The title of the project is Decoding Hidden Legacy, and it is the ongoing digitization. Of manuscripts from major collections as part of this project, which will be discuss in this blog post. The main manuscript collection is estimat to contain about a few pages of manuscripts. Making it one of the largest collections of folklore material in Western Europe. This would have been a major challenge to. Digitize if the technology for handwriting recognition, a technology on which. Decoding Hidden Legacy relies heavily, had not been develop over the past few. Years and allow the project to make its own handwriting recognition engine.
Having spent a few days
From the collection on a specific scribe basis. Since South Sudan Email List our researchers start working with the software at the beginning of the month, we have manag to produce three known handwriting recognition engines that are as accurate as one for Ó á, one for á Ó and one for three for the committee Work for the most passionate collectors. Ó á is collecting folklore from á á, and the manuscripts he wrote bas on the recordings of Ó í ú, which were transcrib to Ó í ú. Accurate transcriptions faithful to the collector’s manuscript page handwriting and dialects requir to train the engine.
This means that it is not
It turns out that transcribing about BS Leads half a page in this way is enough to train a handwriting model, and the accuracy rate ( ) is quite satisfactory. The approach of the following project is to automatically transcribe a large number of pages with the help of a team of students and research assistants (íó and óí) and gradually correct them, and retraining the language model on a wider database gave us a more accurate language model . Learning curves using the language model made so far: á Ó á (left) á Ó (middle) and right.
