Introduction

1Over the past few decades, there has been a growing interest in emigrant letters. Emigrant letters have been used by social historians to investigate push/pull factors and the role of institutions and communities in the process of emigration; they have also been used to explore and understand the conditions and daily lives of emigrants and how individuals and groups adapted to the New World (see Thomas and Znaniecki, 1958; Erickson, 1972; Miller, 1985; Kamphoefner, Helbich and Sommer, 1988; and Fitzpatrick, 1994). Letters are expressive and indicative of correspondents’ identities, values, preoccupations and beliefs; they are a powerful source of information and understanding about migration issues, provide a colourful picture of domestic life from an emigrant perspective, and shed light on processes of language change and variation (on this last point, see, for example, Dossena, 2007; Elspaβ, 2002; and McLelland, 2007).

2Much discussion of the use of “ego documents” or “documents of life” is built around the two editions of Ken Plummer’s book, Documents of Life: An introduction to the problems and literature of a humanistic method (1983) and Documents of Life 2: An Invitation to a Critical Humanism (2001). Plummer has an interesting section on the use of letters and the reading of letters. However, Plummer, following on from Eugene Webb’s concerns, from the 1966 book Unobtrusive Measures, highlights a problem, and to categorise this problem he uses the term “dross rate”: “Letters are not generally focused enough to be of analytic interest – they contain far too much material that strays from the researcher’s concern” (2001: 55). This would seem to be the case in some socio-historical studies. Arnold Schrier’s Ireland and the American Emigration, 1850-1900 (1958) and Kerby Miller’s Emigrants and Exiles (1985), for instance, are less interested in the formulaic aspects of emigrant letters and more interested in what letter writers have to say about, for example, their reasons for migration, experiences of the New World and memories of home. It is worth noting that for Webb and his colleagues “dross rate” is a social science technical term, which can be applied to any kind of data, not just letters (Webb et al., 1966: 31).

3Liz Stanley, however, takes issue with Plummer’s use of the phrase “dross rate”, as applied to letters: “[…] The features of letters he perceives as problems are the very things I find interesting and deserving sustained attention as analytical problematics” (Stanley, 2004: 202). For an example of how information perceived as unimportant by one scholar might be perceived as important by another scholar, David Fitzpatrick’s work on Irish emigration to Australia is especially useful here. Fitzpatrick, in the introduction to his 1994 book, Oceans of Consolation, explores earlier books of and on emigrant letters, the work of his precursors and peers. He looks especially at the editing practices of Blegen (Norwegian letters), Conway (Welsh letters) and Kamphoefner (German letters). Conway, “with due acknowledgement to Blegen, likewise excluded ‘dross’…” (that word again), while Kamphoefner deleted “ritualized pious reflections” and “endless lists of persons to whom the letter-writer sends his or her regards…” – although, to be fair to Kamphoefner, this was to do with space limitations of the printed book (Fitzpatrick, 1994: 20, referring to Kamphoefner, 1991: 46). Fitzpatrick’s own practice is to print his Irish emigrant letters in full, and his method is to take those lists of persons and make maps. Thus, on page 46, there is figure 10, “A Geography of Affection”, mapping the neighbours and relatives mentioned in one letter collection. Fitzpatrick’s maps are curiously moving; they provide a map of a mind, a map of memories – the landscape, as the emigrant remembers it.

4There are two main issues being touched on in the discussion so far. The first relates to how letters are used in different ways across the disciplines: what is one scholar’s “dross” is another scholar’s gold. This can lead to the loss of information and/or the reduplication of work, with letter collections often being transcribed several times by different projects – projects which have their own research aims and, quite often, their own transcription and markup practices. The second issue is to do with interdisciplinarity: while different disciplines might use letter collections, the research is rarely interdisciplinary. Emigrant correspondence projects have often evolved independently of one another, and although project teams have been successful in tackling important research questions relating to, for example, social history and immigration studies they have rarely joined forces, or engaged with stakeholder groups from other disciplines.

5There seems to be a good case, here, for developing a collaborative, cross-disciplinary approach to the digitisation and markup of historical emigrant letter collections. Presently, most existing digital letter collections consist of unannotated versions of original manuscripts. The digitisation process has made the letters more accessible to academics and the general public, and has also increased their searchability, at least to a certain extent. However, relatively few projects have moved beyond the digitisation stage to exploit text content and enhance usability and searchability through the use of digital technologies. Different emigrant letter collections cannot easily interconnect if they are simply digitised without markup, and some search pathways through the material will remain unavailable if software tools are not employed to process this encoding.

