Archaeology specific BERT models for English, German, and Dutch” (Brandsen 2024) explores the use of BERT-based models for Named Entity Recognition (NER) in archaeology across three languages: English, German, and Dutch. It introduces six models trained and fine-tuned on archaeological literature, followed by the presentation and evaluation of three models specifically tailored for NER tasks. The focus on multilingualism enhances the applicability of the research, while the meticulous evaluation using standard metrics demonstrates a rigorous methodology.

The introduction of NER for extracting concepts from literature is intriguing, while the provision of a method for others to contribute to BERT model pre-training enhances collaborative research efforts. The innovative use of BERT models to contextualize archaeological data is a notable strength, bridging the gap between digitized information and computational models.

Additionally, the paper's release of fine-tuned models and consideration of environmental implications add further value.

In summary, the paper contributes significantly to the task of NER in archaeology, filling a crucial gap and providing foundational tools for data mining and reevaluating legacy archaeological materials and archives.

Reference

Brandsen, A. (2024). Archaeology specific BERT models for English, German, and Dutch. Zenodo, 8296920, ver. 5 peer-reviewed and recommended by Peer Community in Archaeology. https://doi.org/10.5281/zenodo.8296920

The paper by A. Lien-Talks [1] presents a small project looking at the use of crowd sourced data collection and particpatory GIS. In particular it looks at the potential of these tools in response to socially disruptive and isolating events such as the COVID-19 pandemic as well as the potential role of digitially mediated heritage initiatives in tackling some of the challenges of changing demographics and life styles.

The types of technologies employed are relatively mature, the project identifies potential for such approaches to be used within the local-history/local community settings, though is also a reminer that depsite the much broader adoption of technology within all areas of society than even a few years ago many barriers still remain. While the the sample size and data collected in the project is relatively modest, the focus on empathy toward the intended audiences from the design process, as well as some of the qualitative feedback reported serve as a reminder that participatory, or crowd-sourced data collection initiatives in heritage can, and perhaps should place potential social benefit before data-acquisition of objectives.

The project also presents a demographic that is not often represented within the literature and the publication and as such the publication of the article represents a meaningful contribution to ongoing discussions of the role heritage and digitally mediated community archaeology can play a role in developing our societies.

References

[1] Lien-Talks, A. (2024). The Ashwell Project: Creating an Online Geospatial Community. Zenodo, 8307882, ver. 4 peer-reviewed and recommended by Peer Community in Archaeology. https://doi.org/10.5281/zenodo.8307882

“What is a form? On the classification of archaeological pottery” by P. Boissinot (1) is a timely contribution to broader theoretical reflections on classification and ordering practices in archaeology, including type-construction and justification. Boissinot rightly reminds us that engagement with the type concept always touches upon the uneasy relationship between the abstract and the concrete, alternatively cast as the ongoing struggle in knowledge production between idealization and particularization. Types are always abstract and as such both ‘more’ and ‘less’ than the concrete objects they refer to. They are ‘more’ because they establish a higher-order identity of variously heterogeneous, concrete objects and they are ‘less’ because they necessarily reduce the richness of the concrete and often erase it altogether. The confusion that types evoke in archaeology and elsewhere has therefore a lot to do with the fact that types are simply not spatiotemporally distinct particulars. As abstract entities, types so almost automatically re-introduce the question of universalism but they do not decide this question, and Boissinot also tentatively rejects such ambitions. In fact, with Boissinot (1) it may be said that universality is often precisely confused with idealization, which is indispensable to all archaeological ordering practices. 

Idealization, increasingly recognized as an important epistemic operation in science (2, 3), paradoxically revolves around the deliberate misrepresentation of the empirical systems being studied, with models being the paradigm cases (4). Models can go so far as to assume something strictly false about the phenomena under consideration in order to advance their epistemic goals. In the words of Angela Potochnik (5), ‘the role of idealization in securing understanding distances understanding from truth but […] this understanding nonetheless gives rise to scientific knowledge’. The affinity especially to models may in part explain why types are so controversial and are often outright rejected as ‘real’ or ‘useful’ by those who only recognize the existence of concrete particulars (nominalism). As confederates of the abstract, types thus join the ranks of mathematics and geometry, which the author identifies as prototypical abstract systems. Definitions are also abstract. According to Boissinot (1), they delineate a ‘position of limits’, and the precision and rigorousness they bring comes at the cost of subjectivity. This unites definitions and types, as both can be precise and clear-cut but they can never be strictly singular or without alternative – in order to do so, they must rely on yet another higher-order system of external standards, and so ad infinitum. 

