When Data Meets the Past: Data Collection, Sharing, and Reuse in Ancient World Studies

2.1 The Digital Turn for AW Studies

The DH have evolved rapidly, integrating technology with traditional humanities research methods. In AW studies, this integration has opened up new dimensions for research, allowing scholars to manage and study vast amounts of ancient material in ways that were previously unimaginable. Since the digitization of the work by Thomas Aquinas with the Index Thomisticus (Busa, 1974–1980), many other projects such as the Perseus Digital Library (Crane et al., 2006; Crane, 1987) digitized large corpora of ancient texts, making them freely available for scholars worldwide. These resources have enabled researchers to apply new digital tools to ancient texts, conducting analyses that reveal patterns across languages, cultures, and historical periods. One core objective of DH in AW studies is to facilitate access to primary sources and secondary literature, bringing previously isolated or difficult-to-obtain materials into the public domain. Databases containing digitized texts, inscriptions, archaeological data, and other similar sources of information on the AW (for our definition of “data” in AW studies, refer to Section 3.1) now serve as essential resources for AW specialists. They provide a foundation for a range of computational analyses, from linguistic studies to geographic information system (GIS) mapping of ancient sites. Additionally, the integration of standardized and interoperable annotation systems, metadata standards, and data formats has strengthened the field’s capacity to interpret and share research findings on an unprecedented scale. The DH in AW studies also encourage a data-driven mindset that promotes collaboration and reproducibility. By enabling scholars to organize and publish data systematically, digital resources allow for interdisciplinary research and shared data sets that are reusable and expandable by others in the field. This collective approach to data management offers classicists the opportunity to contribute to and draw from a shared pool of knowledge, which in turn fosters new lines of inquiry and a more interconnected research community.

2.2 Databases and Digital Tools as Foundations for Data-Driven Research on the AW: Some Examples

Digital databases have become indispensable resources in the field of AW studies, providing structured information essential to data-driven analysis. These databases are not just repositories of ancient texts and artifacts, but they allow AW scholars to collect and study data that may reveal new insights into language, culture, and material history. Each database type exemplifies the diverse range of structured data on which digital AW studies research depends. Here, we comment on a few of these databases, stressing their potential across disciplines.

One of the most well-known resources for Classicists, the Perseus Digital Library, hosts an extensive collection of Ancient Greek and Latin texts, alongside translations and linguistic tools. It allows researchers to search for specific words and cross-reference sources, facilitating data-driven linguistic analysis. The structured data provided by Perseus and its integration with other resources such as the Morpheus system (Crane, 1991) support studies in different disciplines, ranging from philology to historical linguistics and cultural history, enabling classicists to quantitatively trace linguistic trends and textual relationships across time periods and authors. Similarly, the Thesaurus Linguae Graecae (Pantelia, 2000) and the Thesaurus Latinae Linguae (Baraz, 2007; Krebs, 2009) digitize vast corpora of ancient texts, enabling systematic analysis on a larger scale. New efforts towards a quantitative representation of Latin and Ancient Greek have been made, with the creation of the LatinISE corpus for the former (McGillivray & Kilgarriff, 2013) and the Opera Graeca Adnotata for the latter (Celano, 2024). These resources allow classicists to explore not only individual texts, but also patterns and variations across languages, authors, and genres, making them extremely important for comparative studies and long-term linguistic analyses.

Beyond textual databases, specialized platforms such as Papyri.info (Sosin, 2010) or the Europeana network of Ancient Greek and Latin Epigraphy (Amato et al., 2014; Mannocci, Casarosa, Manghi, & Zoppi, 2014; Orlandi, 2016) extend data-driven approaches into the realms of documentary evidence. Papyri.info focuses on ancient papyri, providing transcription tools and searchable data to facilitate papyrological studies. By organizing documents from various regions and time periods, both platforms enable researchers to conduct large-scale analyses of documentary texts, revealing, for instance, shifts in administrative practices, social relationships, and economic trends across the ancient world. Other databases such as the Portable Antiquities Scheme (Hobbs, 2003; PAS, 2012) by the British Museum and Amgueddfa Cymru (Museum Wales) enable data-driven approaches to archaeological studies. The Portable Antiquities Scheme provides detailed information on artifacts including their descriptions, imagery, and find locations. This structured data are essential for analyzing distribution patterns of material artifacts, supporting research in archaeology, economic history, and cultural exchange.

