Language Agnostic Data-Driven Inverse Text Normalization

  • 2023-01-20 10:33:03
  • Szu-Jui Chen, Debjyoti Paul, Yutong Pang, Peng Su, Xuedong Zhang
  • 4

Abstract

With the emergence of automatic speech recognition (ASR) models, convertingthe spoken form text (from ASR) to the written form is in urgent need. Thisinverse text normalization (ITN) problem attracts the attention of researchersfrom various fields. Recently, several works show that data-driven ITN methodscan output high-quality written form text. Due to the scarcity of labeledspoken-written datasets, the studies on non-English data-driven ITN are quitelimited. In this work, we propose a language-agnostic data-driven ITN frameworkto fill this gap. Specifically, we leverage the data augmentation inconjunction with neural machine translated data for low resource languages.Moreover, we design an evaluation method for language agnostic ITN model whenonly English data is available. Our empirical evaluation shows this languageagnostic modeling approach is effective for low resource languages whilepreserving the performance for high resource languages.

 

Quick Read (beta)

loading the full paper ...