Computational Intelligence in Archaeology

As some of you may know, my masters topic was “A Computational Intelligence Approach to Clustering Temporal Archaeological Data”. As Masters sometimes do, that topic ended up having to change, removing the archaeology šŸ˜¦ and resulting in just “A Computational Intelligence Approach to Clustering Temporal Data”.

Removing the archaeology also meant removing everything from the literature study relevant to archaeology, except maybe some examples I managed to keep in. This post contains theĀ literature that had to be removed. The post begins giving a short overview of archaeology, explaining the role of computational intelligence in archaeologyĀ and concluding with a variety of examples where computational intelligence has beenĀ used in archaeology.

Overview of Archaeology

Archaeology [1] is a field in humanities which studies the material remains of former societies. The field also aims to preserve and reconstruct these remains in order to find out more about the past. The preservation of the remains allows for the knowledge arising from the discovery to be passed to future generations, as well as for more complete interpretations of the past to be documented.

The archaeological process takes place in five stages [2]. These stages are discovery and investigation, excavation, post-excavation analysis and lab work, dating discoveries and, lastly, reporting and interpretation. An archaeologist will work with large amounts of data during each step, either by using previous datasets or creating new ones. This data often becomes so vast it may lose manageability and introduce difficulties in the interpretation process. For this reason, computers began to be used by archaeologists not only for storing data and as statistical tools, but also for managing and analysing the data [3].

Technology inspired by various fields is widely used by archaeologists today. Ground penetrating radar (GPR) [4] and metal detectors [5] are devices used to map out and detect certain objects buried underground.Ā  Virtual reality [6] and Laser Scanning [7] areĀ  used to document and generate virtual models of archaeological discoveries.Ā  Applications such as Google Sketchup [8] and Photomodeller [6] are also used to model artefacts and structures. Geographical information systems (GIS) [9] andĀ  geographical positioning systems (GPS) [10] are used to handle spatial data. Tablets, such as iPads [11] have previously been used by Pompeii archaeologists to ease the documentation process. Satellite images [12] are used to detect possible site locations. X-ray diffraction [13] and Raman spectrometry [14] are used for analysing the physical composition of an artefact. All these technologies are used with the intention of easing the archaeological process and centralising storage.


The Role of Computational Intelligence in Archaeology

Computational Intelligence (CI) techniques can be used in archaeology to aid the analysis and interpretation of data. The more common use of CI in archaeology is classification [15], which involves categorizing objects and predicting typology according to the attributes of the data.

Archaeologists need to often group similar objects together, such as artefacts manufactured or designed in similar manners [16]. An automated classification of such objects saves the archaeologist time and results in a more accurate and unbiased classification of the data. Certain CI algorithms need to be trained [17] by the archaeologist, using known classifications which may be based on prior research. In the case where prior classifications are unavailable, algorithms which require no prior training in order to perform classification can be used, such as data clustering algorithms.

Object recognition is another growing archaeological interest where CI techniques can be used [15]. The CI technique chosen for this purpose must deal with data that is not purely numeric or textual. For example, the input to the algorithm may be an image and the output may be the position of an object with specific characteristics. In order for the algorithm to be able to process this kind of data, the data must either be converted to a format that can be understood by the algorithm, or the algorithm needs to be adapted to accept such an input format.

CI may also be used in archaeology as a data mining technique in order to determine interesting patterns among archaeological entities [18]. Techniques that can be used for this purpose include artificial neural networks (ANNs), data clustering algorithms and genetic algorithms (GAs) [19]. Visualisations may be performed before, during, or after the use of CI techniques in order to gain a more general and compact view of the datasets being analysed [19].

CI can also be used in robotics as robots can be programmed to be ā€œintelligentā€ using CI algorithms [20], requiring less or no user interaction. Marine archaeologists often use robots which are built to withstand harsh environments [21]. These harsh environments include depths the human body cannot withstand [22] or dangerous sites with limited access. Robots aid the archaeologists in exploring archaeological sites which would otherwise be hazardous to a diver.