Digital methods in AW studies rely on large, structured data sets allowing for innovative analyses, such as natural language processing (NLP) and GIS mapping. NLP tools such as the Classical Language Toolkit (Johnson et al., 2021) enable researchers to apply morphological tagging, syntactic parsing, and frequency analysis to ancient texts. By processing classical languages computationally, NLP opens new avenues for large-scale linguistic studies, providing insights into vocabulary distribution, syntactic structures, and semantic patterns across different corpora. These tools can rely on databases of digitized texts (see above) that can be used to train the models to systematically identify and analyze patterns in language use over time and across genres. GIS technologies have integrated spatial analysis into historical studies. Projects such as ORBIS: The Stanford Geospatial Network Model of the Roman World (Scheidel, Meeks, & Weiland, 2012) allow researchers to visualize ancient travel routes, trade networks, and geographic boundaries, revealing how spatial relationships influenced ancient societies. GIS projects rely on databases that organize spatial data (e.g., place names, coordinates, historical maps), allowing scholars to explore the intersection of geography and human activity in the ancient world. By combining spatial data with historical and textual information, GIS mapping enables more comprehensive studies of regional interactions, political boundaries, and economic flows.

Through the structured and searchable data sets provided by these and other similar platforms, databases and digital tools enable AW scholars to take a data-driven approach to ancient materials. These resources facilitate large-scale analyses uncovering unique patterns and supporting interdisciplinary collaborations that draw on methodologies from linguistics, archaeology, history, and data science. However, integrating these digital resources into AW studies is not without challenges, as issues of standardization, accessibility, and technical expertise continue to shape the field (Sections 2.3 and 3.2).

2.3 Fostering Collaboration, Accessibility, and Interdisciplinarity

As outlined in Section 2.2, one of the major outcomes of digital methods in AW studies is the potential to foster interdisciplinary collaboration and broad accessibility. The vast, interconnected data within databases about the AW enables scholars to draw on shared resources and expand each other’s research horizons. For instance, resources such as those described in Section 2.2 provide scholars with immediate access to digitized texts, inscriptions, and archaeological data that others have already collected, cataloged, and structured. This ready availability of data encourages researchers to leverage previous work to explore new questions and patterns, rather than starting from scratch. Such collective access enhances the potential for shared discoveries, making it possible for AW scholars to derive new insights from other people’s contributions, fostering a collaborative environment that builds on a common foundation of AW data. The open-access nature of many digital resources has further democratized AW studies, ensuring that such data is (as much as possible) Findable, Accessible, Interoperable, and Reusable according to the FAIR principles (Wilkinson et al., 2016). By adhering to FAIR standards, digital AW studies projects ensure that data can be easily located, accessed, and utilized in new research contexts, supporting sustainability and promoting a wide-ranging impact within and beyond the academic community.

As digital archives continue to grow, AW specialists can explore new possibilities for large-scale studies, enabling them, for instance, to trace linguistic, cultural, or societal changes over centuries and across vast data sets. This capacity to analyze long-term trends across a large body of data represents a significant advancement in AW studies, underscoring the transformative potential of DH. However, despite these advancements, data collection remains challenging for many AW researchers, and a substantial number still hesitate to rely on digital approaches as primary research methods. It is important to note, however, that computational methods are not strictly necessary for data collection in AW studies. Many scholars still gather extensive data manually on computers, yet often without recognizing that they are in fact creating data sets. This can lead to a few notable issues. First, they may structure their data in ways that hinder reusability, such as by using Word tables instead of interoperable formats (Section 4.1). Second, they might not check whether similar data has already been compiled in related studies, thus missing the chance to align with and possibly expand existing resources (Section 4.1). Finally, they often value the intellectual work of interpretation of their data and overlook the work lying behind data collection, so they keep their data private rather than sharing them with the wider scholarly community (Section 5).

Beyond these technical and methodological challenges, a fundamental debate persists over what exactly constitutes “data” in the context of AW studies. The lack of consensus on this issue raises important questions about how information from ancient texts, artifacts, and historical records should be defined, structured, and analyzed in digital form. Understanding what counts as “data” in AW studies is critical to advancing shared methodologies of data collection and sharing. This complex question will be explored in Section 3.

