New pathways towards cultural heritage
We know how to digitize our heritage, so what is the next step: making our Cultural Heritage more accessible to the general public / researchers, and even accessible when it is not there anymore.
In recent years, the application of Artificial Intelligence (AI) approaches has increased rapidly in cultural heritage (CH) management and research. A main driver is the availability of remote sensing data, allowing to detect new archaeological sites and to monitor the preservation of known monuments. Due to advances in computer power and a wide range of free machine learning tools, large amounts of remote sensing data can be processed automatically for CH purposes instead of covering only small areas by expert inspection.
But AI may also be applied for other tasks in cultural heritage research including automated classification of archaeological pottery or bones from excavations, classification of object images in CH collections, symbol and text recognition in ancient inscriptions, detecting relevant terms (often consisting of several words) in site report repositories with limited metadata, mining historical texts, expert systems in restoration, knowledge representation by ontologies, simulation of crowds in buildings (past and present: e.g. museums, prehistoric caves, palaces). Mixed reality apps using AI technology as well as Ambient Intelligence approaches support the creation of new pathways towards CH for the public. CH may also benefit from robotics with integrated AI applications, e.g. vehicles searching for sites in inaccessible areas such as unmanned submarines used for detecting archaeological remains in lakes and the sea.
“Is it possible to build a machine to do archaeology? Will this machine be capable of “interpreting” and “explaining” cultural heritage?” (Juan A. Barceló, Computational Intelligence in Archaeology. State of the Art, CAA 2009)
This conference will showcase best-practice AI applications but also discuss the potential and limits of various AI approaches such as the amount of labelled data required.