Boissinot (1) advocates a mathematical and thus by definition abstract approach to archaeological type-thinking in the realm of pottery, as the abstractness of this approach affords relatively rigorous description based on the rules of geometry. Importantly, this choice is not a mysterious a priori rooted in questionable ideas about the supposed superiority of such an approach but rather is the consequence of a careful theoretical exploration of the particularities (domain-specificities) of pottery as a category of human practice and materiality. The abstract thus meets the concrete again: objects of pottery, in sharp contrast to stone artefacts for example, are the product of additive processes. These processes, moreover, depend on the ‘fusion’ of plastic materials and the subsequent fixation of the resulting configuration through firing (processes which, strictly speaking, remove material, such as stretching, appear to be secondary vis-à-vis global shape properties). Because of this overriding ‘fusion’ of pottery, the identification of parts, functional or otherwise, is always problematic and indeterminate to some extent. As products of fusion, parts and wholes represent an integrated unity, and this distinguishes pottery from other technologies, especially machines. The consequence is that the presence or absence of parts and their measurements may not be a privileged locus of type-construction as they are in some biological contexts for example. The identity of pottery objects is then generally bound to their fusion. As a ‘plastic montage’ rather than an assembly of parts, individual parts cannot simply be replaced without threatening the identity of the whole. Although pottery can and must sometimes be repaired, this renders its objects broadly morpho-static (‘restricted plasticity’) rather than morpho-dynamic, which is a condition proper to other material objects such as lithic (use and reworking) and metal artefacts (deformation) but plays out in different ways there. This has a number of important implications, namely that general shape and form properties may be expected to hold much more relevant information than in technological contexts characterized by basal modularity or morpho-dynamics.

It is no coincidence that ‘fusion’ is also emphasized by Stephen C. Pepper (6) as a key category of what he calls contextualism. Fusion for Pepper pays dividends to the interpenetration of different parts and relations, and points to a quality of wholes which cannot be reduced any further and integrates the details into a ‘more’. Pepper maintains that ‘fusion, in other words, is an agency of simplification and organization’ – it is the ‘ultimate cosmic determinator of a unit’ (p. 243-244, emphasis added). This provides metaphysical reasons to look at pottery from a whole-centric perspective and to foreground the agency of its materiality. This is precisely what Boissinot (1) does when he, inspired by the great techno-anthropologist François Sigaut (7), gestures towards the fact that elementarily a pot is ‘useful for containing’. He thereby draws attention not to the function of pottery objects but to what pottery as material objects do by means of their material agency: they disclose a purposive tension between content and container, the carrier and carried as well as inclusion and exclusion, which can also be understood as material ‘forces’ exerted upon whatever is to be contained. This, and not an emic reading of past pottery use, leads to basic qualitative distinctions between open and closed vessels following Anna O. Shepard’s (8) three basic pottery categories: unrestricted, restricted, and necked openings. These distinctions are not merely intuitive but attest to the object-specificity of pottery as fused matter. 

This fused dimension of pottery also leads to a recognition that shapes have geometric properties that emerge from the forced fusion of the plastic material worked, and Boissinot (1) suggests that curvature is the most prominent of such features, which can therefore be used to describe ‘pure’ pottery forms and compare abstract within-pottery differences. A careful mathematical theorization of curvature in the context of pottery technology, following George D. Birkhoff (9), in this way allows to formally distinguish four types of ‘geometric curves’ whose configuration may serve as a basis for archaeological object grouping. The idealization involved in this proposal is not accidental but deliberately instrumental – it reminds us that type-thinking in archaeology cannot escape the abstract. It is notable here that the author does not suggest to simply subject total pottery form to some sort of geometric-morphometric analysis but develops a proposal that foregrounds a limited range of whole-based geometric properties (in contrast to part-based) anchored in general considerations as to the material specificity of pottery as quasi-species of objects.

As Boissinot (1) notes himself, this amounts to a ‘naturalization’ of archaeological artefacts and offers somewhat of an alternative (a third way) to the old discussion between disinterested form analysis and functional (and thus often theory-dependent) artefact groupings. He thereby effectively rejects both of these classic positions because the first ignores the particularities of pottery and the real function of artefacts is in most cases archaeologically inaccessible. In this way, some clear distance is established to both ethnoarchaeology and thing studies as a project. Attending to the ‘discipline of things’ proposal by Bjørnar Olsen and others (10, 11), and by drawing on his earlier work (12), Boissinot interestingly notes that archaeology – never dealing with ‘complete societies’ – could only be ‘deficient’. This has mainly to do with the underdetermination of object function by the archaeological record (and the confusion between function and functionality) as outlined by the author. It seems crucial in this context that Boissinot does not simply query ‘What is a thing?’ as other thing-theorists have previously done, but emphatically turns this question into ‘In what way is it not the same as something else?’. He here of course comes close to Olsen’s In Defense of Things insofar as the ‘mode of being’ or the ‘ontology’ of things is centred. What appears different, however, is the emphasis on plurality and within-thing heterogeneity on the level of abstract wholes. With Boissinot, we always have to speak of ontologies and modes of being and those are linked to different kinds of things and their material specificities. Theorizing and idealizing these specificities are considered central tasks and goals of archaeological classification and typology. As such, this position provides an interesting alternative to computational big-data (the-more-the-better) approaches to form and functionally grounded type-thinking, yet it clearly takes side in the debate between empirical and theoretical type-construction as essential object-specific properties in the sense of Boissinot (1) cannot be deduced in a purely data-driven fashion.