3.1 What is “Data” in the Humanities and in AW Studies?

To define data in the field of AW studies, it is essential to first establish a general understanding of the term “data” and its application within the Humanities, comparing it to its use in Science, Technology, Engineering, and Mathematics (STEM), and Health disciplines. The difference between what is considered data in STEM and Humanities heavily depends on the difference between these two research areas. By nature, STEM disciplines are based on numerical values, measurable experimental results, structured data sets, and formulas, which better allow for the exploitation of quantitative methods (Valiela, 2001). The main purpose across STEM disciplines is to test hypotheses, find patterns, develop models, predict certain outcomes, and give empirical validation to research questions (Develaki, 2020; NSTA, 2011). On the other hand, Humanities aim to interpret, contextualize, and understand human experiences or cultural products through comparative and/or analytical approaches (Given, 2008). However, data-driven approaches in the Humanities are experiencing significant growth (Section 2.1). They are always used in combination with interpretive methods to analyze complex and often ambiguous research objects. Such complexity and ambiguity arise from several factors intrinsic to the nature of the humanistic enquiry. Human experiences, texts, and artifacts are always embedded in unique cultural, historical, and social contexts that cannot be fully understood in isolation. These contexts introduce layers of meaning that are often interdependent and require careful interpretation – think, for instance, of the meaning of a single word varying depending on its linguistic, social, historical, and even political contexts. Moreover, many Humanities research objects are inherently open to multiple interpretations (Section 4.1). This ambiguity increases when dealing with incomplete or fragmentary sources, such as damaged manuscripts or archaeological finds, where researchers must fill gaps relying on their assumptions and interpretations. The Humanities include a wide array of fields such as History, Literature, Art, Linguistics, Anthropology, and Film Studies, along with specialized subfields defined by time, geography, or language. This heterogeneity is reflected in the great variety of research data used and produced by each of them. Data sources from Humanities research areas range from texts, images, artifacts, oral records, video recordings, among others. The presence of so many different types of data, often not comparable and not measurable with similar metrics or systems, makes the effort of finding a good definition of data for the Humanities an extremely challenging task (McGillivray et al., 2022a, p. 5). In literature, there is no consensus on how data is defined within Humanities research. Apart from a few studies (e.g., Kalinin & Skvortsov, 2023), research on how humanists conceptualize their research data is extremely limited (Poljak Bilić & Posavec, 2024, p. 3). From some surveys aiming at enquiring the attitude of scholars in the Humanities towards the concept of data in their fields, a notable reluctance has emerged to even define their evidence as data per se (Allen & Hartland, 2018). For this reason, DARIAH-DE^[7] proposes an umbrella definition of data for the Humanities as “all sources, materials, and results that have been collected, recorded, or evaluated in the context of research and answers to research questions in the field of humanities and culture, as well as computer-processed data for permanent storage, citation, and further processing”.^[8] The challenge of identifying and defining Humanities data also applies to AW disciplines since, as subcategories of the Humanities, they share their interpretive and analytical approach to data. However, the field of AW studies presents unique characteristics and challenges due to its interdisciplinary nature and the heterogeneity of its data sources. AW studies encompass a wide range of disciplines, including, for instance, History, Archaeology, Linguistics, Literature, Philology, Art History, and Museum Studies, each with its own subfields. Correspondingly, the types of data widely vary and include textual materials (e.g., manuscripts, papyri, inscriptions, translations, annotated linguistic data, word lists), physical artifacts (e.g., tablets, coins, artworks, archaeological objects such as vases or miniatures), and conceptual entities (dates, events, geographical locations). Multiple types of data may often coexist within a single source. For instance, a cuneiform tablet may provide archaeological data (e.g., its provenance, chronological context, and physical features) and linguistic/literary data (e.g., its textual content, linguistic elements, and historical themes). This multiplicity necessitates the application of diverse methodologies, theoretical frameworks, and interpretive standards tailored to the specific nature of each data set.