In order for the use of CI to be of advantage to archaeologists, the input and output of the CI algorithms chosen must be considered. Archaeologists use various types of data daily, including spatial data [23], temporal data [24], numerical data [25], visual data [26], textual data [26], categorical data [26] and 3D models [27]. The CI techniques used by the archaeologist must be implemented to handle these different types of data and the output must be interpretable by the archaeologist. An algorithm that cannot handle the data that has to be processed is useless, as is output that cannot be understood by a human.


Computational Intelligence for the Analysis of Archaeological Data

Although the discussed algorithms have not previously been used to analyse archaeological data, a variety of other computational methods have been analysed for this purpose. Hodson [28] discusses various examples of cluster analysis methods for the classification of fibulae. The research mentions the use of single-link cluster analysis, average-link cluster analysis, double-link cluster analysis and the K-means clustering approach. Single-link cluster analysis [28] involves joining the two most similar data patterns first, followed by the next two most similar data patterns until all data patterns are joined.Ā  Already linked data patterns merge their entire group with similar data patterns or groups, eventually resulting in a set of N clusters representing the classification of the data.Ā  Double-link cluster analysis [29] is an adaptation of single-link cluster analysis where clusters can overlap by a certain number of data patterns before merging to become one cluster. Average-link cluster analysis [30] allows for a data pattern to join an existing cluster only if the average similarity between all its data patterns is at a predefined level. Lastly, K-means [31] involves selecting an optimal number K of cluster centers, assigning each data pattern to the closest cluster center, calculating the average distance between data patterns in a cluster, and setting the cluster center to this average value.

Van den Dries [32] describes an expert system and an ANN developed for the purpose of use-wear analysis of prehistoric flint tools. The expert system, called WAVES, is a menu based expert system designed to be used by people with limited computer knowledge by having a user friendly and intuitive interface. The system consists of a set of rules, which can be added or removed. The ANN implemented was a multi-layer feed-forward ANN with back-propagation learning. After testing both WAVES and the ANN, van den Dries [32] concluded that the systems would be useful to archaeologists as they may save time, reveal possible analysis deficiencies and be usable by non-experts.

Van den Dries [32] also built a table containing names and descriptions of previous artificial intelligence systems built for the purpose of archaeology. This table included, among others, the following examples:

  • An expert system developed for the purpose of identifying and classifying ceramic beakers.
  • An expert system developed for the interpretation of ancient iconography.
  • A system developed for the purpose of classifying bronze age axes.
  • An expert system for identification of artefacts found in excavations.
  • An expert system for the purpose of analysing bird bones.
  • A rule-based system for the purpose of dating Japanese keyhole tombs.
  • A hybrid \gls{ANN} for the purpose of ageing of artefacts.
  • A pattern-recognition system for classification of Phoenician pottery.

ANNs have also been used extensively in archaeology. Toyota et al [33], Lopez-Molinero et al [34] as well as Tanevska et al [35] have all published research done on the use of a self-organising maps (SOMs) in archaeology. SOMs [36] are unsupervised ANNs where neurons form neighbourhoods and a response across all neurons takes place for each input. Toyota et al [33] used these SOMs to evaluate Marajora ceramic composition, while Lopez-Molinero et al [37] used SOMs to classify ancient Roman glazed ceramics and Tanevska et al [35] to determine the origin of the terra cota icons found in the Vinicia Fortress. Alternatively, Barcelo and Fauna [38] used a standard ANN to determine the correlation between the quantity of pottery found in an area and time. This was also done by Ma et al [39], who classified Chinese pottery from the neolithic age using an ANN.

An area not yet mentioned is the use of evolutionary algorithms in archaeology. Although evolutionary algorithms have been used for this purpose, they are not as popular as ANNs. Kontogiorgos and Leontitsis [40] used a genetic algorithm to estimate the weight of microartefacts in order to aid archaeological interpretation of a Greek neolithic site. Maiza and Gaildrat [41] also used a genetic algorithm to determine the best position of a pottery fragment based on a previously calculated distance measure. Coevolution, has been used by IbaƱez et al [42] to handle landmark location uncertainty in skull-face overlay. A fuzzy evolutionary method was applied to the same problem and the two methods were compared.



