What is German, what is French

The schizophrenic situation of having somehow a French hemisphere and a German one sometimes leads me to awkward contradictions.

On the one hand, my mother tongue and culture, the whole system in which I grew up and that taught me how to think. On the other one, the culture I have been professionally active in for over ten years and which I feel I have been adopted by. Of course, having a foot in each of those cultures means having two feet in none and hence be rootless everywhere. That is the deal. As much as I figured, during my first five years in Berlin, that we would eventually get back to France, it has been a while now since we decided to  stay here for good, at least until the kids are big enough to cope with us moving to a place that has never been their home (that is, probably, not before they are 18, which leaves us some 14 years).

For a long time, I only occasionally accepted to acknowledge the fact that I know French in my scientific activity. When I realized how useful this linguistic competence could be  to study precisely the German texts I was interested in, I gave in and took advantage of it. That never made me a specialist of French literature though. I know the basics. But if I had stayed in France and made a PhD in French literature instead of German studies, I would know immensely more. My only advantage is of linguistic more than of scholarly nature. Does it count in terms of qualifications? I still don’t know.

There is another level on which the conflict between both hemispheres is more complex to handle: the method. The first aspect is the team constellation. Such a work environment as working in a team is still almost unheard of in French literary studies. It is considered at best typical for natural sciences. To be honest, it is weird in the context of German literary studies as well, even if the fundings that lead to that kind of group configuration are more widely spread in Germany.

The second aspect is deeper and harder to formulate. The methods I have grown in still define the questions I ask to the corpora I am interested in. It became clear when I had to define the profile of our edition. Of course, the whole “it is the research question that defines the edition” thing veiled it for a while and I could not see it. But it finally became obvious that the prevalence of genetic over historical-critical edition in my eyes has in many ways to do with the teachings I received in my student years, when historical-critical was considered a nice, pointless commentary that does not deal  in depth with the definition of the text layers. The question I am also trying to deal with on a more theoretical level (which could be as plainly as pompously summed up in four words: What is the text?) is certainly a reflection of the conflict between the two scholarly cultures that live in me, sometimes antagonistically.The text is what culture makes it be. Maybe my mission is to make this cultural discrepancy bear fruits within text interpretation. But should I accept it?

And who wants me to work on that? The French or the Germans? Back to square one…



Genetics and pdf

I have been reluctant to practice genetic editing so far. Not that I disregard the importance and/or meaning of such editions, but I did not think that we were fit for that. My main worry was that it would take too much time working with elaborate genetic tools like critical apparatus, that it would slow down our publication rhythm for a meager benefit. I had decided to solve the tension between content tagging and genetic tagging in favor of the first, which made it necessary to encode a pretty big text quantity to be able to make something out of it. Time-consuming genetic encoding was made superfluous by the necessity of delivering text masses.

When we started approaching canonical authors, it became clear that my solution was not necessarily the best one. At first, I thought it had to do with the difference between letters and texts: we cannot do as much content tagging on the literary works as we do on the letters, considering people, works, places don’t play such a great role there. So the tagging of the “Romantic Key Texts” mostly concentrates on rendering the materiality of the manuscript. I realized during the work on the Chamisso letters that the imperative of genetic encoding had to do with the necessity of an authorial enunciation (e.g. to provide for an ultimate text and not for any text) in the case of authors that cannot tolerate to be treated like we don’t care what the word it is that has just been stroked through.

In the end, we developed a complex tagging system that tries to deal with both identifying content entities and rendering manuscript materiality. It has taken a great deal of time to conceive said system, it still takes plenty of time to implement it (we haven’t encoding half of the third of the fourth of what I had hoped for this fall). What is more, I find it less and less satisfactory. One of the reasons why it is so is that it addresses the genetic issue not as a research question, but as a problem to get rid off.

The preparation of the grant application for the August Boeckh Encyklopädie achieved to make a shift in the conception necessary. This only became clear to me when I changed the title of the project  last week. From a “historisch-kritisch”, I switched to “genetic”, because this is what the project was describing.

Now that the grant application is out (or will be in one hour or two), I can finally show you what the pdf test page looks like! – Looking forward to hearing your commentaries on it, because I intend to improve the method for our other documents.