3.2 Collecting, Curating, and Analyzing Data in AW Studies: A Hard Task

Given the peculiar nature of data in AW studies (Section 3.1), a number of challenges arise with respect to its collection, curation, and analysis. A significant challenge in AW studies is often represented by the scarcity and fragmentation of data (Oldman, Doerr, de Jong, Norton, & Wikman, 2014). Unlike other Humanities disciplines, where data sets can be expanded through methods such as web crawling, e.g. the TenTen corpora (Jakubíček, Kilgarriff, Kovář, Rychlý, & Suchomel, 2013) or interviews with speakers, AW studies must rely on fixed-size corpora. For example, linguists working on modern languages can create extensive corpora containing millions of tokens or generate new data through interviews with native speakers. In contrast, researchers studying historical languages are limited to the finite textual evidence preserved from antiquity. New discoveries are rare, and existing corpora are often small in scale, significantly restricting the potential for large-scale quantitative analyses compared to the resources available for modern languages. Additionally, many texts remain inaccessible in their digital form and only exist as physical editions. Even when digitization is feasible via OCR techniques, this process demands significant time and resources, especially for non-Latin scripts, and may face legal and copyright barriers if critical editions are held by publishers that do not permit open access (Tóth-Czifra, 2019). Similarly, archaeologists rely on the finite evidence preserved from excavations. New discoveries are infrequent and often localized, while existing data are frequently fragmented or incomplete. Moreover, much of the archaeological record is inaccessible, either because it remains unexcavated, is poorly or inaccurately documented in earlier reports, or is restricted by legal or institutional barriers. Even digitization of archaeological material requires substantial investment in resources and faces issues such as inconsistencies in recording standards or loss of context due to incomplete documentation (Kintigh, 2006). These limitations have cascading effects. The lack of standardized encoding for linguistic and textual data in ancient languages, coupled with insufficient data sets, hinders the development of computational tools such as lemmatizers, morphological analyzers, and syntactic parsers for historical languages. Efforts to expand data sets often face challenges related to interoperability and adherence to FAIR principles. The inability to ensure open and standardized formats for digital resources further restricts collaborative research and the creation of derivative tools, such as TEI-XML^[9] encoded digital editions of ancient texts. Moreover, data in AW studies often demands the preservation not only of the digital or physical object (e.g., papyri, artifacts) but also of the scholarly interpretation embedded in its study. Digital editions and databases, therefore, must encompass both the raw data and the scholarly work that contextualizes and analyzes these materials (Almas & Beaulieu, 2013). For instance, Papyri.info (Section 2.2), offers both digitized images of papyri and their philological reconstruction, encoding textual problems such as lacunae in TEI-XML language. In general, traditional approaches within AW studies often primarily focus on the interpretive analysis of data, with the ultimate aim of producing scholarly articles for prestigious journals that would provide valuable academic recognition. These publications, while essential, tend to focus on the scholar’s intellectual and interpretive work, while overlooking the processes underlying the production of the data resources themselves. Yet, these resources represent the foundation that enables both individual scholarship and the broader advancement of the field. As a result, AW scholars sometimes lack awareness not only of what data is but also of what constitutes a data set and of the potential impact that it may have on the wider academic community. This potential for reuse i.e., the property of a data set developed for a specific project to become a resource for possibly entirely different studies, is often underestimated (McGillivray et al., 2022b). There is no definitive minimum size for a collection of data to be considered a data set. However, while a data set does not necessarily need to be vast in scope, it should still have enough depth and coverage to describe a specific phenomenon in a representative way. This is essential for the data set to be impactful and useful to the community. For instance, a data set with only a handful of entries would generally not be considered robust, unless the research object inherently involves only a small amount of data. Consequently, producing data sets that are both specific to their original research questions and broadly reusable is key to fostering collaborative and scalable scholarship in AW studies.

3.3 Data Collection and Data Standardization

The first step an AW researcher faces when conducting data-driven analyses is data collection. For more traditional specialists, who may not typically rely on computational methods, learning to systematically collect data across different areas of AW studies is essential. This process involves developing a structured and intentional approach to data handling, which not only benefits their immediate research but also enhances the reusability and impact of their data sets within the broader academic community. Firstly, it is essential to approach data collection with the understanding that any observed phenomenon in a textual passage, tablet, or artifact can be transformed into a valuable data set, regardless of the methodology employed. Understanding this potential is key to adopting a more data-driven mindset. A straightforward way to adopt this approach is to shift from unstructured documents or personal notes to data formats that are interoperable and open for reuse by other researchers, such as tab-separated formats, e.g. CoNLL-U files for syntactic annotation (Buchholz & Marsi, 2006),^[10] or XML-compliant formats for encoding philologic or semantic annotation (TEI Consortium, 2024). This allows for a more systematic and easier way to retrieve data and patterns, through programming languages and tools for visualization and distant reading. In more recent times, research on data structures is moving towards the creation of Linked Open Data infrastructures by encoding information in RDF^[11] format. This approach aims at reducing data fragmentation, so that different resources, originally scattered across various project websites and resource infrastructures can be linked and “talk” to each other, providing researchers with a large amount of information in one single place. Latin is a particularly notable case, as it is arguably the most well-resourced historical language, benefiting from a range of linguistic resources and NLP tools. In this context, the ERC project “Lila: Linking Latin” (2018–2023) has been building a knowledge base of linguistic resources using the Linked Open Data model (Passarotti et al., 2020). Providing AW scholars with training on new advancements in the field would not only help integrate them into the broader discussion on open data-sharing practices and resource infrastructures but also inspire others to follow the example set by LiLa. Consistency is another critical factor in the collection and creation of new data sets. Often, especially in the past, each research project defined its own labels for encoding the phenomena under study. AW researchers should strive to use controlled and standard vocabularies for encoding the phenomena for which the data set is being created. This would ensure clarity and uniformity, making the data more accessible for both personal use and external collaboration. A significant obstacle to this lies in the fact that many datasets remain inaccessible to the public, as they are often closely guarded by the researchers who created them. This lack of widely shared open-access practices makes it challenging to obtain information that could otherwise contribute to the establishment of shared standards for encoding similar – or even identical – phenomena across related fields, while highlighting best practices and providing examples of how to structure data effectively. In relation to the notion of open data, another fundamental step is that the community gains awareness about existing FAIR principles for data collecting and sharing (see Section 2.3). Considering FAIR principles when collecting data enhances the ease of data sharing and increases the potential impact of both the data set and its associated research. This approach involves learning about data standards in one’s field, such as TEI-XML for philological and linguistic data, as well as familiarizing oneself with data-sharing platforms like DARIAH for the Arts and Humanities (Section 3.1) or CLARIN for linguistic data. CLARIN not only serves as a repository for humanities data but also offers training resources for scholars, covering the use of tools within its infrastructure and the implementation of FAIR principles throughout a research project’s lifecycle.^[12]

