Text Recognition for South Asian Scripts - Using Transkribus

AI-based approaches to text recognition have taken OCR for historical documents a big step forward, especially for documents written in non-Latin scripts. Since 2019, the Naval Kishore Press - digital project has been using the Transkribus platform for the entire recognition process (layout analysis, text recognition, post-processing) for the Devanagari texts in the collection. The project "Hindu Temple Legends in South India" conducted at the Heidelberg Academy of Sciences and Humanitites has trained text recognition models with Transkribus for printed Tamil/Grantha texts. In the first part of the talk Nicole Merkel-Hilf will present the workflow and results from the NKP project by using the Transkribus Expert Client. In the second part of the talk, Sowndarya Sriraman will present the workflow and the results for Tamil and she will also give a live presentation on how to use the Transkribus' web application.

Zoom Link