In preparing an annotated edition of the complete correspondence between the Flemish author Stijn Streuvels (1871-1969) and his Dutch and Flemish publishers, a corpus of a couple of thousand letters need to be transcribed and encoded in a machine readable format. Following up on previous research (Vanhoutte 2001), thorough analyses of the TEI tag sets currently available and the structural and 'letter-specific' features to be encoded in the corpus, show that the TEI tag sets have to be stretched in order to be useful in the project. It was necessary for instance to develop new tags for encoding the information on the envelope, such as the postmark, which can be of importance for a correct situation of the letter in the reconstruction of the dialog between both the letter writers (Eide 2001). The most radical extension, however, is the introduction of a means to encode the complex network of different "layers" which we can distinguish in a letter. In the case of the Streuvels corpus, the text of a letter is for instance often written on headed notepaper, and accompanied by notes which show the reading marks of both the recipient of the letters and readers other than the respective authors, such as secretaries, etc. This way the letter as a physical document often holds multiple texts by multiple authors. It is obvious that when using the corpus as a basis for stylometrical research or to study the evolution of language in letter writing, we do want to be able to distinguish among the several authors, in order to get correct results. Therefore, the tags <layer> and <layerList> were introduced in the DTD (developed by Bert Van Elsacker at the Centre for Scholarly Editing and Document Studies). By making use of the attributes author, scribe, style, ink and character, we can give a global identification a layer in a document. (This element has been modelled after <hand> in the TEI additional tag set for the tanscription of primary sources). We defined <layerList>(after <handList>) as a child of <sourceDesc> inside the <teiHeader>. An empty element <lsh /> with attributes place, right, left, and character marks the shift from one layer to the next. This triumvirate is inspired by the additional tag set for the transcripion of primary sources, but differs from it in a couple of aspects. Whereas the "hand-system" assumes that the text is sequentially organised, and uses <handShift> to mark a change in a continous stream of text (e.g. a different colour of ink, a feeble hand, a different scribe), the "layer-system" is explicitly developed to handle discontinuous texts, in which "isles of texts" by different authors, scribes, or of a different type - it sometimes happens that the letter is organised around a newspaper clipping which of course becomes part of the letter, but is not by the author - can appear in a discontinous order, and can refer to each other. That is why <lsh /> has the attribute place. Notwithstanding the fact that an XML-document is sequentially organized, adapted software can separate the "isles of text" and render them according to user preferences.
In this paper I will comment on the advantages of applying the extended DTD to the material and the envisioned uses of the corpus in fields other than literary criticism. Further I will situate the project in the larger Digital Archive of Letters written by Flemish authors and composers in the 19th and 20th century (DALF) and will demonstrate an early version of the electronic edition.