Boissinot’s proposal to re-think archaeological types from the perspective of different species of archaeological objects and their abstract material specificities is thought-provoking and we cannot stop wondering what fruits such interrogations would bear in relation to other kinds of objects such as lithics, metal artefacts, glass, and so forth. In addition, such meta-groupings are inherently problematic themselves, and they thus re-introduce old challenges as to how to separate the relevant super-wholes, technological genesis being an often-invoked candidate discriminator. The latter may suggest that we cannot but ultimately circle back on the human context of archaeological objects, even if we, for both theoretical and epistemological reasons, wish to embark on strictly object-oriented archaeologies in order to emancipate ourselves from the ‘contamination’ of language and in-built assumptions.

Bibliography

1. Boissinot, P. (2024). What is a form? On the classification of archaeological pottery, Zenodo, 7429330, ver. 4 peer-reviewed and recommended by Peer Community in Archaeology. https://doi.org/10.5281/zenodo.10718433

2. Fletcher, S.C., Palacios, P., Ruetsche, L., Shech, E. (2019). Infinite idealizations in science: an introduction. Synthese 196, 1657–1669. https://doi.org/10.1007/s11229-018-02069-6

3. Potochnik, A. (2017). Idealization and the aims of science (University of Chicago Press). https://doi.org/10.7208/chicago/9780226507194.001.0001

4. J. Winkelmann, J. (2023). On Idealizations and Models in Science Education. Sci & Educ 32, 277–295. https://doi.org/10.1007/s11191-021-00291-2

5. Potochnik, A. (2020). Idealization and Many Aims. Philosophy of Science 87, 933–943. https://doi.org/10.1086/710622

6. Pepper S. C. (1972). World hypotheses: a study in evidence, 7. print (University of California Press).

7. Sigaut, F. (1991). “Un couteau ne sert pas à couper, mais en coupant. Structure, fonctionnement et fonction dans l’analyse des objets” in 25 Ans d’études Technologiques En Préhistoire. Bilan et Perspectives (Association pour la promotion et la diffusion des connaissances archéologiques), pp. 21–34.

8. Shepard, A. O. (1956). Ceramics for the Archeologist (Carnegie Institution of Washington n° 609).

9. Birkhoff G. D. (1933). Aesthetic Measure (Harvard University Press).

10. Olsen, B. (2010). In defense of things: archaeology and the ontology of objects (AltaMira Press). https://doi.org/10.1093/jdh/ept014

11. Olsen, B., Shanks, M., Webmoor, T., Witmore, C. (2012). Archaeology: the discipline of things (University of California Press). https://doi.org/10.1525/9780520954007

12. Boissinot, P. (2011). “Comment sommes-nous déficients ?” in L’archéologie Comme Discipline ? (Le Seuil), pp. 265–308.

In this paper [1], the authors describe how they apply machine learning with YOLOv5 to classify visual data, aiming to enhance understanding of archaeological phenomena before conducting destructive fieldwork. Despite challenges, the integration of machine learning with remote sensing technology was seen as transformative, enabling precise recording of areas of interest and assessment of environmental risk factors. The paper discusses successes, failures, and future directions in machine learning research, emphasising the need for standardisation and integration of streamlined methods. The application of machine learning techniques facilitates non-destructive analysis of material culture records, improving conservation efforts and offering insights into both past and contemporary phenomena. While the initial use of YOLOv5 showed potential for consistent detection of archaeological features, further refinement and dataset enlargement are deemed necessary for broader application in non-destructive archaeological surveying. The authors advocate for the integration of machine learning tools in archaeological research to save time, resources, and promote ethical digital recording practices. They highlight the importance of standardised methodologies to enhance credibility and reproducibility, aiming to contribute to the ongoing dialogue in computational archaeology.

Overall, I think this paper is a good step forward in detecting small objects in UAV data, and contains useful information for similar studies. The aim towards greater reproducibility and standardisation is of course shared more widely in the machine learning community, and this study is a good example of how to approach this.

