Algorithms help unravel the secrets of ancient documents

The texts include fragments of Jewish prayer books from the Land of Israel.

The unique algorithm being used to determine the wording wasdeveloped by BGU computer scientists. The documents are searchedelectronically, letter by letter, for similarities in handwriting whichhelp determine the date and author of the texts.

The documents being deciphered at BGU are degraded texts fromsources such as the Cairo Geniza, the Al-Aksa manuscript library inJerusalem, and the Al-Azar manuscript library in Cairo.

All together, the base consists of "100,000 medievalHebrew codices and their fragments [that] represent the book productionoutput of only the last six centuries of the Middle Ages," BGU computerscience professor Klara Kedem said this week.

Kedem has presented the new algorithm to other universities and research centers so a large number of texts can be deciphered.

The purpose of the project is to classify the handwritten documents and determine their authorship.

One problem is that many of the original Hebrew texts which werefound in the Cairo Geniza have been scratched off, and the parchmentused to write an Arabic text.

Ehrlich, head of the Prayer Research Project at BGU, explained,"There was one book, found in the Geniza and which is now in Italy, butwhich was originally used as a Hebrew prayer book."

However, the book had been rewritten as an Arabic text. "Ouraim was to read the first book and not the second book. So we needed tofind out how the Arab book could disappear and would leave only theHebrew letters of the original book. This is why the computer sciencesand humanities departments at BGU decided to collaborate," Ehrlichsaid.

"Although the texts are in Hebrew, the task of deciphering whatis written is difficult because the historical documents have degradedover time," Kedem said. "Now, the foreground and background letteringare hard to separate and there are smudges on the ink of much of thetext which intensifies the background coloring. Furthermore, ink fromthe alternate side of the document adds blotches to the lettering."

To solve the problem, the algorithm is used to cover the textin a dark grey color, which then highlights lighter colored pixels asbackground space and identifies the darker pixels as outlining theoriginal Hebrew lettering.

There are two separate academic disciplines interested indriving this project forward. First, linguistic specialists seek togain a deeper appreciation of the origins of the Hebrew language.Second, Jewish philosophers are interested in studying ancient forms ofprayer that are thought to be contained in the texts.

Ehrlich is particularly interested in deciphering the textswhich are thought to include ancient prayers, so that the translationscan be added to BGU's The Prayer in Rabbinic Literature DatabaseCollection.

Most of the fragments which contain prayers were found in theCairo Geniza. However, few documents remain fully intact, and "we aretrying to piece them together like a big puzzle," Ehrlich said.

It is unsurprising that prayers were found in the Cairo Geniza,because it was there that Jews buried their sacred material when theyhad no further use for it. Many of the fragments originate from Jewishcommunities in the Land of Israel between the 9th and 11th centuries.The texts arrived in Cairo because there were two big Jewishcommunities in that city. One synagogue followed the Babylonian Jewishtraditions and prayers, and the other followed the Land of Israeltraditions and liturgy.

"It is very difficult to estimate the value of these fragments, but they would be very expensive," said Ehrlich.

Most fragments which have been discovered at the Cairo Genizaare now in museums at Cambridge and Oxford universities, The BritishMuseum and in Israel and Paris. Until now the documents have not beenresearched systematically.

With the new algorithm, researchers hope to create a catalogueof all the texts and piece together the ancient prayers and otherdocuments, including those citing Jewish law.