Artificial Intelligence interaction in artefact identification

Archaeological experts from the United States and Canada recently were on the Idaho State University campus to learn how artificial intelligence could aid in artifact identification.

Experts in the nationally recognized ISU National Information Assurance Training and Education Center (NIATEC) enabled researchers from different academic disciplines to design and implement applications for an artificial intelligence classification system. The system is called SIGGI, a nickname for Sigmoid Archaeological Automated Classification System. It mimics the ways humans think, comprehending large numbers of variables that in human minds would require time-consuming analysis.

With SIGGI programmed and operational, archaeologists working in the field could photograph unearthed objects like arrowheads or spear tips, send the pictures to SIGGI and receive an accurate classification report in only a few seconds.

For more about SIGGI see the dedicated Object Classification website. It has a Flash animation to introduce you to the programme - a talking arrowhead, which personally I find more than a little irritating, and that maybe because I used to develop software for a living, but could equally be be due to the fact that last night's visit to the opera was such a dissapointment. Was it really necessary to set the entire first act of Poppea on and around a diving board and to clothe half of the cast in bright pink cellophane? I digress. In spite of talking to you as though you are three years old Siggi-the-talking-arrowhead actually does quite a thorough, if long-winded, job of explaining how the programme works. The "Use Siggi" link refused to do anything at all in Firefox but work fine in Explorer. The link launches a web-based interface with the following introductory text:

The SIGGI – AACS project is a prototype of an expert system for the identification of archaeological materials. To test the validity of the concept, projectile points were selected as the initial type of object to identify. To generate classifications SIGGI, the computerized agent, utilizes an artificial neural network that has been trained to identify projectile point types from three distinct regional typologies:

• Columbia Plateau
• Upper Snake River Basin
• Northwest Plains

Almost as important as the actual identification of the archaeological materials is the user interface developed for the interactions between the computer and the user. A set of interfaces have been designed for varying levels of users ranging from the general public to dedicated research professionals.

More about this and related projects can be found at the Informatics Research Institute website.