Understanding the importance of metadata is another crucial aspect of adopting a data-driven and FAIR-oriented approach to the study of the AW. A clear and comprehensive description of data sets ensures that they remain intelligible to researchers who are unfamiliar with the original study for which they were created, facilitating reuse and interdisciplinary collaboration (Section 5). One of the major challenges in managing data within AW disciplines lies in the inherent heterogeneity of these fields (Section 3.1). The diversity of disciplines and the variety of their data types make it difficult to devise universal standards. As it occurs within the Humanities in general (Section 3.1), this is further compounded by the broad temporal range covered by AW studies, often spanning millennia. For instance, Latin and Greek have a tradition of more than two millennia, where each period reflects not only significant linguistic changes, but also transformations in art, culture, and history – consider, for instance, the profound influence of Christianity on the evolution of Latin (McGillivray et al., 2022b). This heterogeneity is reflected in the nature of the data, necessitating an interdisciplinary approach to the collection, metadata creation, processing, and curation of data sets. Additional challenges include the scarcity of data (Section 3.2), its proliferation without adherence to common standards, and the fragmentation that characterizes many AW disciplines. For instance, in historical linguistics, particularly for Latin, syntactically annotated corpora (treebanks) are an essential resource for researchers. There are currently six Latin treebanks: the Latin Dependency Treebank (Bamman & Crane, 2006), PROIEL (Haug & Jøhndal, 2008), the Index Thomisticus treebank (Passarotti, 2011), UDante (Cecchini, Sprugnoli, Moretti, & Passarotti, 2020), the Late Latin Charter Treebank (Korkiakangas, 2021), and UD_Latin-CIRCSE^[13] (still ongoing). Initially, each Latin treebank (with the exception of the more recent UDante and UD_Latin-CIRCSE) was annotated using different annotation standards. However, the lack of common annotation guidelines highly limited the interoperability of these resources, affecting both ancient and modern languages. Efforts towards a standardization process have been tied in the last 8 years to the Universal Dependencies project (Nivre et al., 2016),^[14] which aims at establishing a cross-linguistically valid framework for treebank annotation. While significant progress has been made in harmonizing these resources for Latin, discrepancies in the annotation of specific syntactic phenomena remain, alongside overlaps in the texts included in different treebanks (Gamba & Zeman, 2023). Such inconsistencies can hinder interoperability and limit the potential for large-scale comparative studies. The field of Assyriology presents additional unique challenges, such as the coexistence of multiple languages (e.g., Sumerian, Akkadian, Hurrian, Babylonian, Assyrian, and Hittite) and the intricate relationship between written and spoken forms. Scribes, for example, often wrote in ways that diverged from everyday spoken language, introducing a layer of complexity in interpreting the data. One of the most significant obstacles is the identification of cuneiform signs. Variations in the interpretation of a single sign across studies can make it challenging to standardize and compare data sets (Section 4.1), complicating the broader goal of establishing shared standards. In conclusion, developing a mindset oriented towards collaboration and reuse can significantly benefit the field. Researchers should consider how their data sets might be useful to scholars in other, related areas. By adopting practices that promote transparency, consistency, and accessibility, traditional classicists can transform their data collection methods into a more structured, impactful process, bridging the gap between traditional and computational approaches in AW studies. The following section will discuss the limitations and advantages of data-driven approaches to AW studies, providing solutions to the former through examples of previous works.

