The IHR-Nom project

The IHR-Nom project is our project for preserving an old Vietnamese language called ChuNom. ChuNom had been using to record the history of Vietnam for more than a thousand years. As results, there is a huge amount of ChuNom documents that had been writing thoughout many generations and many of those have been collected by Vietnamese agencies and world-wide foundations. However, over 90% of the collected documents have not yet been translated into modern Vietnamese language, and very few of them have been text-digitized. Furthermore, according to the Vietnamese Nom Preservation Foundation, less than 100 scholars world-wide can read Nom script today. Therefore, it is extremely necessary to develop a system for text-digitization and futhermore, translation from ChuNom to modern Vietnamese language.

Example images of Nom documents and vertical text image segmentation

This project is about applying new technologies of deep learning for ChuNom text-digitalization and translation to modern Vietnamese language.


The IHR-NomDB is a database for ChuNom language in handwritten, which is described on the paper "IHR-NomDB: The Old Degraded Vietnamese Handwritten Script Archive Database".
The database consists of two parts including the handwriting part and the Synthetic Nom String part. Totally, 34,991 instances of handwritting characters and 704, 991 instances of generated characters are distributed over 106,622 columns of text characters for the whole database. The underlying lexicon includes 13,254 different ChuNom characters. The database described in this paper is freely available for researching purpose.

To obtain this database, you agree with the following terms of use:

  • The dataset and annotations are available for non-commercial research purposes only.
  • You agree with all terms of use, copyrights & restrictions from
  • You agree not to reproduce, duplicate, copy, sell, trade, resell or exploit for any commercial purposes, any portion of the images and any portion of derived data.
  • You agree not to further copy, publish or distribute any portion of annotations of the dataset. Except, for internal use at a single site within the same organization it is allowed to make copies of the dataset.

There are four parts of the database, which is including:

  1. The Handwritten: contains the handwritten dataset download
  2. The Synthetic Nom String dataset without augmentation download
  3. The Synthetic Nom String augmented dataset download
  4. The Natural Language Processing dataset: contains ChuNom sentences and its translation download