Riva faces difficulty in encoding a specific word (original) (raw)

Hardware - GPU (T4)
Hardware - CPU (n1-standard-4)
Operating System - Linux
Riva Version - 2.8.1
TLT Version (if relevant)
How to reproduce the issue ? (This is for errors. Please share the command and the detailed log here)

Currently we are using Riva ASR to infer our audios. I attempted to enhance specific words using the method outlined in this article:
https://docs.nvidia.com/deeplearning/riva/user-guide/docs/tutorials/asr-python-advanced-wordboosting.html#asr-inference-with-word-boosting

I provided a word list to enhance the ASR transcription for specific words, which worked successfully for all words except one: ডিপিএসের. Unfortunately, an error was encountered during the process, displaying the following message:
E0531 07:38:56.434091 91 ctc-decoder.cc:328] Inference failed in ASR decoder: Unable to encode the word ডিপিএসের

It is important to note that no errors occur when using the base word ডিপিএস, but the error arises when attempting to use ডিপিএসের. I tried using this word by normalizing with NFKC, NFKD, NFD. But none of these approaches resolved the issue.

I would greatly appreciate receiving your insights and responses on this matter. Thank you.

After adding the word in the lexicon file, the encoding issue was solved.

system Closed December 17, 2025, 4:46am 3

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.