4.1 Overcoming Barriers: Addressing Common Data-Driven Hesitations in AW Studies

One of the main issues usually raised by AW researchers when discussing data-driven approaches is the reduction of more complex information. For example, in the case of structured data sets created from literary or documentary texts, the overall meaning of the text, its context, and its themes might be lost when prioritizing the quantitative over the qualitative. An over-reliance on quantitative approaches may lead to the risk of overlooking qualitative analyses, favoring data over interpretation and flattening complexity, ambiguity, and rich qualitative analysis. To address this, some practical solutions have been implemented in specific research contexts. An example of historical linguistics is offered by the annotation of modal passages in a corpus of Latin texts, carried out in the framework of the project “WoPoss: A world of possibilities. Modal pathways over an extra-long period of time: the diachrony of modality in the Latin language” (Dell’Oro, 2019–2025). In the framework of the WoPoss project, modality is defined as the expression of possibility, necessity, probability and volition. The annotation of modality is a challenging task, as the type of modality expressed in a passage highly depends on its linguistic and extralinguistic context, and sometimes implies a certain degree of ambiguity. Moreover, the WoPoss annotation schema is very fine-grained, including 25 labels for types and sub-types of modality (Dell’Oro, 2023). For those cases in which the modal passage is considered ambiguous between two modal readings, or an agreement is not reached among annotators on the reading of the same passage, the WoPoss annotation guidelines allow for double annotation. This type of solution ensures the preservation of the linguistic data’s complexity, rather than oversimplifying it to fit the need for assigning unambiguous labels to a complex linguistic phenomenon – in this case, the semantic ambiguity of modal expressions in context. Adaptations such as these ensure that qualitative research is still possible, even with data-driven approaches. Furthermore, the problems associated with the simplification of complex information vary across the different fields that comprise AW studies. In Assyriology, for instance, cuneiform signs can be read in different ways, and scholars choose which readings fit best in the context of the said manuscript through a process called “normalization” (Finkel & Taylor, 2015). By the first millennium BCE, the learned scribes of Assyria and Babylon had developed incredibly sophisticated literary compositions full of duplicitous meanings (Worthington, 2020), rendering modern “normalization” of these manuscripts incredibly difficult. Choosing a single consistent reading of a cuneiform sign can oversimplify the nuanced meanings embedded in Mesopotamian linguistic and cultural contexts. For example, double consonants in a word can alter its meaning, such as changing a verb from active to passive, though cuneiform spelling does not always indicate this distinction. Additionally, the adaptation of cuneiform from Sumerian to Akkadian introduced further ambiguities, as the same symbols represented different sounds, with or without emphasis. This can lead to multiple interpretations, such as in manuscript K2821, where the scribe alternates between the spellings for šahāṭu ‘attack’ and šahātu ‘rinse’. Such complexities require subjective judgment by Assyriologists but can often be overlooked in standardized, data-driven approaches. Another key challenge faced by AW researchers in embracing data-driven approaches to the field is, sometimes, the lack of more technical expertise in areas such as data collection and systematic data organization. For instance, some AW scholars may not be familiar with basic data management tools like Excel. Instead of filling in spreadsheets to collect their data, they may rely on tables created on Word, which, however, make it extremely difficult for their data sets to be queried. Unlike Excel spreadsheets, Word tables do not allow for filtering, so data cannot be systematically organized, retrieved, or analyzed. Moreover, simple text files are not easily shared or integrated with other resources. On the other hand, a well-structured data set stored in a format such as comma separated values (CSV) allows for easy sharing, cross-referencing, and analysis by others in the field. Creating a CSV file is not as difficult as it might seem to AW researchers who are not used to digital methods:.xlsx files or spreadsheets manually created with Google Sheets may be easily converted into a CSV file. Thus, an AW specialist could easily create a CSV file where each column holds a specific piece of information (e.g., title, author, location, date, sentence) and each row represents an entry (e.g., an inscription, a word, an artifact). The resulting data set could then be shared and reused by other scholars in the field (see Sections 3.2 and 5). Examples of different data sets compiled for AW studies can be retrieved from the articles published by Burns, Farina, Marongiu, and Rodda (2023–2024). Usefully, once the data set is in CSV format, it can also be imported into various software tools for further and finer analysis, enhancing its potential. This approach allows classicists to work with their data in a