Artificial intelligence “unlocks” the secrets of ancient texts

Artificial intelligence "unlocks" the secrets of ancient texts
© naftemporiki.gr Intelligence “unlocks” the secrets of ancient texts

The library of St. French Abbey in Switzerland houses some 160,000 volumes of historical and literary manuscripts dating back to the 8th century; many of which have been written in languages rarely spoken today; and such historical archives are kept in libraries and monasteries around the world. 

Much of this material is available to the public through digital images and technology, but it is too much and the material has never been read.

In this context, scientists at the “University of Notre Dame” are developing a network of artificial neurons to read complex ancient handwritten text, based on human perception to improve the capabilities of the so-called “deep learning transcription”. 

“We are dealing with historical documents written in styles that have been out of ‘fashion’ for a long time, perhaps centuries, and in languages like Latin, which are rarely used anymore,”

said Walter Shirer, a professor in the University’s Department of Computer Science.

“You can take nice pictures… but what we underway to do is to automate the transcription in a way that mimics the perception of the page through the eyes of the expert reader and provides a quick, “searchable” reading of the text”.

The relevant research is presented in the “Transactions on Pattern Analysis and Machine Intelligence” of the “Institute of Electrical and Electronics Engineers”. It presents the combination of traditional machine learning methods with optical psychophysics; a method of measuring the connections between physical stimuli and spiritual phenomena, such as the time it takes for a particular reader to recognize a particular character, assess the quality of the writing, or determine the use of some abbreviations.

Shirer’s team studied digitized manuscripts in Latin written in the 9th century; readers entered their hand transcriptions to special software and measured the reaction times during the transcription to understand which words/characters/excerpts were difficult or easy; this data created a network closer to human mode of operation, reduced errors and led to a more accurate, more realistic reading of the text.

Researchers are now trying to improve the accuracy of transcripts, especially in cases of documents that are incomplete or damaged. 

Source:

naftemporiki.gr

Σᾶς ἀρέσει τὸ ἂρθρο; / Do you like this post?
+1
0
+1
0
+1
0
+1
0
+1
0
+1
0
+1
0
(Συνολικές Επισκέψεις: / Total Visits: 10)

(Σημερινές Επισκέψεις: / Today's Visits: 1)

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.