1. [Publications](/publications)
2. A Chat about Boring Problems: Studying GPT-Based Text Normalization
 
 # A Chat about Boring Problems: Studying GPT-Based Text Normalization

  ![Publication image](/sites/default/files/styles/wide/public/default_images/default.jpeg?itok=qUFsuJCP "Publication image")

 Text normalization - the conversion of text from written to spoken form - is traditionally assumed to be an ill-formed task for language modeling. In this work, we argue otherwise. We empirically show the capacity of Large-Language Models (LLM) for text normalization in few-shot scenarios. Combining self-consistency reasoning with linguistic-informed prompt engineering, we find LLM-based text normalization to achieve error rates approximately 40% lower than production-level normalization systems. Further, upon error analysis, we note key limitations in the conventional design of text normalization tasks. We create a new taxonomy of text normalization errors and apply it to results from GPT-3.5-Turbo and GPT-4.0. Through this new framework, we identify strengths and weaknesses of LLM-based TN, opening opportunities for future work.



 ## Authors



Yang Zhang (NVIDIA)

Travis M. Bartley (University of New York)

Mariana Graterol-Fuenmayor (NVIDIA)

Vitaly Lavrukhin (NVIDIA)

Evelina Bakhturina (NVIDIA)

Boris Ginsburg (NVIDIA)

 

 

 ## Publication Date



Monday, March 18, 2024

 

 ## Published in



[ICASSP](https://ieeexplore.ieee.org/xpl/conhome/10445798/proceeding)

 

 ## Research Area



[Machine Translation](/research-area/machine-translation)

[Natural Language Processing](/research-area/natural-language-processing)

[Speech Processing](/research-area/speech-processing)

 

 

 ## External Links



[Paper](https://ieeexplore.ieee.org/abstract/document/10447169)