7The research network project started in April 2013 and finished in July 2014. Through a series of workshops, the project brought together scholars from different disciplines currently working with emigrant letters as a primary data source, to explore the digital potential of these perhaps iconic documents. In these workshops network participants examined how letters are being used across the disciplines, identifying where there are similarities and differences in transcription, digitisation and annotation practices. The aims were to discuss issues and challenges surrounding digitisation, build capacity relating to correspondence markup, and initiate the process of interconnecting resources to encourage cross-disciplinary research. Central to this was the development of a system of correspondence markup to represent the linguistic, structural, discoursal, contextual and physical properties of the letters, thus offering different layers of meaning and ways-in to the texts, allowing for more sophisticated searches, and also the presentation of outputs through meaningful visualisations.

9The following sections will briefly talk through the network process.

Towards a System of Markup for Emigrant Letter Collections

10The process of encoding is an intellectual activity, which involves thinking about which features of the document to represent, the relationship between those features and how those features should be named, described and categorised in a formalised way. The way a text is encoded will reveal something about what we believe to be important or salient about the original document. Encoding makes explicit our interpretation of a document and as such it is never a neutral process. The same text, then, can be encoded in many different ways, drawing out features that are most relevant to our own research interests, or projects. Encoding makes things visible – it enriches the text; it allows us to look at a text in new ways and from many disciplinary perspectives, offering different ways-in and providing different layers of interpretation. For the “Digitising experiences of migration” project, the encoding process was divided into two stages: document analysis and markup.

Stage One: Document Analysis

11The first stage in our encoding project was to carry out a detailed analysis of the document/s to be digitised and marked-up, to identify features that are important in and across the various emigrant letter collections. This first step is necessary before encoding if the resulting markup is to be useful (Stadler, 2013). When analysing a selection of emigrant letters, the project partners were asked to consider two key questions:

• What do different researchers use correspondence collections for?

• What features of the letters are considered important across the disciplines?

12In many ways, the document analysis stage of the process is the most challenging. Looking at Figure 1, for example, a letter by an Irish emigrant called Julia Lough, one can see a wealth of information that might be of interest to a broad range of scholars, with a broad range of research questions. A paper conservator or paper historian may, for instance, be interested in the “material letter” (Daybell and Hinds, 2010 and Steen, 1994), including the quality of the paper and how the paper has been folded (vertically through the centre and horizontally one third from the top/bottom of the sheet), whereas a graphologist may wish to capture information about the handwriting style, whether there are multiple authors contained within the document, whether the original letter is written in ink or in pencil and how the document is structured, perhaps distinguishing between writing that is contained in the margins and writing that is contained in the main body – information which might provide valuable clues as to the author’s circumstances and background. Additionally, whereas linguists may be more interested in capturing information about the language of the document, such as spelling variations (circled in red), unusual syntax (circled in green), omissions or repetitions, historians, on the other hand, may be more interested in any references to people, places and significant events. There will also be contextual information that is not explicitly stated within the document content, but which is useful to capture nonetheless. This information may be inferred from the document itself or it may be obtained from outside sources. Archival notes, for instance, confirm that the letter shown in Figure 1 was written by someone called Julia Lough, a female Irish emigrant from a town called Meelick in (what was then called) Queen’s County, Ireland. Additionally, previous research by immigration historians suggests that Julia had five sisters, three of whom, like Julia, emigrated to America in the 1870s and 1880s. And, finally, there is information available to us about where the Lough collection came from: a few of the letters were donated to Arnold Schrier, Professor Emeritus at the University of Cincinnati, by Canice and Eilish O’Mahony of Dundalk, Co. Louth and the remaining letters were donated to Professor Kerby Miller, University of Missouri, by Edward Dunne and Mrs Kate Tynan of Portlaoise, County Laois.

13Document analysis, then, is the starting point of the encoding process, looking at how the various disciplines, with their diverse range of research questions, use emigrant letters. Through this process it was possible to identify where there were commonalities across the disciplines and where there were differences, and, finally, where reduplication was taking place. There is not enough space to go through all of the features that were examined, so this paper will focus on just three features of the letters that were deemed to be important across the disciplines, namely: person, location and date. Having examined a selection of emigrant letters, taken from different archives, the network partners agreed that – where possible – the following information should be captured in relation to person, location and date (see Table 1). There are two things to note here: 1) the lists contained within Table 1 are not exhaustive (depending on the research project other or additional information may need to be captured); and 2) it is very rare that all of this information is available for any one letter – quite often the researcher is working with a letter extract, for instance, and has very little information about the sender and/or the addressee.

