- 1 I thank Dr. Jean-Pierre Le Glaunec for access to this work before publication.
- 2 Appendix B lists examples of such sites.
1Archives dedicated to the documentation of slavery and to slave resistance have become an essential part of historical scholarship. Since the emergence of the World Wide Web, the publication of these archives has also migrated online. This movement has brought the discussion around the “marginalized or discriminated against” outside the walls of academia (Dinsman 2016). As researchers developed and debated the concepts of humanity, agency and resistance (Johnson 2003), they mobilized different types of documents to bring to light the oppression and the struggle of millions of individuals, largely Africans, who were subjected to this practice. One of the first collections used to convey life under slavery came out of oral history, demonstrating the variety of materials required to give a voice to the oppressed. Slave narratives collected by the Federal Writers’ Project were transcribed and then made available on microfilm, in print and now online (Federal Writers’ Project, 1941). Less common occurrences such as large-scale slave rebellions have also gathered significant attention as flash-points of resistance. The web sites developed to present these events provide in-depth analysis supported by maps or by powerful illustrations. In between, printed collections of runaway slave advertisements (Windley 1983) have emerged as “one of the staples of the historiography of eighteenth-century slavery” (White & White 1995). These short notices (Figure 1), were originally designed to report escapes and to provide information for the recapture of self-liberating slaves. As with many other documents created to enforce slave systems, they have become important sources for the study of slavery and resistance (Le Glaunec 2021).1 Given the nature of the material (brief, descriptive, varied and powerful), advertisements are now also a staple of online archives dedicated to the history of slavery and to its teaching.2 The richness of the documents has been brought out through new analyses, creating a prism where individual actions are refracted to bring out the conditions of slave humanity.
Figure 1. Example of a runaway slave advertisement (Affiches Américaines, January 1, 1766)
Source: Jean-Pierre Le Glaunec & Léon Robichaud, “Le marronnage dans le monde atlantique : sources et trajectoires de vie,” hereafter Marronnage project
- 3 Tom Costa, The Geography of Slavery in Virginia, Charlottesville, University of Virginia, 2005 (ht (...)
- 4 David Gwynn, North Carolina Runaway Slave Advertisements, 1750–1865, Greensboro, University of Nor (...)
- 5 Simon P. Newman, Runaway Slaves in Britain: bondage, freedom and race in the eighteenth century, G (...)
- 6 Edward E. Baptist, Freedom on the Move, Ithaca, Cornell University, 2017- (https://freedomonthemov (...)
- 7 Matrix, the Center for Digital Humanities and Social Sciences, Enslaved. Peoples of the Historical (...)
2Online archives have varied scopes. Certain projects are regional in nature, such as Virginia and Maryland (4,000 advertisements),3 North Carolina (2,400),4 or Great Britain (836).5 All of these sites present both the image and the text of each advertisement. The smallest project also offers detailed information to facilitate search based upon a broad range of variables (including skills, physical characteristics and rewards) while larger projects focus on document metadata (for proper source attribution, copyright information, etc.) and basic information on persons mentioned in the documents (name only or name, gender, age, height and skills). As the number of items increases, project leaders have to make choices as to how detailed the personal data can be. To overcome the financial constraints of humanities research, some projects rely on crowdsourcing to create extensive collections, as in the case of Freedom on the Move (27,422 advertisements in May 2020).6 With a robust infrastructure, this strategy holds long-term promise to create a vast database with verified transcriptions and detailed data entry. To facilitate access to the growing number of separate databases, efforts are ongoing to federate online advertisements and other documents related to slavery in a linked and open platform.7
- 8 Jean-Pierre Le Glaunec & Léon Robichaud, Le marronnage dans le monde atlantique, 1760-1848. Source (...)
- 9 In many slave societies, black persons were deemed to be slaves unless they carried papers proving (...)
- 10 We would like to thank Dave Roy for the data entry. An analysis of persons who did not declare a n (...)
- 11 Jean-Pierre Le Glaunec & Léon Robichaud, Le marronnage dans le monde atlantique…, op. cit. The col (...)