Multilingual improvements (j’ai la #frt!)

If you are, my beloved reader, paying just as much attention as I expect you to, you will have noticed that we have been improving the French interface of the edition over the past 10 days – with a variable success, some of the fields now appearing in all three languages at a time in all the versions. In preparation for the Frankoromanistentag (aka #frt2012), where I will present technical issues connected to the integration of francophone texts to a German edition, I have worked on all the small details of the interface that you only notice when you really do not understand them: the tags for the 6 different views of the document, the presentation texts for the corpora, the search page, the metadata view. You will certainly have noticed that the English equivalent is still missing, ahem. I spent hours improving the French part and will certainly try to move on with the English one as well soon, although this is far more painstaking, since translating German or French into English takes me considerably more time than translating German into French. But I promise I will have my English translations copyedited (for once, you will be able to read *real* English, not just my English).

We have also been working on the multilingual version of the metadata. To make those as accessible to data mining as humanly possible, we encode them in German, French and English. But is it necessary to translate all fields of the metadata? Everything that contains adresses, norms like dates or the size of the page in cm clearly does not need translation and is self-explainable in all three languages. While asking myself what really does deserve to appear in three languages, I found the translating process so boring and useless that I have quickly been dividing the fields concerned by the translation in two categories: those that need to be translated by a human being on the one hand (typically, full-text zones describing the overall looks and preservation of the manuscript) and on the other hand, those that should be translated by a computer. Countries names for instance. There are also fields that need not be translated at all because the XSLT-Stylesheet does it for me, such as the expression “Transcription by” or “Annotation by”. This lead me to some thinking about the level at which such translations are useful. Obviously, when it comes to reading the texts or searching marked-up elements of our digital edition, the translation through the XSLT-stylesheet suffices. But when it comes to reusing our XML file? I will have to look more precisely into that.

There was some hesitation as to the technical implementation as well. Just like for the first letter from Tieck to Raumer, on which I had done a series of fairly unsucessful multilingual experimentations, it turns out to be quite tricky to have three different variations of the same field that will only show in the one view corresponding to the language. For the notes  for instance, I should not add after one another < note resp=”#anne.baillot” xml:lang=”de”>Bla bla.< /note> and < note resp=”#anne.baillot” xml:lang=”fr”>Bla bla.< /note> and < note resp=”#anne.baillot” xml:lang=”en”>Blah blah.< /note>, but have only one < note> divided into three < seg>, each one of those being garnished with the capital xml:lang-information. It turns out to work the exact same way for most of the metadata fields. Except, said Alex, for title and titleStatement. What the heck! One way or another, I will probably begin using xml:lang while speaking very soon, considering how frantically I have been using while encodig lately.

This is almost what I will talk about in my paper tomorrow. Almost, because what I wrote today is more a further developpement of the ideas of my paper, and of the editorial praxis that I have tried to make fit to the paper’s ideas. The main argument of my paper is that it really does not matter in which language the text is that you want to edit. What matters is the language in which you edit it.

Which is why my paper is in French, the congress is taking place in Germany and I will be tweeting in English!

Of ghosts and metadata

While writing my paper for the Biography Conference and writing my paper for the Frankoromanistentag and trying to finish up the grant application for the Boeckh Encyklopädie and settling on the OCR issue, thus turning my intention to very different fields and objects, I have been wondering which of the research objects I am manipulating are real and alive and which are more speculative and less substantial, constructions of a ghostly nature if you want.

It was Markus Schnöpf who gave me the first hunch when we discussed the composition of the Berlin Digital Humanities Group and he underlined how important the museology fraction is. I was not really able to process all the implications of that consideration in that moment. I only thought: The hegemony of text sciences is a very narrow point of view, come to think about it. I have been so infatuated in texts for so long that I forgot that other items can be placed on the same level as texts. Especially items that have just the same archival function: the function to keep things alive through memory. But are archives alive, or are they only ghostly versions of what they used to be, fading away, unsubstantial, flickering flames? Are stones or papyri more alive than texts? Do we need the original materiality, the manuscript itself, almost as an object of adoration, to keep things like texts alive – knowing perfectly well that they decompose anyway, never stay the same, never are the same than they just have been a second before?