Stage Two: Markup

14Once the research network had examined the various ways in which letter collections are put to use, specifically identifying those features of letters that are important across the disciplines, the next stage was to agree on how to model that information using TEI markup, thus formalising and standardising this metainformation and allowing letter collections to potentially interconnect.

16The TEI Special Interest Group on Correspondence, established in 2008 and led by Peter Stadler (Universität Paderborn), Marcel Illetschko (Austrian National Library, Vienna) and Sabine Seifert (Humboldt University, Berlin), “seeks to bring together scholars interested in creating digital scholarly editions of correspondence… to discuss and develop sample tagsets (including suggesting additions/modifications to the TEI Guidelines) for varying forms of correspondence” (TEI, 2014). In particular, the TEI Correspondence SIG are developing a special purpose element, <correspDesc>, which will allow features of letters such as sender and addressee to be represented in a standardised way. At the time of writing, this special purpose element is embedded under <profileDesc> within the TEI header.

18The markup shown in Figure 2, above, captures information about the sender <ct:sender>, the addressee <ct:addressee>, and the date of the letter <ct:dateSender>. Embedded within <ct:sender> is the element <persName> that contains a unique identifier, or “key”, which links to a separate personography file (in this case LOUGHPers_0001). This individual personography file contains more detailed information about the sender (information such as that described in Table 1). Similarly, embedded within <ct:placeSender> is the element <placeName> that contains a unique identifier, or “key”, which links to a separate placeography file (in this case LOUGHPlace_0001). This individual placeography file contains more detailed information about the sender’s location (information such as that described in Table 1). Information about the addressee <ct:addressee> is organised in the same way.

19The personography and placeography files are effectively the same as authority files – a term used by archivists and librarians to describe bibliographic “master” files. There is one personography file (or authority file) for each person and there is one placeography file (or authority file) for each place. These authority files for person and place are each given a unique identifier, which is then referred to in the markup. In other words, LOUGHPers_0001, for instance, points to a separate file which contains information about this particular participant (Julia Lough) such as date of birth, first name, surname, maiden name, nicknames (including all spelling variations), sex, occupations, date of emigration, date of death, etc. Similarly, LOUGHPlace_0001 points to a separate file which contains information about this particular place (Winsted) including, for instance, its geographical coordinates. Having separate personography and placeography files (or authority files) for each person and place makes it much easier to manage changes to related metainformation at a later date. It is easier to change one master document – i.e. the personography or placeography file – than it is to change hundreds of documents.

20In addition to the personography and placeography information <ct:dateSender> contains details of when the letter was dated. There are different ways to capture this information within the header. If the letter contains a date, then the day, month and year can be represented in the markup (as in Figure 2). However, quite often an exact date is missing and it is up to the researcher to make an educated guess as to when the letter was written. In such instances, the notBefore and notAfter attributes can be used to place the letter within an approximate timeframe:

21<date notBefore="1800" notAfter="1899"/>

22It is not possible, in the space of this paper, to discuss how the personography and placeography information can be modelled using TEI, however further information about modelling the emigrant letter, together with links to example xml files, can be found on the project blog.

Interconnecting Letter Collections

26Even with minimal metadata it is possible to create meaningful visualisations which allow the user to notice patterns within the data (i.e. what is there) and, just as important, gaps (i.e. what is not there). This paper focuses on metainformation relating to participants (the sender and the addressee – their names and locations) and the date of the letter. This metainformation can be stored in XML documents, spreadsheets or databases, and (provided it is well structured, with established fields relating to sender, addressee, location and date) it is relatively easy, from a technical perspective, to access and retrieve that information. Additionally, the metainformation relating to lots of letters can be stored and retrieved at once. The key question, then, is how to “read” this metainformation.

27Obviously it is possible to read the metainformation relating to each letter one at a time (see Figure 6). One might even be able to identify trends through close reading of the data in this format. Ultimately, though, very large datasets may prove hard to grasp in their entirety, and spotting trends can be virtually impossible. Even a relatively small dataset of, for example, 5,000 letters, can be difficult to analyse in this way.