3The Marronnage in the Atlantic World web site8 (Marronnage project) participates in this broad scholarly dynamic and adds an original feature because it also includes prison lists (Figure 2). In these lists, authorities announced, enumerated and described persons who had been captured and deemed to be fugitive slaves.9 These documents remind us that freedom could be temporary and they provide us with a new source to evaluate how prevalent marronnage could be. Using documents from 1776 in Saint-Domingue as a sample, we discovered that there are almost 4.5 times as many people enumerated in prison lists as there are in traditional advertisements (Table 1).10 In Guadeloupe, there are twice as many prison lists as there are advertisements, even without counting the individuals mentioned.11
Figure 2. Example of a prison list (Affiches Américaines, January 3, 1776)
Source: Marronnage project
Table 1. Number of persons advertised and captured, Saint-Domingue (Affiches Américaines, 1776)
Month |
Individuals in advertisements |
Individuals in prison lists |
January |
59 |
107 |
February |
25 |
192 |
March |
54 |
172 |
April |
46 |
174 |
May |
55 |
204 |
June |
44 |
177 |
July |
47 |
210 |
August |
35 |
208 |
September |
37 |
193 |
October |
23 |
158 |
November |
21 |
179 |
December |
20 |
125 |
Total |
466 |
2099 |
Source: Marronnage project
4Creating a collection of such documents presents its own challenges. Records created by institutions dedicated to upholding (or ending) slavery such as trade records, plantation records, slave registers or freedom papers can be harvested with a certain level of efficiency because a high proportion of the content is specific to that issue. Advertisements and lists, for their part, are sprinkled in newspapers which cover a broad range of topics. How does a research team tackle their digitization, description and publication? As researchers familiar with the process are well aware, every issue of every newspaper published in the slave societies selected for a project has to be scrutinized to extract individual notices (Figure 3) which then have to be described and added to a digital collection. This paper focuses on how the Marronnage project, co-directed by Jean-Pierre Le Glaunec and the author of this article, tackled these issues. This partnership brought together expertise on fugitive slave advertisements from different locations (Le Glaunec 2007), and in the development of historical web sites, including a project highlighting the presence of slavery in New France (Beaugrand-Champagne & Robichaud 2006). Colleagues at the Université de Sherbrooke, the co-directors could rely on their combined skill-sets to manage and operate the project with a minimal team and limited funds.
Figure 3. Fugitive slave advertisements and prison lists on a newspaper page (Affiches Américaines, January 1, 1766)
Source: Marronnage project
- 12 The Affiches Américaines have since been digitized as part of the Digital Library of the Caribbean (...)
- 13 Previous versions of the site can be viewed on the Internet Archive’s Wayback Machine (https://web (...)
5At first (in 2008-2009), the objective was to publish a collection of documents which were hard to access: the fugitive slave advertisements published in the Affiches Américaines, the official newspaper of the French colony known as Saint-Domingue which became Haiti when it gained independence in 1804. Selected advertisements from this newspaper had been used by historians, but the full extent of fugitive advertisements was unknown, and the newspaper was only available on microfilm.12 We hoped that the digitization of all of the advertisements would make this corpus accessible and would shed some light on the diversity of marronnage. Given its focus, the original web site was entitled Marronnage in Saint-Domingue (Haïti), history, memory, technology.13 Given the interest garnered by the original version, we built on this expertise and expanded the site (with new funding) to include fugitive slave advertisements from other regions (Guadeloupe, Louisiana, South Caroline, Jamaica and Lower Canada) to facilitate comparative studies, and add prison lists to offer a new type of document. Over time, the layout and the interface were improved and the project name was changed to Marronnage in the Atlantic World to reflect the broader scope of the project.
6When developing such a project, teams have to define objectives and strategies that match their resources. In the case of Marronnage, limited means curtailed our ambitions with regards to infrastructure and procedures. The emphasis has therefore been on accessibility, diversity and usability in an effort to bring attention to the documents themselves and to individuals who were maroons. At each phase of our occasional funding, we added new content and improved the interface, always wondering if it would be the last round. We were also prepared to transfer the data to other projects should we no longer be able to maintain the site. Having survived its first decade, Marronnage has nevertheless become a reference for scholars, students and members of the public (Appendix 1). Our considerable and varied corpus provides easy access in one interface to more than 15,000 fugitive slave advertisements and nearly 7,000 prison lists extracted from twenty-one newspapers published in seven different regions (Figure 4). This original contribution allows users to compare slavery, resistance and capture across societies or to drill down and study individual acts of defiance and self-liberation.