Then the ghosts turned up again in another, completely different discussion. While Laurent was (yet again) trying to convince me to have the OCRs done (don’t worry, Laurent, I am convinced now), he said to me that images like scans without searchable (even badly OCRed) text were invisible, nonexistent as digital research items. Here are my ghosts!, I thought. Unsubstantial images that are connected to nothing. What makes digital ghosts is their lack of connection, of traceability, their absence of metadata. Metadata are existence factors.

I understand and accept the crucial role metadata play in the digital world. Do I really have a choice about that? Whomever refuses to admit that metadata play such a central role inevitably gets lost in the feeling that the web is unstructured, that you can as much find anything as you can find nothing, that it is all a big fascinating whirl you cannot have any kind of grip on. A dance of shapeless ghosts. In that sense, I don’t think that any form of digital science can do without a really deep epistemological thinking on metadata and their structure. In other words: it is essential that we, the researchers, take part in the shaping of the metadata that are relevant for our field. And that we do it now, because in twenty years, we will only have our eyes to cry because others have done it for us in a way that does not fit our research objects.

On the other hand, it can and should not all be about the metadata. If I consider things not from a structural point of view, but from the point of view of the inevitably super-specialized research that makes me want to go to work every day, metadata are dead. Real life is elsewhere. Real life is when meaning emerges, and meaning can emerge from anything, but most likely it emerges from something. From a manuscript, from a discussion, from a fight for a scholarship that is able to see ghosts and keep their trace – whether alive or not.


Positively Dawesome Feature

There is no such word as dawesome, I know. But I could not think of an adjective starting with a d that would express fully my enthusiasm for our progress on the production of pdf-files on the basis of our XML/TEI documents.

It is only natural that a publisher would not be very accommodating when you come to him saying: “Hey, we will have everything online first: would like you to then make a book out of it?”. As much as I understand the publisher’s  reluctance though, fact is that this is what we need (and by we I don’t just mean me, but also the scientific community at large). The digital edition of the Encyklopädie will provide a dynamical work and query environment.  What should the paper version provide for?

I gave that question some thought yesterday and came up with a list of ideas and questions:

On the level of a single dataset

  • 2 buttons: pfd this document / pdf this page
  • On each pdf:

–          Metadata in italic on top

–          Abbreviations expanded

–          No line breaks, no gaps, no unclear passages, no strikethrough

–          Additions appear, being marked as different from the text (bold?)

–          Page breaks marked (page nb?)

–          What about third hands? Do we keep them in the pdf?

–          Changes in script (latin/german) marked

–          In case of marginalia, separation of the text’s body in 2 columns (maybe we need to encode this in the XML document?)

–          Tagged persons, works, places are ignored

–          <note> with a resp-attribute at the bottom of the page (separated by a short line)

–          Act and scene appear at the top of the page

–          Styles for titles to be defined

–          Possibility to generate several pdfs depending on the text layer? (Boeckh-Encyklopädie)


On the level of a subcorpus (1 „pdf this subcorpus“ button)

–          Front page (define styles)

–          Add a table of contents after front page (with page numbers)

–          Index of persons, works and places: after the main text – with or without page numbers?(the pdf is not supposed to be there to search something…)


In the following minutes, Anna reacted by saying we need to show gaps, etc. And it turned out, especially, when Bertrand Gaiffe sent us the first conversions he had made (without having gotten much information from us as to what we want), that it would be much better to work with a double footnote apparatus than with a second column.

Of course, making pdfs doesn’t equal making a book, by far. But concerning the Encyklopädie for instance, it would offer the possibility to generate a pdf for each historical layer of the text (1809, 1810/11, 1816, etc.), and the book version would provide for a recapitulation of the text in the version of 1865. My fear (that we would in the end not do something so much different from what Bratuschek did in 1877, e.g. a cumulative compilation) would not be justified. We really are doing something different. And we hope to find a publisher that sees the potential and the meaning of this too.

There is still a lot of work to do there. I hope to be able to post a nice sample in a few days already!