References

[1] Sharp, K., Christofis, B., Eslamiat, H., Nepal, U. and Osores Mendives, C. (2024). Machine Learning for UAV and Ground-Captured Imagery: Toward Standard Practices. Zenodo, 8307612, ver. 5 peer-reviewed and recommended by Peer Community in Archaeology. https://doi.org/10.5281/zenodo.8307612

Managing data in archaeology is a perennial problem. As the adage goes, every day in the field equates to several days in the lab (and beyond). For better or worse, past archaeologists did all their organizing and synthesis manually, by hand, but since the 1970s ways of digitizing data for long term management and analysis have gained increasing attention [1]. It is debatable whether this ever actually made things easier, particularly given the associated problem of sustainable maintenance and accessibility of the data. Many older archaeologists, for instance, still have reels and tapes full of data that now require a new form of archaeology to excavate (see [2] for an unrealized idea on how to solve this).

Today, the options for managing digital archaeological data are limited only by one’s imagination. There are systems built specifically for archaeology, such as Arches [3], Ark [4], Codifi [5], Heurist [6], InTerris Registries [7], OpenAtlas [8], S-Archeo [9], and Wild Note [10], as well as those geared towards museum collections like PastPerfect [11] and CatalogIt [12], among others. There are also mainstream databases that can be adapted to archaeological needs like MS Access [13] and Claris FileMaker [14], as well as various web database apps that function in much the same way (e.g., Caspio [15], dbBee [16], Amazon's Simpledb [17], Sci-Note [18], etc.) — all with their own limitations in size, price, and utility. One could also write the code for specific database needs using pre-built frameworks like those in Ruby-On-Rails [19] or similar languages. And of course, recent advances in machine-learning and AI will undoubtedly bring new solutions in the near future.

But let’s be honest — most archaeologists probably just use Excel. That's partly because, given all the options, it is hard to decide the best tool and whether its worth changing from your current system, especially given few real-world examples in the literature. Bastien Rueff’s new paper [20] is therefore a welcomed presentation on the use of Omeka S [21] to manage data collected for the Timbers in Minoan and Mycenaean Architecture (TiMMA) project. Omeka S is an open-source web-database that is based in PHP and MySQL, and although it was built with the goal of connecting digital cultural heritage collections with other resources online, it has been rarely used in archaeology. Part of the issue is that Omeka Classic was built for use on individual sites, but this has now been scaled-up in Omeka S to accommodate a plurality of sites. 

Some of the strengths of Omeka S include its open-source availability (accessible regardless of budget), the way it links data stored elsewhere on the web (keeping the database itself lean), its ability to import data from common file types, and its multi-lingual support. The latter feature was particularly important to the TiMAA project because it allowed members of the team (ranging from English, Greek, French, and Italian, among others) to enter data into the system in whatever language they felt most comfortable.

However, there are several limitations specific to Omeka S that will limit widespread adoption. Among these, Omeka S apparently lacks the ability to export metadata, auto-fill forms, produce summations or reports, or provide basic statistical analysis. Its internal search capabilities also appear extremely limited. And that is not to mention the barriers typical of any new software, such as onerous technical training, questionable long-term sustainability, or the need for the initial digitization and formatting of data. But given the rather restricted use-case for Omeka S, it appears that this is not a comprehensive tool but one merely for data entry and storage that requires complementary software to carry out common tasks.

As such, Rueff has provided a review of a program that most archaeologists will likely not want or need. But if one was considering adopting Omeka S for a project, then this paper offers critical information for how to go about that. It is a thorough overview of the software package and offers an excellent example of its use in archaeological practice.

NOTES

[1] Doran, J. E., and F. R. Hodson (1975) Mathematics and Computers in Archaeology. Harvard University Press.

[2] Snow, Dean R., Mark Gahegan, C. Lee Giles, Kenneth G. Hirth, George R. Milner, Prasenjit Mitra, and James Z. Wang (2006) Cybertools and Archaeology. Science 311(5763):958–959.

[3] https://www.archesproject.org/

[4] https://ark.lparchaeology.com/

[5] https://codifi.com/

[6] https://heuristnetwork.org/

[7] https://www.interrisreg.org/

[8] https://openatlas.eu/

[9] https://www.skinsoft-lab.com/software/archaelogy-collection-management

[10] https://wildnoteapp.com/

[11] https://museumsoftware.com/

[12] https://www.catalogit.app/

[13] https://www.microsoft.com/en-us/microsoft-365/access

[14] https://www.claris.com/filemaker/

[15] https://www.caspio.com/

[16] https://www.dbbee.com/

[17] https://aws.amazon.com/simpledb/

[18] https://www.scinote.net/

[19] https://rubyonrails.org/

[20] Rueff, Bastien (2023) Dealing with Post-Excavation Data: The Omeka S TiMMA Web-Database. peer-reviewed and recommended by Peer Community in Archaeology. https://zenodo.org/records/7989905

[21] https://omeka.org/