flexible, scalable, and accessible manner, enabling them to make valuable contributions to digital classics projects without needing extensive technical expertise. Although CSV is a beginner-friendly format, it does not fully meet interoperability standards, limiting its long-term integration with other data sets and digital resources. More robust and interoperable formats, such as XML – particularly TEI for textual data – offer greater flexibility and sustainability for data structuring and exchange. TEI-XML is widely used in digital humanities projects (Cummings, 2018), ensuring consistency and compatibility with other scholarly resources. To address this, workshops and training sessions introduce traditional AW researchers^[15] or humanists more in general^[16] to the implementation of the TEI-XML format in their workflows, alongside simpler formats like CSV. Such initiatives help these scholars gradually develop digital skills, making it easier to transition from basic to more advanced data management and digital skills. A third challenge may be represented by the lack of interoperability between existing data sets. Different digital repositories in the AW fields often use incompatible formats or structures, making it difficult for researchers to combine data from multiple sources. For example, a researcher might have a data set on inscriptions from a particular region, but this data set may not align with a larger collection of archaeological data from a different source. If these data sets do not share common standards or metadata (e.g., location, date, type of artifact), it becomes extremely difficult to merge them (Section 3.3). However, working towards standardization and adopting shared labels and frameworks is an immensely valuable endeavor. By aligning data sets with standardized metadata fields or classifications, researchers can enhance the usability of their data across different projects and disciplines. Standardization requires effort and adaptation, but it unlocks the potential for larger-scale, collaborative analyses that would be impossible – or that would require more effort and therefore more time – with isolated or non-integrated data sets. One example is given by Farina (2023a), where the morphological features of the annotated word tokens follow the same notation as the Universal Dependencies (see Section 3.3). It is essential for AW scholars to research existing data sets before creating their own, in the same or in similar fields (e.g., searching data sets on modern languages before annotating historical languages). This helps avoid duplication and promotes integration with established resources: this way, scholars may ensure that new data aligns with broader disciplinarity efforts, rather than remaining isolated. Another challenge for traditional researchers in AW studies may be a lack of familiarity with secure data archiving standards. The urgency of data security was highlighted by the data breach at the British Library, carried out by the hacker group Rhysida – one of the most severe data security incidents in the United Kingdom (British Library, 2024). As a result, approximately 600GB of copyrighted material was illegally leaked, and the British Library suffered significant financial losses in restoring its security systems. While incidents of this scale are unlikely to directly impact AW researchers, the academic community has increasingly adopted a framework of data preservation and sharing fully aligned with the FAIR principles. In most cases, aside from specific circumstances involving copyright restrictions or ethical concerns, the primary goal is to ensure that data sets remain accessible and reusable for scholarly research. To achieve this, established research infrastructures such as CLARIN offer step-by-step guidance for depositing data sets, ensuring compliance with data and metadata standards. These platforms also require careful selection of an appropriate license, and, where access restrictions are needed, they manage access control on behalf of researchers, granting entry only to authorized users. This approach relieves researchers of the burden of acquiring complex technical expertise on data security while simultaneously requiring them to develop an informed understanding of best practices in open science and the licensing frameworks applicable to research data. Finally, the integration of digital AW studies within the AW curricula is still in its infancy but is growing as research increasingly adopts data-driven approaches. Many Classics departments still need to incorporate training on digital tools, data management, and interoperability into their programs, leaving students and researchers unaware of new approaches. Participation in data-driven AW studies is therefore limited, with a generalized lack of awareness about the usefulness of sharing data and the misconception that publishing data sets requires advanced technical expertise or should only be done by those specializing in digital methods. This perspective overlooks the fact that data sharing does not need to be one’s main research focus, but it can be a natural product of a broader project. For instance, data sets compiled during textual or archaeological analyses can be cleaned, structured, and published alongside traditional research outputs (Farina, Marongiu, & Rodda, 2024), greatly enhancing their accessibility and reusability.