Figure 4. Spatial distribution of the documents
Source: Marronnage project, created with Carto mapping platform
Table 2. Distribution of documents by region and by period
Region |
Newspapers |
Period |
Documents |
French Guyana |
1 |
1829-1829 |
77 |
Guadeloupe |
1 |
1815-1833 |
451 |
Jamaica |
5 |
1800-1815 |
1,784 |
Louisiana |
7 |
1801-1815 |
993 |
Lower Canada |
4 |
1765-1810 |
87 |
Saint-Domingue |
1 |
1766-1791 |
17,308 |
South Carolina |
2 |
1802-1815 |
1,785 |
Total |
21 |
n/a |
22,485 |
Source: Marronnage project
7As described in the following section, short and simple documents such as advertisements and lists contain a wealth of information which has to be structured into a database. Given that the co-directors had no immediate plans to analyze the data for publication, the model, described in the next section, was designed to facilitate access to the content. The following section will describe the project’s evolution, including experiments which did not provide the anticipated results, an aspect of academic research which is not usually revealed by scholars. These forays into pilot initiatives have confirmed that the project should remain focused on its strengths. As the conclusion will demonstrate, the site’s adoption across the world and its effective use in research have consolidated its place in the landscape of online archives dedicated to the enslaved.
8Fugitive slave advertisements have already been analyzed to highlight discourse on race, violence, bodies and the banality of chattel slavery in slave societies. This paper will therefore focus on descriptive information which can be transposed into a database to facilitate search and analysis, starting with metadata which has to be captured from the newspaper issue as a whole (Figure 5).
Figure 5. Newspaper header (Affiches Américaines, January 1, 1766)
Source: Marronnage project
- 14 There can be gaps in the numbering due to images spread over two columns or two pages being merged (...)
9As expected, we can identify the newspaper name, the date of publication and the issue number from the header. The place of publication will sometimes be declared outside of the header, or is found in registers of historical newspapers. The Affiches Américaines has other specificities. Aside from the regular issues published in Cap Français (today Cap Haïtien), there are also supplements and distinct issues for Port-au-Prince. This issue type (type_parution) can be used to study local variations in marronnage. Apart from the general information regarding the issue, the metadata also includes the page number on which a document is located, and the order in which the items appear in the issue.14
10In the database, information related to the newspaper is entered in a reference table (Figure 6). The city and region (both period name and current name) are entered to facilitate groupings. The newspaper identifier (idjournal) can then be used in each document entry to reference which publication it was from. The newspaper code (code_journal) can link items back to references used in J.-P. Le Glaunec’s research. The journal table currently contains 21 records, one for each newspaper from which we have extracted documents (Figure 7).
Figure 6. Structure of journal table
Source: Marronnage project
Figure 7. Content of journal table
Source: Marronnage project
11Other metadata is specific to the issue (date, issue number, issue type) or to the item described (page number, order in the issue). We find this information in the document table (Figures 8 and 9), where each record is identified by a unique number (iddocument). The type_document field distinguishes advertisements from prison lists, all documents being grouped in this generic document table as opposed to our original concept of creating a different table for each document type. Again, a field was created to reference items back to J.-P. Le Glaunec’s earlier advertisement collection (code_jplg). Occasionally, we could not identify the exact publication date (9 cases). The precision_date field allows us to specify if the date entered is exact (le) or approximate (vers le).
Figure 8. Metadata structure for the document table
Source: Marronnage project
Figure 9. Sample of metadata content for the document table
Source: Marronnage project
12Other parts of the document table contain the text transcription and procedural metadata (status, name of person who performed different certain tasks…). The image is not stored in the table. Two fields point to the directory and file names to create the link between the data and the image file. Other fields were created to add comments specific to a document (remarques) or to relations between documents (intertexte). We included republications of advertisements so that we can compare them with the original publication. Given that some republications have modifications, we could calculate the percentage of the repetition using a Levenshtein distance.15 Certain fields are remnants of the original design which have yet to be removed (type_annonce_fr and type_annonce_en) while others were created during a phase when we planned to translate the documents (transcription_en for English and transcription_cr for Haitian Creole).
