Joint Language Identification of Code-Switching Speech using Attention based E2E Network

  • 2019-07-15 06:30:15
  • Sreeram Ganji, Kunal Dhawan, Kumar Priyadarshi, Rohit Sinha
  • 1

Abstract

Language identification (LID) has relevance in many speech processingapplications. For the automatic recognition of code-switching speech, theconventional approaches often employ an LID system for detecting the languagespresent within an utterance. In the existing works, the LID on code-switchingspeech involves modelling of the underlying languages separately. In this work,we propose a joint modelling based LID system for code-switching speech. Toachieve the same, an attention-based end-to-end (E2E) network has beenexplored. For the development and evaluation of the proposed approach, arecently created Hindi-English code-switching corpus has been used. For thecontrast purpose, an LID system employing the connectionist temporalclassification-based E2E network is also developed. On comparing both the LIDsystems, the attention based approach is noted to result in better LIDaccuracy. The effective location of code-switching boundaries within theutterance by the proposed approach has been demonstrated by plotting theattention weights of E2E network.

 

Quick Read (beta)

loading the full paper ...