4.2 The Benefits of Adopting a Data-Driven Approach in AW Studies

Building and utilizing data sets in AW studies offers numerous advantages, transforming the ways in which scholars approach their research. First, data sets enable more systematic analyses, allowing AW researchers to uncover patterns, trends, and connections that might remain invisible through traditional qualitative methods. This systematic approach fosters a deeper and more comprehensive understanding of AW-related phenomena. Digitizing AW data is particularly relevant in this context, as it enhances accessibility, making it easier for researchers to access, share, and utilize materials. Moreover, digitization helps mitigate issues of data sparsity and redundancy, ensuring that resources are not only preserved but also effectively utilized by a wider community. The creation and sharing of data sets play a pivotal role in fostering interdisciplinary research, providing AW scholars with opportunities to collaborate across disciplines and bringing fresh perspectives and methodologies to the study of the AW. By structuring data in accessible formats and making it openly available, data sets enable scholars from various fields to engage with AW data, facilitating the exchange of ideas and methods. Interdisciplinary collaboration, increasingly encouraged by universities, funders, and research groups, benefits AW scholars by introducing new tools and techniques while offering insights into how their work might intersect with other domains. Importantly, AW scholars are not expected to shoulder the entire burden of acquiring technical skills in data science and DH. Many successful projects in AW studies have received substantial funding precisely because they incorporate interdisciplinary teams, bringing together domain experts alongside specialists in computer science and DH. The ERC-funded project LiLa (Passarotti et al., 2020), the SNSF-funded project WoPoss (Dell’Oro, 2019–2025), and the project Ithaca, funded by the EU Horizon 2020 (Assael et al., 2022) exemplify how collaboration between AW scholars and technical specialists can effectively manage the complexities of creating, sustaining, and analyzing data sets or integrating them with AI. Such partnerships allow AW scholars to focus on their subject’s expertise while benefiting from the technical proficiency of DH and computer science professionals, ensuring that data sets are not only created but also maintained and enhanced with the latest methodologies. For example, partnerships with DH or computer science can yield advanced computational tools, such as taggers and parsers for ancient languages, which are essential for scaling up research and analyzing vast corpora of ancient texts. This is particularly critical in areas like papyrology and Assyriology, where enormous quantities of texts remain unstudied (Cobanoglu et al., 2024). Furthermore, making data sets reusable and reproducible not only enhances the reliability and replicability of findings but also supports the development of resources that amplify the reach and impact of AW scholarship. The process of publishing data sets also offers significant benefits to AW specialists, especially those with limited experience in this area. Data set publication typically follows a structured layout, providing a clear framework for organizing and presenting data in line with best practices. This process also fosters professional growth by exposing scholars to constructive feedback from peer reviewers, a practice traditionally reserved for research articles. Applying classical peer review to data sets not only strengthens the resource itself but also enhances its usability and reusability, while promoting collaboration and networking. A critical step toward fostering interdisciplinary collaboration is the sharing of one’s data. Publishing a data set in open-access repositories such as Zenodo or Figshare ensures it is freely available and subject to scrutiny by experts, enriching its value to the academic community. For even greater impact, scholars are encouraged to publish an accompanying data paper in conjunction with an open repository (Section 5).

4.3 Encouraging a Broader Participation in Data-Driven Approaches to AW Studies

Some of the limitations discussed in Section 4.1 can be, to a large degree, offset by encouraging a broader participation in data-driven AW studies. This can be done by presenting examples of data sharing and showing the benefits of this approach through organized events. Successful initiatives such as the workshop “Data Driven Classics: Exploring the Power of Shared Datasets,” organized by Andrea Farina and George Oliver (King’s College London) in July 2024, have gathered together PhD students and early career researchers interested in learning more about creating and sharing data sets. Such events not only introduce participants to practical approaches to data set creation but also provide hands-on opportunities for training in data conceptualization and collection, interoperability standards, and data publishing. Communication among scholars is essential to the process of data sharing: other events such as the “Data in Historical Linguistics Seminar Series,” launched by Andrea Farina and Mathilde Bru (King’s College London) in 2023–2024 have successfully created research collaborations and partnerships, as well as spreading awareness of the sorts of data being collected within the field of historical linguistics. Promoting greater participation requires consistent and systematic efforts to engage scholars at various stages of their careers, from students to senior researchers. Initiatives like DARIAH-EU, which offers resources and training for DH, are pivotal in equipping AW specialists with the tools and methodologies necessary for successful data set creation and publication. Encouraging institutions to incorporate such training into their curricula could ensure that emerging scholars are better prepared to engage with data-driven research from the outset of their careers. Academic working groups such as that of Pleiades for ancient geographical data (NYU’s Institute for the Study of the AW; cf. Elliott, 2021) are vital for advancing shared standards in data set creation and interoperability. Participating in collaborative efforts – such as contributing to repositories or establishing discipline-specific metadata standards – can empower scholars to make their data sets more widely usable and integrated with larger research projects. As these collaborations grow, they create a feedback loop, where best practices evolve and become embedded within the discipline, ultimately broadening the impact of data-driven approaches in AW studies. In addition to academic events and working groups, data journals are increasingly being considered as a means of promoting data-driven approaches in the field. For example, the Journal of Open Humanities Data’s special collection “Representing the AW through Data” (Burns et al., 2023–2024) attracted considerable interest within the field of AW studies (Farina et al., 2024), with the subdisciplines of data sets ranging from Latin literature to archaeological data from Roman rural landscapes to historical linguistics and cuneiform studies. For many contributors, this was their first time publishing a data set and accompanying data paper. For example, Bru (2023a, b) presented a data set on the average length of Greek words from the 5th century BCE to the 2nd century CE, demonstrating an increase in average word length over time. The publication of this data set in the Harvard Dataverse repository was prompted by the journal guidelines. As a direct result of publishing her data set, the author collaborated with other linguists and received invaluable feedback from peer reviewers, notably regarding visually presenting the data through boxplots and performing further statistical tests on the data to strengthen the validity of her conclusion. The call for papers for this collection made it clear what constituted a data set and gave helpful instructions to authors publishing one for the first time. By providing detailed submission guidelines and peer review processes, these journals also play an educational role, ensuring that contributors develop essential skills in organizing, documenting, and structuring their data.