Figure 10. List of document information fields in the document table
Source: Marronnage project
Figure 11. Sample of content from document information fields in the document table
Source: Marronnage project
13The transcription and link to the images are the crucial fields for users viewing the information for each document on the interface. The digitization and transcription process will be described in the next section. For any given document, there is at least a digital image (6,248 items) or a transcription (4,562 items), or both an image and a transcription (11,600 items).
14To calculate how may words were transcribed and to facilitate certain indexing operations, a separate table contains individual words, numbers and abbreviations (1,174,390) links back to each document in their order of appearance and flags stop words in French for a total of 892,986 significant words, numbers or abbreviations. The list of stop words was created based on their frequency and therefore included certain verbs; it was recently modified to include abbreviations which could be of interest for some researchers. Plans are underway to add English stop words.
Figure 12. Structure of the document_mot table
Source: Marronnage project
Figure 13. Sample content from the document_mot table
Source: Marronnage project
15Given that one objective was to offer a window into individual experiences, we invested a lot of resources in recording every person mentioned in the advertisements, adding their name and gender when indicated, for a total of 17,856 men or boys and 4,293 women or girls. We found more than 3,700 distinct first names, but we also created a record for unnamed persons (4,539) who escaped, alone or as part of a group. There are therefore many entries, without names, which nevertheless record a quest for freedom. This information is in the esclave table, where we enter a unique identifier for each mention of a person (we have not yet grouped individuals mentioned more than once), as well as the principal name and gender; it is also in the esclave_nom table, where we enter all names used.
Figure 14. Name and gender in the esclave and esclave_nom tables
Source: Marronnage project
16Advertisements and prison lists contain varying types of information and levels of precision for each person mentioned, creating a challenge for database design. Should the database structure reflect all types of content, or should it be restricted to predetermined research objectives? Given that we could not fund the development of a robust data management interface, and that we could not ensure complete data entry for every document, we created a number of fields with a simple interface to evaluate the types of information available for some sample years. After collecting data for the years 1766, 1776 and 1786 for Saint-Domingue, we concluded that resources were not available to extend this type of data entry for all cases. As a result, fields related to age, height, ethnic origin, language(s), skills, descriptions of body, clothing, moral character, etc. were not used for other years and locations. An analysis of places from which slaves escaped or where they took refuge would have been interesting, but also required more data entry than we could manage. For Saint-Domingue, we were able to identify ship names by searching for certain words in the documents (capitaine, navire, etc.) for which a specific table (negrier) was created, and which was later used in a research project.
Figure 15. Structure of the negrier table
Source: Marronnage project
Figure 16. Sample content of the negrier table
Source: Marronnage project
17The fundamental fields of the database serve the basic purposes of the project: reference the documents, enumerate the individuals whose memory the project evokes, and identify the collaborators who took part in the process. Recognizing the work of the participants is important as demonstrated by H. Lovejoy in his contribution to this issue (Lovejoy 2020). To facilitate links with other projects, this structure will be converted to an XML format as in the case of Liberated Africans (Lovejoy 2015-2020). We will also explore the creation of RDF triplets to facilitate the integration of our project in a linked and open data world (Michon 2016). Given the evolution of the Marronnage project over the past decade, we believe that by streamlining content to what can be managed with limited means, this project also defines some basic practices that can be used by other online archive projects.
- 16 Unfortunately, many historical photographs are also published with incorrect attributions or conte (...)
18Online publication of historical material has matured since the pioneering era of the 1990s (Brüogger & Milligan 2019). Institutions and research teams have developed methodologies to ensure the authenticity and quality of the archives and to facilitate access to them.16 The Marronnage project has striven to achieve scientific and usability requirements while working with limited resources. The project’s objective has therefore been focused on documents from different locations, improving accessibility to lesser-known collections such a that from Saint-Domingue, and to publish prison lists which have less potential for story-telling but offer more data for analysis. Given that documents were acquired by different methods (transcription only, paper archives, microfilm, digitized collections), we faced difficulties in standardizing their presentation across such a large corpus, a process which is still ongoing. Although we could not create a uniform visual appearance for all documents, we nevertheless reached our objective of remembering the many stories of marronnage and, through documents designed to oppress and maintain slavery, of giving a voice to those who regained, even if only temporarily in some cases, a measure of freedom.