[15] J. A. Barcelo. Computational intelligence in archaeology. Idea Group Inc, 2009.

[27] J. A. Barcelo, O. de Castro, D. Travet, and O. Vicente. A 3d model of an archaeological excavation. The Digital Heritage of Archaeology. Computer Applications and Quantitative methods in Archaeology, pages 85-90, 2003.

[38] J. A. Barcelo and J. M. Fauna. Time series and neural networks in archaeological seriation: An example on early pottery from the near east. In Proc. Archaeology in the Age of the Internet: Computer Applications and Quantitative Methods in Archaeology, pages 91-102, 1997.

[6] J. A. Barcelo, M. Forte, and D. H. Sanders. Virtual Reality in Archaeology, volume 1. Archaeopress, 2000.

[17] C. M. Bishop. Neural networks for pattern recognition. Clarendon Press, 1995.

[10] G. Bitelli, V. A. Girelli, M. A. Timi, and L. Vittuari. Low-height aerial imagery and digital photogrammetrical processing for archaeological mapping. In Proc. International Society for Photogrammetry and Remote Sensing, volume 35, pages 498-503, 2004.

[4] A. T. Chamberlain. Cave detection in limestone using ground penetrating radar. Journal of Archaeological Science, 27(10):957-964, 2000.

[11] D. Chartier. Apple posts ‘ipad in pompeii’ archaeological case study [online]. Available: http://www.macworld.com/article/154717/2010/10/ipad_archeology_pompeii.html (Accessed: 19 February 2013), October 2010.

[14] P. H. Colomban. Raman spectrometry, a unique tool to analyze and classify ancient ceramics and glasse. Applied Physics A: Materials Science & Processing, 79(2):167{170, 2004.

[5] M. Connor and D. D. Scott. Metal detector use in archaeology: An introduction. Historical Archaeology, 32(4):76-85, 1998.

[2] J. Grant, S. Gorin, and N. Fleming. The Archaeology Coursebook: An introduction to themes, sites, methods and skills. Routledge, 3 edition, 2008.

[1] K. Greene and T. Moore. Archaeology: an introduction. Taylor & Francis, 5 edition, 2010.

[28] F. R. Hodson. Cluster analysis and archaeology: some new developments and applications. World Archaeology, 1(3):299{320, 1970.

[25] F. R. Hodson, P. H. A. Sneath, and J. E. Doran. Some experiments in the numerical analysis of archaeological data. Biometrika, 53(3-4):311-324, 1966.

[42] O. IbaƱez, O. Cordon, and S. Damas. Two different approaches to handle landmark location uncertainty in skull-face overlay:coevolution vs fuzzy landmarks. Advances in Intelligent Systems Research, 1(1):334-341, 2011.

[29] N. Jardine and R. Sibson. The construction of hierarchic and nonhierarchic classifications. Computer Journal, 11(2):177-184, 1968.

[9] E. H. Khoumeri and J. F. Santucci. Gis in archaeology. In Proc. First international Symposium on Environment Identities and Mediterranean Area, pages 441-446, 2006.

[36] T. Kohonen. The self-organizing map. In Proc. The IEEE, volume 78, pages 1464-1480, 1990.

[40] D. Kontogiorgos and A. Leontitsis. Microartefacts’ weight estimation by genetic algorithm minimization. Journal of Archaeological Science, 32(8):1275-1282, 2005.

[24] A. Koussoulakou and E. Stylianidis. The use of gis for the visual exploration of archaeological spatial-temporal data. Cartography and Geographic Information Science, 26(2):152-160, 1999.

[20] N. K. Kubota. Computational intelligence for structured learning of a partner robot based on imitation. Information Science, 171(4):403-429,2005.

[23] K. L. Kvamme. Geographic information systems in regional archaeological research and data management. Archaeological Method and Theory, 1:139-203, 1989.

