Hello everyone,
this is the first in a series of posts dedicated to Open Source AI (Artificial Intelligence) applied to archaeology. During my work, I have often found myself, lately, working with various AI tools.
Yesterday, for example, I was engaged in the recovery of old typewritten documentation regarding Lake Antholz/Anterselva, where Alessandro Bezzi and I will soon be involved in some underwater archaeological explorations (as part of a research project on the lakes of South Tyrol, in Northern Italy, entrusted to Arc-Team). To expedite my work, I used Open Source software called Transkribus, developed by the University of Innsbruck (Austria). In short, Transkribus is "an AI-powered platform for text recognition, transcription, and searching of historical documents – from any place, any time, and in any language."
The image below is an example of how easy it is to use Transkribus, thanks to the web interface developed by the University of Innsbruck. As you can see, I just had to upload the scan of the document I had available (unfortunately of low quality and with strong distortions on the left margin) to obtain a fairly faithful transcription in just a few seconds. Transcription errors are mainly due to deformations in the document (unfortunately, I didn't have anything better available). Nonetheless, even with my limited knowledge of German, I managed to correct the text and obtain a good translation into Italian (my native language) thanks to other AI tools (which I will discuss in other posts).
This post is primarily dedicated to OCR (Optical Character Recognition) and HTR (Handwritten Text Recognition), two branches of research that help us archaeologists (and obviously historians and archivists, more qualified in this) quickly read ancient documents. Transkribus can perform both OCR and HTR, and indeed I have often used it for reading cursive documents (for example, related to World War I or archaeological documentation from the late 19th to early 20th century, but these will be topics for other posts).
What I want to convey here is that AI is forcefully entering (for some time now) the world of archaeology (I will soon provide more examples in other posts), even in areas where it seems unlikely to happen, and there are often many Open Source options to use these "new" tools. For instance, there is Kraken, "a turn-key OCR system optimized for historical and non-Latin script material," released under the Apache 2.0 license, very useful, for example, for documents in Arabic or Greek (but not only). Kraken itself is linked to the "eScriptorium" platform, also developed with open-source code by various partners, including the University of Paris (France).
If you are interested in Open Source tools for OCR and HTR, I recommend starting with eScriptorium and from this page, where you can upload your first documents to run tests.
Have a nice day!
No comments:
Post a Comment