19Given that Marronnage was not at the core of either co-director’s research and publication efforts, work on the project was occasional, with intense phases and long lulls. We documented decisions in various ways, but not in a continuous logbook. A single method for documenting the evolution of the project should have been implemented from the outset, for example in a GitHub where the project code could have been stored.
20To start the process, we chose three sample years for Saint-Domingue to test the methodology and evaluate the amount of work to complete the corpus: 1766, 1776 and 1786. Procedures for selecting documents for research assistants (RAs) unfamiliar with the material (identifying what is an advertisement and how to digitize from microfilm) were put in place. We verified that the RAs captured all the documents, and in a sample test only one out of a total of 359 had been missed. Transcription rules were also defined to ensure that RAs maintained the spelling as found in the documents, identified the long “s” and transcribed it as the letter “s” and not as the letter “f” and described letter sequences printed upside down, backwards, or both (Figure 17).
Figure 17. Example of stamps that need to be described in the transcription (Affiches Américaines, January 15, 1766)
Source: Marronnage project
- 17 New tests would be required to verify if handwritten text recognition (HTR) would achieve better r (...)
- 18 It would actually be possible to analyze this transition among Saint-Domingue printers using this (...)
21We chose to rely upon human transcribers rather then optical character recognition (OCR) because of the varying quality of the images taken from the microfilms.17 Given the short length of the documents and the repetitive nature of the text, we were able to achieve a quality of transcription well above 95%, with sample corrections confirming an error level below 3%. The most common error was for students to type the verb “connaître” and its various forms with modern spelling, as opposed to the “connoître” form which was common until the 1780’s in Saint-Domingue.18
22The available digitizer created pdf files which were then batch converted and resized into the jpg format using ImageMagick19 software. To facilitate the creation of metadata, each image file was renamed to indicate the date of publication, the issue type and number and the order number of the advertisement or list in the issue. The file name was then copied and pasted as the first line of each transcription to facilitate loading the information into the database. The importation script created the item metadata for each record by parsing each file name. Another script read the transcription text file, using the file names as separators between each item, and copied the transcribed text into the appropriate record in the database. Errors in linking were detected and corrected in the data entry phase, from basic information about each enslaved person mentioned (Figure 18) to details about physical appearance, languages, origin or skills (Figure 19).
Figure 18. Entry form, basic individual information
Source: Marronnage administration section
Figure 19. Entry form, detailed individual information
Source: Marronnage administration section
23This sample of Saint-Domingue documents was made available to the public in the fall of 2009. As we reflected on the next phase, it became evident that the growing number of documents over the course of the time period (196 advertisements in 1766, 346 in 1776, 580 in 1786) would make it impossible to add detailed information for every person. The second phase was therefore focused on digitizing all advertisements from the rest of the Affiches Américaines. From May to December 2010, the items were digitized and imported into the database. A new interface was developed to transcribe the text directly (Figure 20), and only the name and gender of each enslaved person was entered (Figure 21). A professional graphic artist improved the site’s design and usability so that we could present the full extent of advertisements from Saint-Domingue in the most efficient manner. Once the documents were transcribed, we noticed that 201 documents provided information about a ship or a captain. This information was compiled in the database and was used in student projects at Duke University (Zinna 2012).
Figure 20. Transcription interface
Source: Marronnage administration section
Figure 21. Data entry for name and gender
Source: Marronnage administration section
- 20 Parts of lists had to be copied and pasted into a new image to show a complete view of the list.
24Having achieved the original goal, we experimented in 2011 and 2012 with two options for further development. Content was tested for teaching purposes: bibliography, other printed sources, images, and selected advertisements. Pursuing that track was however likely to prevent the addition of documents which were being neglected by historians: prison lists. A test sample from 1776 had demonstrated the potential of these documents, and efforts were therefore focused on adding 6,535 prison list images to the database. Given the length of many lists, more work was required to create a single image from items which covered more than one column in the newspaper20. The volume of documents and the number of people mentioned meant that full transcription and data entry was not possible with the available funding. A 10% sample of prison lists was processed in the same manner as the advertisements, which nevertheless provides a usable sample for researchers who would wish to explore those documents further (see Yales 2013 for such an example).
