Ceratosaurus is a tool designed for the analysis and visualization of Character Error Rate (CER) in text comparisons. It allows users to upload text files, configure analysis parameters, and visualize the results through interactive plots.
The genesis of this project was the HTR school preceding the first SCOOP: Source Codes of the Past conference at Princeton's Institute for Advanced Study (IAS). During a lecture, Jan Odstrčilík (IMAFO, ÖAW) mentioned how beneficial it would be to see Character Error Rate (CER) on a page-by-page basis. His colleague, Martin Roček (FF UK; ÖAW), immediately brought the idea to life by coding a functional prototype in Google Colab during the talk. This tool is the direct evolution of that spontaneous effort.
A crucial question in the age of AI and digitalisation is data privacy. Although CERatosaurus uses server to process your data more quickly and to not overload your device, we do not store any of your data permanently. All uploaded files and generated results are deleted from the server immediately after you refresh the browser tab.
For feedback, questions please reach out to us at martin.rocek@ff.cuni.cz.