28This is where data visualisation comes in. By processing the metainformation into a graphical form it is possible to navigate the content more easily, spot trends, concentrate on particular stories within the whole, and bring together content from several different sources. The map shown in Figure 7 (developed using the open source map library, Leaflet), for example, gives a general overview of the distribution of correspondence from both the IED and DIL collections. The blue dots represent destinations while the red dots represent the origins of the letters. Very quickly it is possible to see a cluster of red dots around Northern Ireland – something that one might expect given the nature of the data being used (most of the metainformation here comes from the IED). This is an important point when looking at any visualisation: the visualisation can only reflect the dataset being analysed, including its biases. In other words, the picture is only as holistic as the collection is inclusive. In Figure 7, for instance, there is little by way of correspondence with Asia. This is not necessarily because there were no migrants from Asia to the USA; rather, there are no letters from Asia to the USA within our particular dataset.

30To look more closely at the participants involved in the letter writing process, it is necessary to return to the mapping interface, but this time adding a level of interaction for the user that allows them to identify particular places, people and times. It is difficult to show this with a static image, but Figure 10 is an interactive map and by clicking on one of the bubbles, the user can retrieve additional information about the participants involved in the correspondence.

31It is also possible to display the interaction between participants in another way, focusing more on the individuals that are involved in the correspondence (see Figure  11).

32Narrowing down further on the participants, Figure 12 shows letter writing networks. By clicking on an individual’s name, it is possible to view all persons the individual wrote to, or received letters from.

33The visualisations so far have concentrated on people and places, but there is another aspect to the letters to consider – time. Again, it is possible to take an overview of each collection. This gives us an appreciation of when the main bulk of the letters were written for each dataset. However, by bringing in a faceted means of filtering the data, it is possible to interactively narrow down the analysis. This can be done in a number of different ways, by line-charts, area-charts, step-charts, bubble-matrices and bubble-lines. In Figure 13, a bubble chart is used to track a person’s movements over time. The vertical axis represents location and the horizontal axis represents time. Each bubble represents a person. This is perhaps detailed more clearly in Figure 14, which shows the movements of William Montgomery (an Irish emigrant from Portadown, Ireland).

34William Montgomery appears to have travelled from Portadown to New Orleans sometime between 1831 and 1848. He stayed in America until at least 1850 before returning to Portadown sometime between 1854 and 1857. He then moves to Manchester, after which time there are no more letters.

35Such analysis will not provide all the answers, but it gives a firm basis for further research. Data visualisation allows the user to discover new insights into some very old data, hopefully giving it a new resonance for today. It can also be very cost-effective; all these visualisations shown in this paper use open-source (free) technologies. The scalability of such tools make it possible for multiple projects, with similar content, to be brought together.

Conclusion

36What this encoding process highlighted, in particular, is the potential for working with header information (in this case information embedded within <correspDesc>) relating to person (sender and addressee), location and date, without necessarily having access to the letter itself. One of the biggest challenges that came out of the network process related to accessibility of letter collections and issues to do with intellectual property. In other words, it is often difficult to get access to collections and even more difficult to make collections freely available online – especially when working across disciplines and across cultures. By focusing on metainformation about the letter (rather than the letter itself) there were fewer barriers to overcome with regards interconnecting resources.

37Once the information is digitised, computer programmes can search in and across collections to notice patterns. It becomes possible to step near and step back. Step Near, or step closer, in that we can explore our research material in more detail and explore that detail in new, different ways. We can, to use film metaphors, zoom in, pan round. The detail becomes available to a multitude of disciplines. And Step Back, in that we can observe our material from a distance, observe patterns that might not otherwise be easily visible, though they might have been postulated. Digital technologies can order, and make visual, patterns within our research material – so that, at every stage, there is the ability to Step Near or Step Back. Possible patterns can be explored in detail.

38This is just the first step in terms of developing fully interoperable resources, but hopefully – with further funding – more letter collections will be able to interconnect in this way. Increasingly funding bodies do look for “interdisciplinarity”, but it is not always clear quite what it is that is being looked for, and there are few formal academic careers in interdisciplinary studies (Griffin et al., 2006). Nonetheless, the role of the interdisciplinary interlocutor is worth developing and there are some very simple and obvious rewards to interdisciplinary dialogue: just from the first workshop, for instance, it became apparent that different scholars were working on the same letter collection. For some disciplines a letter is simply a piece of text, trapped on paper and yearning to breathe free. Other disciplines are interested in the actual piece of paper, and argue – quite reasonably – that the physical object is of interest and may hold information that will, in fact, clarify the text and its history. Through examining the emigrant letter within a digital humanities framework it was possible for different disciplines to come together to discuss some of the problems and opportunities of working with correspondence collections, and to discuss best practice within and across disciplines – alliances were formed and the groundwork for future collaborative projects is starting to take place.