25The extraction of names from prison lists in our sample raised another possibility: could we link individuals appearing in both sources? If it was possible, it would open up new avenues for studying escape routes, networks, and individual trajectories. Adapting procedures used in historical demography, we attempted to link individuals appearing in both sources, on the basis of name, age, ethnic origin, height and body markings. In the end, we could only reliably connect 32 people, which was less than 10% of the individuals identified in advertisements and 1,5% of those mentioned in prison lists. Our methodology was simple enough to be converted into an algorithm, but the limited results meant that it did not warrant further development to apply it to the full corpus.
- 21 Such work is being accomplished by Paul E. Lovejoy, Érika Melek Delgado and Kartikay Chadha, “Free (...)
26The frequent association of people mentioned in the documents with various ethnonyms suggests that such a corpus can easily be used to study the origins of the captives who escape from slavery, possibly even of captives in general. On a practical level, there was an issue of standardizing the approximately 200 designations which appear in the documents so that our research assistants could manage the information efficiently. As we attempted to define a vocabulary, it became obvious that more research was required by specialists in the field21 before we could structure the information in a manner which could be used scientifically, instead of reproducing slave trading categories.
27The project was then dormant for a few years while the co-directors focused on other research projects. When they returned to Marronnage in 2016, it had become evident that the site’s success was based upon offering a substantial corpus of documents and that this should remain the core mission of the project. While streamlining the type of content offered, the geographic coverage was broadened and a new visual interface was designed to reflect this orientation. Documents were to be added from Jean-Pierre Le Glaunec’s doctoral dissertation (Louisiana, South Carolina and Jamaica, Le Glaunec 2007), the appendices published in Frank Mackey’s study of the abolition of slavery in Lower Canada (Mackey 2010), and documents photographed on site in Guadeloupe or at the Bibliothèque nationale de France (for French Guyana). We considered adding Mauritius and Réunion, but this would have extended our geographic coverage beyond the Atlantic. With regards to Spanish or Portuguese colonies, we did not have the expertise to manage the content adequately. Given that larger projects had begun the process of covering the United States, we preferred to avoid duplicating the work unless we already had a corpus on hand such as that provided by Le Glaunec. The advertisements of one important slave colony, Martinique, still however remain largely inaccessible; we hope to add them to the corpus in the next round of funding.
28The documents had varied origins and the processing pipeline had to be adapted for each case. In that of Lower Canada, the same procedure defined for Saint-Domingue could be applied, starting the process with microfilm. The Guadeloupe newspaper proved to be richer than anticipated during an exploratory research trip, and digital images were made on site using a smart phone, which required corrections using the GIMP image editor.22 During research at the Bibliothèque nationale de France, a year of the French Guyana newspaper were digitized using the same method. Those documents could then be imported into the database with the same script used for Saint-Domingue, adjusting for metadata differences. A different strategy was required for Jamaica, Louisiana and South Carolina, because only the transcription was available, without the digital image. Each item was copied and pasted into a specifically designed interface, with newspaper and personal metadata being added at the same time.
29Given the increase in the number of documents, the interface was streamlined to simplify access to the most basic data while facilitating the selection of time and place. The period covered varies from territory to territory. For Saint-Domingue and Lower Canada, we begin when newspapers appear in each colony and end with the last published advertisements or prison lists, whether from a revolution (Haiti) or a de facto end to the legal enforcement of slavery (Lower Canada). The documents for Guadeloupe and French Guyana offer a sample to illustrate the types of documents available for those colonies. The temporal coverage for the three other territories (Jamaica, Louisiana, South Carolina) had been defined as part of Jean-Pierre Le Glaunec’s dissertation.
30Given the size and somewhat diverse nature of the content, we needed a simple and effective interface for searching and browsing. Our original version attempted to provide access to as much information as possible, including to some of the attributes entered for the original three sample years of Saint-Domingue data (Figure 21). This provided unexpected results for users, given that few people consult methodological documentation. In the simplified version, we chose only to offer access to the document images themselves and/or their transcriptions. Users first see a splash page with sample documents and a mention of our partners. The presentation page offers basic statistics on the content, a short text and a map showing the locations from which our documents originate. Users can find out more about our project under “À propos” and “Méthodologie,” but the core is the “Corpus” tab.
