Program CAA Session 24
CAA 2018 (Tübingen)
Session 24 Computational classification in archaeology
Date: 2018-03-20 Venue: Room 6, first floor
S24, Module A
Chair: Oliver Nakoinz
- 10:30-10:50 Hinz/Nakoinz: Computational classification in archaeology - an introduction
- 10:50-11:10 Roth: Considerations on archaeologically meaningful Distance Measures for Multivariate
- 11.10-11:30 Zeckey/Langner: 3D Shape Analysis of Ancient Terracottas: Contributions to Automated Object Mining
- 11:30-11:50 Hinz/Heitz: Classifying vessel shapes using automated shape extraction and unsupervised classification
- 11:50-12:10 Lăzărescu/Mom: Digital Pottery Studies of 4th Century Necropolises on the Moldavian plain
12:15 LUNCH BREAK
S24, Module B
Chair: Martin Hinz
- 13:45-14:05 Mehltretter/Langguth/Sieverling/Goesele/Lang: Ceramic profiler – a new tool for classifying of bulks of objects
- 14:05-14:20 Philippe: Partitioning Archaeological Chaînes Opératoires (PACO) on R : how to deal with huge incomplete categorical datasets?
- 14:20-14:40 Yaman: A Model-Based Statistical Classification Analysis for Karamattepe Arrowheads
- 14:40-15:00 Taylor: Computer-aided Classification of British Isles Neolithic Group VI Axes from Photographic and Scanned Image Data
- 15:00-15:20 Bustillo/Arnaiz/Rodrigue: Using Data Mining techniques to evaluate Bronze Age metallic components
15:30 COFFEE BREAK
S24, Module C
Chair: Moritz Mennenga
- 16:45-17:05 Gennaro: Multispectral image analysis applied to identification of archaeological buried remains
- 17:05-17:20 Nenova: Space, typology and mobility: using multivariate clustering technique to explore spatial patterning in prehistoric Thrace
- 17:20-17:40 Cardinal: Matrix in the Network: assemblage co-expression networks to unlock meaning in stratigraphic matrices
- 17:40-18:00 Discussion
Session Abstract
To make inferences on the archaeological material that go beyond the individual object we always have to decide what is similar or equal and what is not. This reasoning is at the heart of the archaeological method since its beginning and describes what we understand as meaningful categories such as a type and what we try to achieve with a typology. We group and label objects on the basis of more or, in most cases, less defined criteria. Predominantly this is still done in an ‘impressionistic’ or ‘ intuitive’ manner since more ‘objective’ and ‘standardised’ methods, combined with automated recording of the artefacts, have not found a wider reception within archaeology. The reason for this might be that most approaches are considered to be complicated, general or reductionistic. What is a classification good for if it can not inter-subjectively evaluate how similarity and dissimilarity are defined, if the typology can not be duplicated by other studies due to imprecise specification, and a new object can not easily be integrated into the original classification? These issues are perfectly well handled when a statistical, computer based classification is applied. Especially the growing interest in pattern recognition, machine learning and the classification of information emerged within the last years, led by major information processing companies (eg. search engines and social networks). Many new and interesting approaches to this topic were developed that hopefully find their way into archaeological reasoning. In our session we would like to survey the current state-of-the-art research for the classification of archaeological datasets. The aim of the session is to provide a better understanding of classification methods and algorithms and of validation techniques since a sound methodological knowledge is required choose the right approaches among many competing approaches. In particular, a tight connection between method and theory which is essential for a valid interpretation of the results has to be based on this kind of knowledge rather than on methodological fashions. We would like to welcome presentations on recent applications of machine learning, clustering approaches, and related regression methods in the field of archaeology. Presentations will explore methods for evaluating the accuracy of classifications, and investigating the implications of different classification methods for archaeological interpretation and understanding. Reports on how to deal with the challenges of applying modern computational methods to sparse and problematic archaeological datasets will also be included in this session. The session is organized by the ISAAKiel group (Initiative for Statistical Analysis in Archaeology Kiel: https://isaakiel.github.io/)
Session Organizers
Oliver Nakoinz, Martin Hinz, Benjamin Serbe, Caroline Tietze, Ben Marwick, Christoph Rinne, Clemens Schmid, Daniel Knitter, Franziska Faupel, Karin Kumar, Moritz Mennenga, Nicole Grunert, Nils Müller-Scheeßel, Wolfgang Hamer, Camille Butruille, Dirk Seidensticker, Joana Seguin