep neural networksKaldiLanguage modelingMorphological tagging

Indexed keywords

Engineering controlled terms:Computational linguisticsDeep neural networksModeling languagesSpeech recognition
Engineering uncontrolled terms'currentAutomatic speech recognitionAutomatic speech recognition systemKaldiLanguage modelMorphological featuresMorphological taggingNeural network featuresSerbian languageState of the art
Engineering main heading:Recurrent neural networks
  • ISBN: 978-166542584-1
  • Source Type: Conference Proceeding
  • Original language: English
  • DOI: 10.1109/TELFOR52709.2021.9653410
  • Document Type: Conference Paper
  • Publisher: Institute of Electrical and Electronics Engineers Inc.


© Copyright 2022 Elsevier B.V., All rights reserved.

Cited by 3 documents

Sečujski, M. , Popović, B. , Pekar, D.
Retrospective and Perspectives of TTS & STT Technology Development and Implementation for South Slavic Under-Resourced Languages
(2025) Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Floriano, A.C. , Avila, S.L. , Fernandes, R.C.
Structured vocabulary specific to power operation control centers
(2024) Energy Systems
Pakoci, E. , Pekar, D. , Popovic, B.
Overcoming Data Sparsity in Automatic Transcription of Dictated Medical Findings
(2022) European Signal Processing Conference
View details of all 3 citations
Inform me when this document is cited in Scopus:
Set citation alert Set citation feed
{"topic":{"name":"Computational Linguistics; Modeling Language; Speech Recognition","id":10377,"uri":"Topic/10377","prominencePercentile":51.398098,"prominencePercentileString":"51.398","overallScholarlyOutput":0},"dig":"38dcaa273b7abf1d0457544ca6a1d5dc490b76547303e933c806daf05a642dda"}

SciVal Topic Prominence

Topic:
Prominence percentile: