Update README.md

README.md

@@ -63,4 +63,14 @@ TRIDIS was trained using a encode-decoder architecture based on a fine-tuned ver
 This final model operates in a multilingual environment (Latin, Old French, and Old Spanish) and is capable of recognizing several Latin script families (mostly Textualis and Cursiva) in documents produced circa 11th - 16th centuries. 
 
 During evaluation, the model showed an accuracy of 94.3% on the validation set and a CER (Character Error Ratio) of about 0.06 to 0.12 on three external unseen datasets 
-and a WER of about 0.14 to 0.26 respectively, which is about 30% lower compared to CRNN+CTC solutions trained on the same corpora.
+and a WER of about 0.14 to 0.26 respectively, which is about 30% lower compared to CRNN+CTC solutions trained on the same corpora.
+
+### Other formats
+A CRNN+CTC version of this model trained on Kraken 4.0 using the same gold-standard annotation is available in Zenodo: 
+
+Torres Aguilar, S., & Jolivet, V. (2024). TRIDIS: HTR model for Multilingual Medieval and Early Modern Documentary Manuscripts (11th-16th) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10800223
+
+### Paper
+A journal paper presenting the scientific basis of this models is also available:
+
+Torres Aguilar, Sergio, Jolivet, Vincent . La reconnaissance de l'écriture pour les manuscrits documentaires du Moyen Âge, Journal of Data Mining & Digital Humanities, 22 décembre 2023 - https://hal.science/hal-03892163/document