Figure 22. Search page for “Le marronnage à Saint-Domingue ”
Source: Marronnage project (archived version of the website)
- 23 The facets use AJAX (Asynchronous Javascript and XML) technology, which makes it possible to reloa (...)
31In keeping with current practice, we offer a faceted search interface with the facets on the left-hand side of the screen, the transcription in the center and a miniature of the image on the right, when the image is available (Figure 24). When only the image is available, a notice informs the user. A click on the miniature opens the full-size version of the digitized image of the document, whether or not the transcription is available. A first page with twenty documents is loaded when the tab is first opened. Users have the option of browsing through the 1,125 pages of documents or using the facets to filter documents matching basic criteria23. The first three filters search the content of the documents: keywords, name or sex. The word search tries to match the string of characters entered, searching for an exact match, including spaces. The names have been indexed and the first letter typed will bring up the list of corresponding names to facilitate the search. The sex filter allows users to target their search to take into account gendered aspects of marronnage. The other four filters reflect metadata about the documents themselves: location (colony or state), newspaper title, document type and year of publication. The location and newspaper filters are linked so that once a location is selected, only the newspapers published in that colony or state are available. The document type is designed to focus a search on advertisements or on prison lists. We chose to offer a slider to define a time interval as opposed to typing in specific dates. This conforms to common interface design practices for academic projects and responds to the needs of most users who tend to generate a subset of data rather than search for individual documents. Finally, all filters can be reinitialized by clicking on a button which reloads the page.
- 24 In the five level (or five star) rating system proposed by Tim Berners-Lee, founder of the World W (...)
32The frontend of the system is currently only available in French, but the backend is designed to accommodate presenting the interface in many languages. Plans are underway to translate the interface into English and Haitian Creole, but the documents will remain in their original language of publication, whether French or English. Another upcoming modification, in line with the third level of open data standards, will be to allow users to download the metadata for their selection in text delimited format.24 Until now, data has been supplied to researchers when requested. The future of this data is another concern. Quebec university libraries are setting up data repositories, but our institution, the Université de Sherbrooke, at this date only offers a document repository. While we have preferred developing our project independently for a decade so that we could focus on publishing using a light approach, we will need to consider depositing the data with a larger project on the history of slavery and resistance.
33Over a ten-year period, we have invested approximately $75,000 (Canadian currency) in transcription, data entry, interface design, programming and translation, an approximate cost of slightly more than three dollars for each of the 22,485 items now online. Contributions by the co-directors in terms of transcription, digitization, image manipulation and programming were a definite factor in keeping the cost per unit as low as possible. If these in-kind contributions were calculated, it would cost closer to eight dollars to integrate and display each document into the system. Given the low cost of this project, the co-directors are always surprised to see how it is used in the community. Users from all over the world have appropriated the content, not only for historical research, but, as we will see in the concluding section, for studies in linguistics, representation, genealogy and local history.
34In preparing this project, we naturally wished to unlock these documents for academic research. We also hoped, however, that making them available would enable non-academics to discover these important traces of history. The usual metrics such as number of visitors, number of pages consulted and duration of the visit are not very helpful in measuring the impact of a small web site. We therefore chose to analyze how the site is used and in what context.
35A first indication of Marronnage’s reach is the diversity of geographical origins of visitors. We have reached visitors in most countries around the world (Figure 26) with the main audience located in Haiti, the United States, France, Canada, and the United Kingdom (Table 3). Other former French slave territories (Martinique, Guadeloupe, Réunion, French Guyana) appear in the top 25 places of origins, as do countries from which people were captured and transported across the Atlantic Ocean (Senegal, Côte d’Ivoire).
