The researchers have been awarded $548,000 through the Digging into Data Challenge, an international competition administered by the U.S.-based National Endowment for the Humanities and funded through 10 agencies in four countries. Of the total, nearly $125,000 for the project will flow from Social Sciences and Humanities Research Council of Canada (SSHRC).
Entitled Digging Archaeology Data: Image Search and Markup (DADAISM), the project's goal is to create an online tool that archaeologists can use to search and analyze images and "grey text" - unpublished information such as field notes. DADAISM is a collaboration of three teams, led by Eramian in Canada, University of York computer science Professor Helen Petrie in the UK, and computer science Professor Maarten de Rijke at the University of Amsterdam in the Netherlands.
A crucial part of the project will be to collaborate with archeologists, both for their input and their help in testing the system. The first test images will be of amphorae, large vase-shaped containers used to ship products across the Roman Empire, with the hope that DADAISM can reveal insights to inform long-running debates about the scale and structure of the imperial economy.
Eramian explained data banks are so large it could be the work of several lifetimes for an archaeologist to find specific answers. This is where DADAISM comes in.
Images are typically searched by keyword, which has serious limitations. These searches depend on metadata, that is, descriptive text that is added to the image files when they are uploaded.
"You type in key words and hope images similar to what you're looking for are tagged with appropriate metadata, but that depends on who put them in and if anybody added metadata at all," Eramian explained. "The quality of the metadata is poor, so the quality of the text-based searches is poor."
Eramian's contribution is in image analysis, something he has applied to everything from high-resolution ultrasound of the reproductive tract and microscope images of cancer cells, to technology to help farmers seed crops more efficiently. For DADAISM, the challenge will be to create an image-based search that recognizes shapes, even if they are only partially there or seen from different angles.
"That involves, first of all, what part of the image actually contains the object, and then analyzing things like shape, colour, texture - basically its appearance," he said. "Then you're boiling that down into kind of a digital fingerprint - sort of a unique signature."
DADAISM will also be designed to improve itself. For example, an archaeologist might search for amphorae. From dozens of results, the researcher would choose perhaps three relevant images and DADAISM would automatically record the search terms and which images were found to be connected, essentially "learning" each time an archaeologist uses the tool.
The team will assemble an archaeology advisory group, led by Julian Richards, director of the Archaeology Data Service with the University of York. Eramian plans to recommend colleagues in the U of S Department of Archaeology for this role as well.
DADAISM is among 14 projects awarded a total of $5.1 million through the Digging into Data Challenge, and one of eight projects to include Canadian university researchers. The competition is funded by 10 international organizations, including three top Canadian research funding agencies: SSHRC, the Canada Foundation for Innovation, and the Natural Sciences and Engineering Research Council.
For more information, contact:
University of Saskatchewan