“The history of knowledge is the history of forms of expression of knowledge, and those forms change.” Johanna Drucker (2011)

In my Digital Humanities course, I was introduced to the ideas of Johanna Drucker, whose critical perspective I particularly appreciate. She strikes me as an independent thinker who is willing to question established assumptions and bring important issues into academic discussion. One example is her article, in which she challenges conventional ways of presenting information and introduces the distinction between data and capta.

Drucker makes an important distinction between data, meaning “given,” and capta, meaning “taken.” She argues that information is not simply given to us as objective facts. Researchers decide what to collect, what to include or exclude, and how to organize and interpret it. Therefore, she suggests that capta is a more appropriate term because it recognizes the role of human judgment in producing knowledge.

Drucker also questions the way information is presented through graphs, charts, and other visual tools. Her main argument is that visual representations should not simply present information as facts; they should also make us aware of the decisions and assumptions behind them.

Ten years later, in 2021, Matthew Lavin responded to this debate in his article. Lavin agrees with Drucker that human decisions influence how information is collected and interpreted. However, he disagrees with her suggestion to replace the word data with capta. He explains that the word data has a long and complicated history and has not always meant information that is completely objective or independent of human judgment.

Lavin argues that changing the word does not necessarily solve the problem. Instead, he proposes the concept of situated data, meaning that information should always be understood within the historical, social, and cultural circumstances in which it was collected and organized.

I find this debate particularly interesting because it raises a larger question about historical research: If information is always collected and organized through human decisions, can we ever consider data completely objective? Ultimately, the disagreement between Drucker and Lavin is not simply about choosing one word over another. It is about how knowledge is created and how we decide what counts as reliable historical evidence.

From Theory to Practice: Preparing Historical Data

A more practical approach to these questions appears in a 2025 article by Clodomir Santana and his fellow researchers. The authors investigate how the preparation of historical information can influence the results of digital research. Their study uses approximately 168,000 historical records from the Portuguese Empire between 1640 and 1822. They apply a computer-based method called topic modeling, which identifies common subjects and patterns in large collections of texts.

Before analyzing the documents, the researchers first prepare the information through a process called data preprocessing. This involves cleaning and organizing the data, correcting inconsistencies, and deciding how the information should be analyzed.

Their research follows five main stages:

-Collect historical documents: Gather the records needed for the research.

-Preprocess the data: Clean and organize the information, including standardizing words and identifying names and places.

-Apply topic modeling: Use three computer algorithms- LDA, LSI, and GSDMM- to identify recurring themes.

-Compare the results: Examine how different methods of preparing and analyzing the information affect the findings.

-Document the process: Record the decisions and methods used so that other researchers can understand, evaluate, and repeat the analysis.

The authors demonstrate that even when researchers use the same historical documents, different methods of preparing and analyzing the information can produce different results. They therefore emphasize the importance of clearly documenting each step of the research.

This approach seems logical to me because it recognizes the problems raised by Drucker and Lavin while offering a practical way to address them. It does not eliminate human interpretation, but it makes the research process more transparent.

Another reading, Tableau’s “Guide to Data Cleaning,” explains the practical side of preparing information for analysis. The guide describes several important steps: removing duplicate records, correcting spelling and formatting errors, examining unusual values, handling missing information, and checking the accuracy of the cleaned data.

My Experience with OpenRefine

Today, researchers have access to free digital tools that can help them organize and clean historical information. One such tool is OpenRefine.

To practice using OpenRefine, I used a historical dataset I had compiled for my final project, a virtual historical tour of Paris focused on the Paris Salon exhibitions. I selected data from 11 specific years between 1789 and 1823.

I created numeric facets to examine the exhibition years, and the number of artworks displayed. I also created a text facet to identify the political regimes in power at the time of each exhibition. This allowed me to compare changes in the scale of the exhibitions across different political periods in France.

OpenRefine also allows users to filter out irrelevant information, correct inconsistencies, and explore data through numerical distributions and facets. Although it can help identify patterns and relationships in historical data, understanding the reasons behind those patterns still requires historical research and interpretation.

For me, this brings the discussion back to Drucker’s opening statement. The ways we express and analyze knowledge continue to change as new digital technologies become available. However, regardless of how advanced these tools become, we must continue to question how information is collected, interpreted, and presented.