Figure 26. Origin of visitors to www.marronnage.info, 2009-2019
Source: GoogleAnalytics
Table 3. Top five countries from which users consult marronnage.info, 2009-2019
Country |
Number of sessions |
Percentage of sessions |
Haiti |
14,333 |
35.38 |
France |
5,274 |
13.02 |
Canada |
5,128 |
12.66 |
United Kingdom |
833 |
2.06 |
Source: GoogleAnalytics
36The site was presented in various conferences and workshops over the past decade, but the co-directors have not yet published any research based upon the documents made available to the public, and the only publication dedicated to the site was a very recent review in the American Historical Review (Geggus 2020). We were therefore surprised by the variety of publications and theses which appear in a Google Books search for “marronnage.info” (see complementary bibliography). As we might expect, the site is used in studies on slavery and resistance, but also on urban history, the arts and even botany. In conferences and blog posts, presenters have used the site as an example in digital humanities and as a resource for exploring slavery and colonial history. University libraries include Marronnage in the lists presented to students who wish to study the history of slavery and resistance, the history of the United States or the history of the Caribbean. The site is also listed as a reference in a YouTube video, in a world history course series. Blog posts on those topics and on the history of rum or on the legacy of slavery also include Marronnage. In genealogy forums, users reference the site for the history of planter families or as a resource for local Haitian history. Elevated to the status of a national hero in Haitian history, identity and popular culture, the marron remains the subject of controversies (Lucas 2002; Béchacq 2006). The contested memories of the marrons were clearly present when Le Glaunec discussed this project in Haiti itself. However, the Marronnage documents have not yet become part of online debates on the subject, except in the case of Boukman (Salnave 2016).
37Over the course of a decade, the Marronnage project has succeeded in making more than 20,000 documents available online and in transcribing more than a million words. Our ambitions and our methodology have been defined by the resources available, keeping the site simple and efficient. This was furthermore achieved at a time when other research projects occupied most of the co-directors’ attention. As the world of digital humanities moves towards big data projects requiring large infrastructures, there is still a place for small digitization and publication projects. These projects can follow basic guidelines to ensure their scientific quality and their medium-term existence. For the long term, project directors need to plan on depositing their data in a repository and/or partnering with larger projects to transfer their corpus. As we reflect upon this experience, we are the first to admit that we were not always able to follow the best practices in the field. Our exploratory approach has nevertheless brought to the fore documents which would otherwise likely remain hard to access. We believe that scholars who have digitized other such collections can also easily make them available online to increase knowledge about different local conditions. Sharing content will avoid duplication of effort, and facilitate endeavors to connect and federate the many thousands of documents that have been digitized.
Phases |
Year(s) |
Tasks |
Contributors |
I |
2009 |
Digitization, transcription, and full data entry. Affiches Américaines, 1766, 1776, 1786 |
Anne Bruneau-Poulin Jean-Pierre Le Glaunec Marie Lemonnier Sarah Maquet |
Database design Interface design Programming |
Paquin Design Léon Robichaud Heather Robichaud |
II |
2010-2012 |
Digitization, transcription and summary data entry. Affiches Américaines, 1767-1775, 1777-1785, 1787-1791 Digitization of prison lists, Affiches Américaines, 1766-1791 Digitization of advertisements, Lower Canada newspapers Verification of translations for advertisements, Affiches Américaines Preparation of a bibliography on the history of slavery and marronnage Preparation of a digital image collection Experimentation with linkages between advertisements and lists Experimentation with teaching resources Preliminary analysis of ethnic origins indicated in the documents |
Jean-Pierre Le Glaunec Marie Lemonnier Julie Marleau Pierre Meese Pierre-Olivier Parenteau Annie Poulin Jonathan Proulx-Salvas Julie Ratté Josianne Ricard Léon Robichaud David Roy Hubert Ste-Marie |
Interface modified for a responsive design Addition of statistics to the site Interface translation Integration of complementary resources to the site (bibliography, images, sources) |
Victor Bilodeau Paquin Design Olivier Robichaud Philippe Tardif |
III |
2015-2019 |
Integration of data from Jamaica, South Carolina and Louisiana Digitization of newspapers from Guadeloupe and French Guyana Transcription of sample prison lists Transcription of new documents Data entry for new documents |
Jacqueline Garriss Thomas Garriss Jean-Pierre Le Glaunec Geneviève Piché Léon Robichaud Mélise Roy-Bélanger |
Refresh interface design AJAX programming |
Paquin Design Heather Robichaud |