Thirty Years of Digital Libraries: What ECDL/TPDL Tells Us About the Field, and Where We Go from Here

In the late 1990s, a new idea began to take shape in the world of information science: the digital library. At the time, this was not merely a website with a few scanned books. It was a complex engineering challenge, a promise to build entirely new systems that could store, organize, and deliver vast collections of digital content. Researchers were busy constructing the digital shelves, writing the software that would allow computers to talk to one another, and figuring out how to make these systems work for people. Over the next three decades, a specific group of scientists gathered annually at a conference in Europe to share their progress. They called their meeting the European Conference on Research and Advanced Technology for Digital Libraries, later shortened to TPDL. For thirty years, this gathering served as a barometer for the field, recording every shift in what researchers considered important, who was doing the work, and what they were actually building.

A recent study by Gianmaria Silvello, a researcher at the University of Padua, has looked back at the entire history of this conference series to understand how the concept of a digital library has transformed. Silvello analyzed every paper presented at the thirty editions of the conference, from its first meeting in 1997 to its most recent gathering in 2026. To make sense of nearly 1,600 academic papers, the researcher used a powerful computer program, known as a large language model, to read the titles and abstracts of each contribution and sort them into categories. This tool helped identify whether a paper was about building a new software system, studying how people use a library, or analyzing the data inside a collection. The researcher then checked a sample of these computer-generated labels by hand to ensure the machine was reading the text correctly. The goal was to trace the evolution of the field, not just by counting words, but by watching how the community itself changed over time.

The story the data tells is one of a fundamental shift in focus. In the early years, from 1997 to 2007, the conference was dominated by the construction of systems. Researchers were focused on the architecture of digital libraries, creating the frameworks and protocols that allowed different computer systems to share information. They were building the infrastructure, designing the metadata standards that would let a library in one country talk to a library in another, and developing the search engines that would help users find what they needed. The papers were full of technical discussions about system designs, interoperability, and the tools required to make these digital repositories function. It was an era of engineering, where the primary question was how to build the library itself.

As the years passed, the conversation began to change. By the middle period, from 2008 to 2016, the focus started to move away from the construction of the systems and toward the people using them and the content they held. Researchers began to study how users interacted with these digital collections, how to preserve digital materials for the long term, and how to manage the growing web of data. The systems that had been built in the first decade were no longer the main subject of research; they had become the background infrastructure, operated by libraries and archives. The new questions were about what could be done with the records inside these systems.

In the most recent decade, from 2017 to 2026, the field has undergone a second, even more dramatic transformation. The research has shifted almost entirely toward the data itself. Today, the papers are less about building the library and more about processing the scholarly and cultural records that the library holds. The dominant topics are now text mining, natural language processing, and machine learning. Researchers are using advanced computer models to analyze vast amounts of historical documents, scientific papers, and cultural heritage materials. They are building knowledge graphs, which are complex maps of how different pieces of information connect, and they are using artificial intelligence to extract meaning from millions of records. The digital library is no longer seen as a system to be built, but as a curated collection of trusted data that can be used to train artificial intelligence and support new forms of discovery.

This shift in what is being studied is matched by a shift in who is doing the work. In the early days, the conference attracted a broad international audience, with a significant number of researchers from the United States. Over time, the community has contracted geographically. The share of authors from North America has declined steadily, while the presence of European researchers has grown to dominate the conference. Today, the vast majority of participants come from Europe, and the number of countries represented has decreased. The community has become more localized, with a smaller core of returning authors and a large number of newcomers who attend only once. The research groups that drive the conference have also changed, with different teams taking the lead in different eras, reflecting the changing priorities of the field.

The study also reveals how the definition of a digital library has evolved in the minds of the researchers. In the beginning, the term referred to a system or a collection of digital objects. Over time, it has come to describe a role: the accountable curation of collections. A digital library is now understood as the entity that ensures a collection is trustworthy, that it knows where its data comes from, and that it can be used safely by both people and machines. This new definition is crucial because it separates a true digital library from a simple search engine or a random dump of data. It emphasizes the responsibility of the institution to maintain the quality and integrity of the information over time.

The researchers found that while the focus has moved from systems to data, the involvement of actual libraries, archives, and museums has remained steady, though their role has changed. In the early years, these institutions were often the primary builders of the systems. Today, they are more often partners in research, providing the curated collections that researchers analyze. The papers written by practitioners from these institutions now frequently involve collaboration with university researchers, combining the deep knowledge of the collections with the advanced analytical tools of computer science.

Looking to the future, the study suggests several directions for the field. As artificial intelligence becomes more powerful, the need for high-quality, curated collections has never been greater. These collections serve as the training data for machine learning models and the ground truth for evaluating their answers. The researchers propose that the field should focus on the quality of these collections, the reliability of the knowledge graphs built from them, and the long-term preservation of the data. They also suggest that the conference should continue to evolve, perhaps by giving more space to the digital humanities and by strengthening its connections with other fields like data management.

The thirty-year journey of the ECDL/TPDL conference shows a field that has matured from a focus on engineering to a focus on knowledge. The builders of the systems have given way to the processors of the data, but the core mission remains the same: to make the world's information accessible and useful. The digital library has not disappeared; it has simply changed its shape, moving from a place where systems are built to a place where the stories of human culture and science are processed, understood, and preserved for the future. The researchers conclude that the most important thing a digital library can do today is to account for what it holds, ensuring that the collections it manages remain trustworthy and available for generations to come.