[34] A. Lopez-Molinero, A. Castro, J. Pino, J. Perez-Arantegui, and J. R. Castillo. Classification of ancient roman glazed ceramics using the neural network of self-organizing maps. Fresenius’ journal of analytical chemistry, 367(6):586-589, 2000.

[39] Q. Ma, A. Yan, Z. Hu, Z. Li, and B. Fan. Principal component analysis and artificial neural networks applied to the classification of chinese pottery of neolithic age. Analytica Chimica Acta, 406(2):247-256, 2000.

[31] J. MacQueen. Some methods for classification and analysis of multivariate observations. In Proc. fifth Berkeley Symposium on Mathematical Statistics and Probability, volume 1, pages 281-297, 1967.

[41] C. Maiza and V. Gaildrat. Automatic classification of archaeological potsherds. In Proc. 3IA International Conference on Computer Graphics and Artificial Intelligence, 2005.

[22] United States. National Oceanic, Atmospheric Administration. Manned Undersea Science, and Technology Office. The Noaa Diving Manual: Diving for Science and Technology. DIABE Publishing, 2 edition, 1992.

[12] G. Philip, D. Gonoghue, A. Beck, and N. Galiatsatos. Corona satellite photography: an archaeological application from the middle east. Antiquity, 76(291):109-118, 2011.

[19] A. K. Pujari. Data mining techniques. Universities Press, 4 edition, 2001.

[26] J. D. Richards and N. S. Ryan. Data processing in archaeology. CUP Archive, 1985.

[16] I. Rouse. The classification of artifacts in archaeology. American Antiquity, 25(3):313-323, 1960.

[21] H. Singh, J. Adams, D. Mindell, and B. Foley. Imaging underwater for archaeology. Journal of Field Archaeology, 27(3):319-328, 2000.

[30] R. R. Sokal and P. H. A. Sneath. Principles of numerical taxonomy. W. H. Freeman, 1963.

[35] V. Tanevska, I. Kuzmanovski, and O. Grup ce. Provenance determination of vinica terra cotta icons using self-organising maps. Annali di Chimica, 97(7):541-552, 2007.

[33] R. G. Toyota, C. S. Munita, C. Boscarioli, E. D. M. Hernandez, E. G. Neves, and C. C. Demartini. Neural network applied to elemental archaeological marajora ceramic compositions. In Proc. International Nuclear Atlantic Conference, volume 41, 2009.

[32] M. H. van den Dries. Archaeology and the application of artificial intelligence : case-studies on use-wear analysis of prehistoric flint tools. PhD thesis, Leiden University, 1998.

[8] B. Volta, T. E. Levy, and G. E. Braswell. The virtual chichn itz project: modelling an ancient maya city in google sketchup. Antiquity, 83(321), 2009.

[7] G. Vozikis, A. Haring, A. Vozikis, and K. Kraus. Laser scanning: A new method for recording and documentation in archaeology. pages 1-16, May 2004.

[13] T. J. Wess, M. Drakopoulos, A. Snigiev, J. Wouters, O. Paris, P. Fratzl, M. Collins, J. Hiller, and K. Nielsen. The use of small-angle x-ray diffraction studies for the analysis of structural features in archaeological samples. Computer-assisted drafting and design. Programs for presenting architectural history and archaeology, 43(1):117-129, 2001.

[3] R. Whallon. The computer in archaeology: A critical survey. Computers and the Humanities, 7(1):29-45, 1972.

[18] Z. Zweig. Using data mining techniques for analyzing pottery databases. Master’s thesis, Bar Ilan University, 2006.


2 comments on “Computational Intelligence in Archaeology

  1. this is very interesting! i am wondering, as an undergraduate myself, what was your undergraduate education in?

    • Hi. My undergraduate degree was computer science but I find most computational archaeologists study archaeology and learn the computer science later. Computer science teaches you to think in a particular manner, so if you’re interested in CI and archaeology I would recommend studying computer science and specialising in CI. However if you maybe just like the idea of a neural network, for example, coursera does offer beginners and advanced courses in such topics. So it all depends on what you want to specialise in. I think I made the right decision for myself by studying computer science.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: