Encoding Guidelines for the ELTeC: level 1

1.1. Principles and sampling method

The goal of CA16204 is to create a benchmark corpus of literature from 1840-1920 adequate to the needs of many different computational distant reading methods for corpus annotation and analysis. The corpus design should support comparison of texts and individual sub-collections selected according to the metadata associated with each text. It should be possible to sample sub-collections from the ELTeC for specific tasks and research questions. In a first step, we focus on the development of clear, operationalized, transparent and motivated selection criteria for the corpus.

It is important to stress that we do not intend to define what a novel is by defining what kind of selection criteria we will use for ELTeC. The category novel may be divided into three groups where at least one of the following core criteria is met: a) textual: length (>10.000 words), prose, fiction, narrative structure b) paratextual (the term ‘novel’ or equivalent appears in the title or subtitle of the text) and c) contextual: the text is bibliographically listed with the UDC: 82-31 Novels. Full-length stories.

We follow a non-normative but metadata-based approach of sampling criteria which will follow a corpus design approach. Corpus sampling criteria are mostly oriented/developed by the research question or/and contexts of the corpus creators group. In CA16204, we have neither only a single research question nor a fixed and previously known corpus creator group. The research context of the Action is more interested in knowledge production in a methodological sense and does not prefer a single method, model or theory. Furthermore, the member group of the Action will fluctuate and consist of researches from different disciplines with different theoretical and cultural contexts. Thus, we need to build the corpus design on a methodical basis. With this method, we will also be able to select canonical texts as well but not exclusively.

Representativeness is a kind of ideal which we would like to pursue but which cannot be achieved as whole. We will therefore aim to represent the variety of a population. In line with the MoU, the ELTeC will be designed as a monitor corpus where texts (from different languages and periods) can be added over time. We then need to decide which criterion is balanced in which way and interplays with other criteria.

1.2. Requirements of sampling criteria

According to the MoU, the corpus design should be balanced with respect to language and publication date of the texts. This means that the corpus should not be based solely on chronological criteria, meaning that we need a text from each year of the period in question. The main sampling criterion ‘language’ will require not to include translations at all. We will prefer to take the first edition of a novel or editions of these novels. By a novel, we prefer to take the edition of the book, hence we don't prefer novels only printed in periodicals, unless a particular literary tradition only features novels printed in serial format. If we consider editions of a novel, these editions should be freely available (free licences for reusing them. The first edition is more interesting from a philological point of view. It represents the authentic texts of the authors. Dealing with historical texts might require some cleaning up or normalizations. We will merge all word forms which are separated by line breaks. At the moment, we must assume that there are no (sufficiently good) normalization tools for every language. Later editions of a novel may be already normalized in some way. This might lead to different text representations in ELTeC which should be indicated in the metadata.

Considering also later freely available editions of a novel has two advantages: First, members of the Action already can provide machine-readable text documents (html, TEI etc.) of later editions and second, in some languages it might be easier to find later editions which already exist in a machine-readable format (in this way we do not have to put effort in digitizing them).

Electronically availability should not be a leading sampling criterion although availability is a limiting factor. A text should not be excluded from ELTeC because it is not digitized, but it should be excluded if the text cannot be made freely available in ELTeC. If we only use availability as a selecting criterion, we are at risk of copying projects such as ‘Gutenberg’ for example. The issue remains of finding additional funds to digitise non-canonical books. Un til that moment, the solution would be to create pilot corpora (that can later be supplemented or substituted by an alterna tive) for literatures that do not have a significant number of digitized texts.

We then need additional criteria which can be applied without having to know (read) the texts in question. The criteria should be checked without a deep knowledge about the texts. Otherwise, this will oppose the goal of the whole Action and the methodical approach of distant reading. The criteria should be operationalizable, meaning decidable from text metadata. Here, we define text metadata in a wider scope than only the classical bibliographical metadata. In this way corpus design interacts with metadata. Some of the text’s metadata can be used as sampling criteria. These criteria are text-external and -internal criteria (cf. Hunston 2008) on which we then need to rely. The selection criteria may be assisted by bibliographical overviews (wherever available) for each language in order to avoid possible canon-derived bias.

We suggest using an online table as a means of collecting nominations for inclusion in the ELTeC but other methods are feasible.

1.3. Sampling criteria

For creating a language collection two steps have to be done: First step is selection: identifying text candidates. Second step is balancing: proportion within the corpus. Both steps are defined in this document.

Principles:

• We will represent the variety of production and aim to maximize the variety within each time period.
• We will prefer to collect novels published as a book in first as well as later editions over novels published in serial publications (journals).
• We will not include translations.
• We will consider only freely available texts and trying to reuse already digitized texts.
• We will use non-normative sampling criteria which allow for selecting canonical novels and non-canonical novels.
• Metadata should indicate the type of text representation (normalized or non-normalized, first or later editions etc.).
• We will strictly follow the selection and balancing criteria and we prefer 80 novels meeting these criteria over 100 novels which do not meet the criteria.

Organization

• Please note: Contact the sampling support Team: Carolin, Lou, Pieter and Diana! Task: To control the corpus sampling according to the selection criteria.

Criteria: Eligibility.

In order to be included (selection), a text must...

• have been first published as a book (or or in case of need as series publication) between 1840 and 1920, sub-grouped and equally split by decades,
• have first been published in a European country. [maybe not "first": within that decade],
• be a novel, i.e. a fictional prose narrative,
• have originally been written in the language of the given subcollection.

Criteria: Composition.

Among the novels in each language the subcollection must contain...

• at least 10%-50% have been written by female authors for the language subcollection.
• 9 to 11 authors are represented with exact three novels.
• at least 30% are highly canonical novels and at least 30% should be novels of low canonicity; canonicity is assessed by investigating the number of times a title was reprinted during the period 1970-2009.
• at least 20% are short novels (10-50k word tokens), at least 20% are long novels (>200k word tokens).

Date : 1840 to 1920 (first iteration)

We will divide into four groups

• group A (1840-1859): code T1
• group B (1860-1879): code T2
• group C (1880-1899): code T3
• group D (1900-1920): code T4

Language of the text

The MoU defines the languages to be sampled. It does not propose distinguishing regional variation (e.g. in German), nor geographical variation (e.g. the French spoken in Belgium, France, or Switzerland). It assumes only European varieties, so English excludes US English; French excludes Quebecois.

We follow a language-based approach (not country-based). This means for example that we include Swiss German texts in the German language collection. We prefer standard varieties over dialect varieties if sampling criteria for text candidates are met.

Reprint count

We propose to use the number of times a work is reprinted as an objective measure of its reception during the period 1970-2009, using categories like the following:

• low: few or no reprints found during this period,
• high: many reprints found during this period
• Note that the reprint count does not include digitizations of texts.

Author gender

We use the following three categories for actual (not claimed) author gender

• male
• female
• mixed (more than one author)
• undefined

Length

We include a variety of lengths

• short (10k-50k word tokens)
• medium (50k-100k word tokens)
• long (>100k word tokens)

2.1. Encoding Principles

The MoU for the project notes that ‘Distant Reading methods cover a wide range of computational methods for literary text analysis, such as authorship attribution, topic modelling, character network analysis, or stylistic analysis.’ The focus of the ELTeC encoding scheme is therefore not to represent texts in all their original complexity of structure or appearance, but rather to facilitate a richer and better-informed distant reading than a transcription of lexical content alone would permit. In designing this encoding scheme, we have applied the following principles:

• There should be a guaranteed minimum level of encoding for all texts in the ELTeC.
• The encoding should reflect the analytic needs of ‘distant reading’, rather than those of the digital editor.
• At the same time, the encoding should not misrepresent the text as originally created and received.
• The encoding must be easy to apply in a consistent way, starting from both basic OCR and pre-existing digital versions.

The goal is not to duplicate the work of scholarly editors or to produce (yet another) digital edition of a specific source document. Rather it is to ensure that the ELTeC texts can be processed satisfactorily, even by simple minded (but XML-aware) systems primarily concerned with lexis, and to make life easier for the developers of such systems.

In selecting features for inclusion in the markup scheme, we have been guided, but not limited, by existing practice as far as possible. Our main goal has been to identify a small core set of textual features which can be readily (preferably automatically) identified in existing digital transcriptions, or easily and consistently provided by new transcriptions.

We distinguish three ‘levels’ of encoding, referred to below as level zero, level one and level two. All ELTeC texts are made available at level zero, the basic encoding format. Some texts may additionally be made available at levels one or two, which provide a richer set of encoded features. For example: a level one text will include semantic information missing from a level zero text; a level two text will include tokenization information missing from a level one text. As far as possible conversion between levels will be automatically scripted, but this is not possible in the general case.

This document lists all the textual features which are to be distinguished in an ELTeC conformant transcription at one of these three levels. Whenever a given feature exists in a text, it will be marked up as indicated here. No other features will be captured by the markup: if some textual feature not provided for here is identified by a marked up source text, that markup will be removed (though it may be retained in a version of the text encoded at a different level).

All ELTeC documents are TEI conformant, and therefore include a TEI Header, as discussed in section 3.1. Metadata in the TEI Header below.

2.2. Basic Transcription Guidelines (all levels)

The basic unit of the ELTeC collection is a single novel, represented by a single <TEI> element, consisting of a <teiHeader> element containing metadata specific to that novel and a <text> element containing a normalised transcription of the text itself. We propose no mechanism (other than metadata) to encode units larger than a single novel, such as multipart novel series like Proust's A la recherche du temps perdu or Balzac's Les Rougon-Macquart. Each text should be transcribed in full from a specific identifiable edition, typically the first, and the source documented in the TEI Header. The original spelling and punctuation of the source should be retained, but details of typography are not required: hence words hyphenated across line or page breaks should be silently reassembled.

To facilitate checking of a transcription against its source during production, the <pb> element may be used to mark the point in a transcription where a new page begins in the source. An identifier for each <pb> element may be provided in a level 1 text to facilitate linkage to a page image of the corresponding source page. This element is not required in a level 0 text.

Running titles, page footers, catchwords and other forms of printed paratext are all omitted from an ELTeC transcription, with the exception of a page number, which may be supplied as value of the n attribute. Note that this attribute supplies the page number as specified by the source. If no page number is given, the value should be enclosed in brackets.

If a page begins with the second part of a hyphenated word, the <pb> tag may appear after that word in order to simplify lexical processing. Otherwise its position should be the same in transcription and source.

As well as a titlepage or a table of contents, a published novel often includes material such as forewords or appendixes additional to the text of the novel itself. This liminal matter is included in an ELTeC text only if it is believed to be authorial. Material before the body of the text begins is collected within a <front> element, and material following the body in a <back> element. In either case, distinct sections of the material, if encoded, are represented by a <div> with its type attribute set to liminal.

At level zero, titlepages and tables of contents are omitted. At level one, they are replaced by a <gap> element. Non-authorial liminal material is silently omitted at all levels.

Within the body of a text, major structural divisions (parts, sections, chapters etc.) will be captured using the generic <div> element, with attributes type, xml:lang, xml:id and n used as further detailed below.

The names used for hierarchic structural divisions of a novel above the chapter are arbitrary, culture-specific, and often inconsistent : in some novels things called ‘part’ contain things called ‘book’ and in others the reverse. We propose to follow TEI in using a single element (<div>) for every hierarchical structural division, down to the level of ‘chapter’.

The type attribute is used to indicate the function of a structural division. It should have one of the following values:

liminal

authorial preface, foreword, back matter, etc.

titlepage

(within front or back) contains title page text: level 1 only

notes

(within front or back) contains authorial notes : level 1 only

part

any structural subdivision of a text larger than a chapter e.g. part, volume etc.

chapter

smallest structural subdivision of the body of a text

A short novel may have no subdivision at all, in which case the <div> element should not be used. No further subdivisions within a <div type='chapter'> are permitted. If the text of a chapter is subdivided in some way, for example by means of a number, a row of stars, a horizontal rule, or similar device, this should be indicated in the markup by means of a <milestone> element. If a chapter contains an embedded text of some kind, for example a quoted letter or other narrative, this should be marked using the <quote> element.

The (human) language in which a text is expressed is indicated explicitly by the xml:lang attribute which supplies the ISO 641-2 letter code for the language concerned. This attribute will always be supplied on the <text> element to specify a default, and may also appear on other elements to indicate passages where the language changes. The various different languages used in a given text will be itemized in its metadata (see <langUsage> element in the header).

A single reference scheme will be defined for the whole corpus, with the following components:

• text identifier : every text will have an identifier consisting of its three letter language code in upper case and a five digit serial number, for example FRA00042
• chapter identifier: each chapter or equivalent will have an identifier concatenating the text identifier and a three digit serial number, for example FR042012 is the twelfth chapter of the 42nd French novel.
• If sub-chapter segmentation is implemented, then the segments will append a further four digit serial number.
• liminal divisions are not identified; the front element has an identifier concatenating the text identifier and the value 000; the back element has an identifier concatenating the text identifier and the value 999.

The identifier will be supplied as the value of an xml:id attribute on each <text>, <div>, front, back, or <s> element as appropriate. Adding this identifier is an easily automated task built into the workflow for accession to the ELTeC.

Note that these identifiers will not necessarily correspond with the numbering used in a particular source text. In a work where the first twelve chapters are considered to form part one, and the next twelve constitute part two, the first chapter of the second part will have an identifier ending 013, even though it may be numbered 1 in a source text.

We do not preserve the lineation of running prose in our source texts, since this is always purely an artefact of the source edition. For the same reason we reassemble words broken across a line break, silently removing any hyphen present. (This will make it impossible to use our texts for hyphenation studies. So be it.)

The title of a chapter, or of any other subdivision, as given in the source should be encoded using the TEI <head> element. There may be more than one such element at the start of a <div> element. Novels occasionally include other initial matter, such as a quotation, or a summary of the content of the chapter; these are not specially treated in ELTeC texts.

The chapters of a novel mostly consist of prose, arranged in paragraphs. It is not unusual to find other structures however, specifically verse, or passages of dialogue presented as if in a play, with speaker labels and even stage directions. Less frequently, novels may contain material presented in list or tabular formats. Graphics with their own associated heading or other text are also frequent.

Novels are also full of direct speech, represented using various different conventions, but almost always distinguished from the narrative voice. The first person narrative is also common, but may be regarded as a special case. How exactly different narrative strands are articulated in a novel, and the extent to which they may be characterised by their lexis has been a preoccupation of many ‘distant reading’ style analyses. Although it might be helpful to distinguish material purporting to be direct speech from material purporting to be narrative in our basic encoding, doing so consistently and accurately would be problematic. ELTeC texts therefore do no more than preserve existing punctuation. The <p> element is used for everything which is typeset as a separate block on the page, including both paragraphs and list items; the <l >element is used for verse lines or similar, typically set off from the rest of the text. Illustrations and any associated text such as a title or heading are excluded. Passages set as if in drama are not specially treated.

Printed texts typically deploy a number of conventions which can cause problems for linguistic analyses of even the most basic kind. Changes of font or style (italicization or use of superscript, for example) usually signal something, which an analysis should take into account. However, determining the function of such typographic variation is not always straightforward. ELTeC texts (at level 0) therefore simply indicate the presence of typographic salience, using the <hi> element.

2.3. Transcription Guidelines (level 1)

ELTeC encoding at level zero aims above all for consistency and transparency in what is reliably achievable, leaving most problematic issues to be addressed by linguistic annotation.

If however a text has been derived from a digital version in which a more ambitious range of textual features has already been captured, whether by means of TEI-style markup or styling information such as that provided by Word, or if there are sufficient resources available to provide a slightly reacher encoded version, a novel may be encoded at ELTeC level 1, using additional elements discussed in this section. Note that a level one text can always automatically be converted to a level zero text, if this is necessary for compatibility or for some other reason. (A script to do this is available on the project website) The reverse conversion, from level zero to level one, requires human intervention.

At level 1, the element <gap> should be used to indicate when something has been omitted from the encoding. For example, a suppressed graphic, or foreword, which at level zero are silently omitted, should be represented at level 1 by an explicit <gap> element, with attributes indicating what has been omitted from the encoding.

At level 1, the following additional elements can be used to mark up the significance of some stretch of text which would otherwise simply be marked as typographically salient using <hi>:

<foreign>

a word or phrase in a foreign language

<title>

the title of a book, piece of music, etc.

<label>

any form of heading or label found within a paragraph of text

<emph>

a word or phrase being linguistically emphasized e.g. by being shouted

The <hi> is also sometimes used for indications of superscript characters (such as French ‘14ᵉ’); these should simply be removed.

When a sequence of verse lines, or a passage from some other narrative level such as a letter is quoted within a text, the <l> or <p> elements representing it should be wrapped in the <quote> element made available by the ELTeC level one schema.

It is not unusual to find special devices such as a row of stars or a rule in the middle of a chapter, usually indicating a discontinuity in the narrative timeline, or structural shift. Such indications may be simply ignored in an ELTeC level zero text; at level 1, the special purpose <milestone> element may be used to mark their presence.

Level 1 texts may also represent authorial notes, if these are present, using the element <note> to contain the body of the note, and the element <ref> to represent the point of attachment for the note. Wherever they appear in the text, notes are always separated from it and encoded in a separate <div> element within the <back> element. See examples.

Occasionally, the text being transcribed contains self-evident errors. Where these are caused by the encoding process (e.g. an OCR error), these are always silently corrected in the transcription. Where however the original itself is faulty, however, and the transcriber (or a textual editor) has corrected it, the correction should be signalled by using the <corr> element available at level one.

2.4. Transcription Guidelines (level 2)

At ELTeC level2, all existing elements are retained and two new elements <s> and <w> are introduced to support segmentation of running text into sentence-like and word-like sequences respectively. Individual tokens are marked using the <w> element, and decorated with one or more of the TEI-defined linguistic attributes pos, lemma, and join. Both words and punctuation marks are considered to be ‘tokens’ in this sense, although the TEI suggests distinguishing the two cases using <w> and <pc> respectively. The <s> (segment) element is used to provide an end-to-end tessellating segmentation of the whole sequence of <w> elements, based on orthographic form. This provides a convenient extension of the existing text-body-div hierarchy within which tokens are located. The elements <p>, <head>, and <l> (which contain just text at levels 0 and 1) at level 2 can contain a sequence of <s> elements. Empty elements <gap>, <milestone>, <pb> or <ref> are also permitted within text content at any point, but these are disregarded when segmentation is carried out. Each <s> element can contain a sequence of <w> elements, either directly, or wrapped in one of the sub-paragraph elements <corr>, <emph>, <foreign>, <hi>, <label>, or <title>. To this list we add the element <rs> (referring string), provided by the TEI for the encoding of any form of entity name, such as a Named Entity Recognition procedure might produce.

This approach implies that <w> elements may appear at two levels in the hierarchy which may upset some software; it also implies that <w> elements must be properly contained within one of these elements, without overlap.

This TEI XML format is equally applicable to the production of training data for applications using machine learning techniques and to the outputs of such systems. However, since such machine learning applications typically operate on text content in a tabular format only, XSLT filters which transform (or generate) the XML markup discussed here from such tabular formats without loss of information are envisaged. At the time of writing, however, Working Group 2 has yet to put this proposed architecture to the test.

2.5. TEI Elements and textual features

The following summary table lists the textual features which every ELTeC text must capture, together with an indication of how that feature should be represented.

Other textual features are treated differently at different encoding levels. They are listed in the following table:

3.1. Metadata in the TEI Header

This section describes the metadata associated with each text (title, authorship, date etc.) and with the collection as a whole. The intention is to provide this in a standardised way to facilitate subsetting of the collection, using (for example) coded values for the descriptive selection criteria associated with the text. As far as possible, our text should represent the first complete printed edition of each novel selected.

The TEI Header provides a very large number of possibilities for encoding such metadata. We will provide a checklist of the TEI Header elements which are always to be provided for each text, possibly in the form of a template. As in the body of the text, the intention is to provide a guaranteed minimal level of information, consistent across all parts of the ELTeC.

Note that metadata may be supplied at (at least) two levels: the level of the ELTeC as a whole, and that of individual texts within it. Information which applies uniformly to all parts of the collection should be supplied in the ELTeC header; information specific to a particular document in the text header.

Every ELTeC text includes a TEI Header supplying metadata to describe it. There is also a TEI Header for the whole collection, which has additional information common to all the texts. This section lists the header components which should be supplied for every text, indicating briefly specific usage rules.

Each of the TEI elements shown above must be provided and used as described below. The ELTeC schemas will reject as invalid a document in which these conventions are not followed.

The natural language used for text in the Header should be that of the language collection to which the text belongs, e.g. French if the text is in French. The attribute @xml:lang may be supplied on any element to indicate its content is in some other language where necessary.

Appendix A.1 Elements

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <elementRef key="teiHeader"/>
  <alternate minOccurs="1" maxOccurs="1">
   <sequence minOccurs="1" maxOccurs="1">
    <classRef key="model.resource"
     minOccurs="1" maxOccurs="unbounded"/>
    <elementRef key="TEI" minOccurs="0"
     maxOccurs="unbounded"/>
   </sequence>
   <elementRef key="TEI" minOccurs="1"
    maxOccurs="unbounded"/>
  </alternate>
 </sequence>
</content>
    

element TEI
{
   att.global.attribute.n,
   att.global.attribute.xmlbase,
   att.global.attribute.xmlspace,
   att.global.rendition.attribute.rend,
   att.typed.attributes,
   attribute xml:id { text },
   attribute xml:lang { text },
   ( teiHeader, ( ( model.resource+, TEI* ) | TEI+ ) )
}

<content>
 <textNode/>
</content>
    

element author { att.canonical.attributes, att.global.attributes, text }

<content>
 <empty/>
</content>
    

element authorGender { attribute key { "M" | "F" | "U" | "X" }, empty }

<content>
 <elementRef key="licence"/>
</content>
    

element availability { att.global.attributes, licence }

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <alternate minOccurs="0"
   maxOccurs="unbounded">
   <classRef key="model.frontPart"/>
   <classRef key="model.pLike.front"/>
   <classRef key="model.pLike"/>
   <classRef key="model.listLike"/>
   <classRef key="model.global"/>
  </alternate>
  <alternate minOccurs="0" maxOccurs="1">
   <sequence minOccurs="1" maxOccurs="1">
    <classRef key="model.div1Like"/>
    <alternate minOccurs="0"
     maxOccurs="unbounded">
     <classRef key="model.frontPart"/>
     <classRef key="model.div1Like"/>
     <classRef key="model.global"/>
    </alternate>
   </sequence>
   <sequence minOccurs="1" maxOccurs="1">
    <classRef key="model.divLike"/>
    <alternate minOccurs="0"
     maxOccurs="unbounded">
     <classRef key="model.frontPart"/>
     <classRef key="model.divLike"/>
     <classRef key="model.global"/>
    </alternate>
   </sequence>
  </alternate>
  <sequence minOccurs="0" maxOccurs="1">
   <classRef key="model.divBottomPart"/>
   <alternate minOccurs="0"
    maxOccurs="unbounded">
    <classRef key="model.divBottomPart"/>
    <classRef key="model.global"/>
   </alternate>
  </sequence>
 </sequence>
</content>
    

element back
{
   att.global.attributes,
   (
      (
         model.frontPart
       | model.pLike.front
       | model.pLike
       | model.listLike
       | model.global
      )*,
      (
         (
            model.div1Like,
            ( model.frontPart | model.div1Like | model.global )*
         )
       | ( model.divLike, ( model.frontPart | model.divLike | model.global )* )
      )?,
      ( model.divBottomPart, ( model.divBottomPart | model.global )* )?
   )
}

<content>
 <alternate minOccurs="0"
  maxOccurs="unbounded">
  <textNode/>
  <classRef key="model.gLike"/>
  <classRef key="model.highlighted"/>
  <classRef key="model.pPart.data"/>
  <classRef key="model.pPart.edit"/>
  <classRef key="model.segLike"/>
  <classRef key="model.ptrLike"/>
  <classRef key="model.biblPart"/>
  <classRef key="model.global"/>
 </alternate>
</content>
    

element bibl
{
   att.global.attributes,
   att.sortable.attributes,
   attribute type
   {
      "firstEdition" | "printSource" | "digitalSource" | "unspecified"
   },
   (
      text
    | model.gLike
    | model.highlighted
    | model.pPart.data
    | model.pPart.edit
    | model.segLike
    | model.ptrLike
    | model.biblPart
    | model.global
   )*
}

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <classRef key="model.global"
   minOccurs="0" maxOccurs="unbounded"/>
  <sequence minOccurs="0" maxOccurs="1">
   <classRef key="model.divTop"/>
   <alternate minOccurs="0"
    maxOccurs="unbounded">
    <classRef key="model.global"/>
    <classRef key="model.divTop"/>
   </alternate>
  </sequence>
  <sequence minOccurs="0" maxOccurs="1">
   <classRef key="model.divGenLike"/>
   <alternate minOccurs="0"
    maxOccurs="unbounded">
    <classRef key="model.global"/>
    <classRef key="model.divGenLike"/>
   </alternate>
  </sequence>
  <alternate minOccurs="1" maxOccurs="1">
   <sequence minOccurs="1"
    maxOccurs="unbounded">
    <classRef key="model.divLike"/>
    <alternate minOccurs="0"
     maxOccurs="unbounded">
     <classRef key="model.global"/>
     <classRef key="model.divGenLike"/>
    </alternate>
   </sequence>
   <sequence minOccurs="1"
    maxOccurs="unbounded">
    <classRef key="model.div1Like"/>
    <alternate minOccurs="0"
     maxOccurs="unbounded">
     <classRef key="model.global"/>
     <classRef key="model.divGenLike"/>
    </alternate>
   </sequence>
   <sequence minOccurs="1" maxOccurs="1">
    <sequence minOccurs="1"
     maxOccurs="unbounded">
     <classRef key="model.common"/>
     <classRef key="model.global"
      minOccurs="0" maxOccurs="unbounded"/>
    </sequence>
    <alternate minOccurs="0" maxOccurs="1">
     <sequence minOccurs="1"
      maxOccurs="unbounded">
      <classRef key="model.divLike"/>
      <alternate minOccurs="0"
       maxOccurs="unbounded">
       <classRef key="model.global"/>
       <classRef key="model.divGenLike"/>
      </alternate>
     </sequence>
     <sequence minOccurs="1"
      maxOccurs="unbounded">
      <classRef key="model.div1Like"/>
      <alternate minOccurs="0"
       maxOccurs="unbounded">
       <classRef key="model.global"/>
       <classRef key="model.divGenLike"/>
      </alternate>
     </sequence>
    </alternate>
   </sequence>
  </alternate>
  <sequence minOccurs="0"
   maxOccurs="unbounded">
   <classRef key="model.divBottom"/>
   <classRef key="model.global"
    minOccurs="0" maxOccurs="unbounded"/>
  </sequence>
 </sequence>
</content>
    

element body
{
   att.global.attributes,
   (
      model.global*,
      ( model.divTop, ( model.global | model.divTop )* )?,
      ( model.divGenLike, ( model.global | model.divGenLike )* )?,
      (
         ( model.divLike, ( model.global | model.divGenLike )* )+
       | ( model.div1Like, ( model.global | model.divGenLike )* )+
       | (
            ( model.common, model.global* )+,
            (
               ( model.divLike, ( model.global | model.divGenLike )* )+
             | ( model.div1Like, ( model.global | model.divGenLike )* )+
            )?
         )
      ),
      ( model.divBottom, model.global* )*
   )
}

<content>
 <empty/>
</content>
    

element canonicity { attribute key { "high" | "low" | "unspecified" }, empty }

<content>
 <macroRef key="macro.specialPara"/>
</content>
    

element change
{
   att.datable.w3c.attribute.notBefore,
   att.datable.w3c.attribute.notAfter,
   att.datable.w3c.attribute.from,
   att.datable.w3c.attribute.to,
   att.global.attributes,
   att.typed.attributes,
   attribute when { text },
   macro.specialPara
}

<content>
 <macroRef key="macro.paraContent"/>
</content>
    

element corr { att.global.attributes, att.typed.attributes, macro.paraContent }

<content>
 <alternate minOccurs="0"
  maxOccurs="unbounded">
  <textNode/>
  <classRef key="model.gLike"/>
  <classRef key="model.phrase"/>
  <classRef key="model.global"/>
 </alternate>
</content>
    

element date
{
   att.global.attributes,
   att.canonical.attributes,
   att.datable.attributes,
   att.dimensions.attributes,
   att.typed.attributes,
   ( text | model.gLike | model.phrase | model.global )*
}

<content>
 <macroRef key="macro.phraseSeq"/>
</content>
    

element distributor
{
   att.global.attributes,
   att.canonical.attributes,
   macro.phraseSeq
}

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <alternate minOccurs="0"
   maxOccurs="unbounded">
   <classRef key="model.divTop"/>
   <classRef key="model.global"/>
  </alternate>
  <sequence minOccurs="0" maxOccurs="1">
   <alternate minOccurs="1" maxOccurs="1">
    <sequence minOccurs="1"
     maxOccurs="unbounded">
     <alternate minOccurs="1" maxOccurs="1">
      <classRef key="model.divLike"/>
      <classRef key="model.divGenLike"/>
     </alternate>
     <classRef key="model.global"
      minOccurs="0" maxOccurs="unbounded"/>
    </sequence>
    <sequence minOccurs="1" maxOccurs="1">
     <sequence minOccurs="1"
      maxOccurs="unbounded">
      <classRef key="model.common"/>
      <classRef key="model.global"
       minOccurs="0" maxOccurs="unbounded"/>
     </sequence>
     <sequence minOccurs="0"
      maxOccurs="unbounded">
      <alternate minOccurs="1"
       maxOccurs="1">
       <classRef key="model.divLike"/>
       <classRef key="model.divGenLike"/>
      </alternate>
      <classRef key="model.global"
       minOccurs="0" maxOccurs="unbounded"/>
     </sequence>
    </sequence>
   </alternate>
   <sequence minOccurs="0"
    maxOccurs="unbounded">
    <classRef key="model.divBottom"/>
    <classRef key="model.global"
     minOccurs="0" maxOccurs="unbounded"/>
   </sequence>
  </sequence>
 </sequence>
</content>
    

element div
{
   att.global.attributes,
   attribute type
   {
      "titlepage" | "notes" | "liminal" | "chapter" | "letter" | "group"
   }?,
   (
      ( model.divTop | model.global )*,
      (
         (
            ( ( model.divLike | model.divGenLike ), model.global* )+
          | (
               ( model.common, model.global* )+,
               ( ( model.divLike | model.divGenLike ), model.global* )*
            )
         ),
         ( model.divBottom, model.global* )*
      )?
   )
}

<content>
 <macroRef key="macro.paraContent"/>
</content>
    

element emph { att.global.attributes, macro.paraContent }

<content>
 <elementRef key="p"/>
</content>
    

element encodingDesc
{
   att.global.attribute.xmlid,
   att.global.attribute.xmllang,
   att.global.attribute.xmlbase,
   att.global.attribute.xmlspace,
   att.global.rendition.attribute.rend,
   attribute n { "eltec-0" | "eltec-1" | "eltec-2" },
   p
}

<content>
 <elementRef key="measure" minOccurs="1"
  maxOccurs="unbounded"/>
</content>
    

element extent { att.global.attributes, measure+ }

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <elementRef key="titleStmt"/>
  <elementRef key="extent"/>
  <elementRef key="publicationStmt"/>
  <elementRef key="sourceDesc"/>
 </sequence>
</content>
    

element fileDesc
{
   att.global.attributes,
   ( titleStmt, extent, publicationStmt, sourceDesc )
}

<content>
 <macroRef key="macro.phraseSeq"/>
</content>
    

element foreign { att.global.attributes, macro.phraseSeq }

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <alternate minOccurs="0"
   maxOccurs="unbounded">
   <classRef key="model.frontPart"/>
   <classRef key="model.pLike"/>
   <classRef key="model.pLike.front"/>
   <classRef key="model.global"/>
  </alternate>
  <sequence minOccurs="0" maxOccurs="1">
   <alternate minOccurs="1" maxOccurs="1">
    <sequence minOccurs="1" maxOccurs="1">
     <classRef key="model.div1Like"/>
     <alternate minOccurs="0"
      maxOccurs="unbounded">
      <classRef key="model.div1Like"/>
      <classRef key="model.frontPart"/>
      <classRef key="model.global"/>
     </alternate>
    </sequence>
    <sequence minOccurs="1" maxOccurs="1">
     <classRef key="model.divLike"/>
     <alternate minOccurs="0"
      maxOccurs="unbounded">
      <classRef key="model.divLike"/>
      <classRef key="model.frontPart"/>
      <classRef key="model.global"/>
     </alternate>
    </sequence>
   </alternate>
   <sequence minOccurs="0" maxOccurs="1">
    <classRef key="model.divBottom"/>
    <alternate minOccurs="0"
     maxOccurs="unbounded">
     <classRef key="model.divBottom"/>
     <classRef key="model.global"/>
    </alternate>
   </sequence>
  </sequence>
 </sequence>
</content>
    

element front
{
   att.global.attributes,
   (
      ( model.frontPart | model.pLike | model.pLike.front | model.global )*,
      (
         (
            (
               model.div1Like,
               ( model.div1Like | model.frontPart | model.global )*
            )
          | (
               model.divLike,
               ( model.divLike | model.frontPart | model.global )*
            )
         ),
         ( model.divBottom, ( model.divBottom | model.global )* )?
      )?
   )
}

<content>
 <alternate minOccurs="0"
  maxOccurs="unbounded">
  <classRef key="model.descLike"/>
  <classRef key="model.certLike"/>
 </alternate>
</content>
    

element gap
{
   att.global.attributes,
   att.dimensions.attributes,
   ( model.descLike | model.certLike )*
}

<content>
 <alternate minOccurs="0"
  maxOccurs="unbounded">
  <textNode/>
  <elementRef key="lg"/>
  <classRef key="model.gLike"/>
  <classRef key="model.phrase"/>
  <classRef key="model.inter"/>
  <classRef key="model.lLike"/>
  <classRef key="model.global"/>
 </alternate>
</content>
    

element head
{
   att.global.attributes,
   att.typed.attributes,
   (
      text
    | lg
    | model.gLike
    | model.phrase
    | model.inter
    | model.lLike
    | model.global
   )*
}

<content>
 <macroRef key="macro.paraContent"/>
</content>
    

element hi { att.global.attributes, macro.paraContent }

<content>
 <alternate minOccurs="0"
  maxOccurs="unbounded">
  <textNode/>
  <classRef key="model.gLike"/>
  <elementRef key="idno"/>
 </alternate>
</content>
    

element idno
{
   attribute type
   {
      "ISBN" | "ISSN" | "DOI" | "URI" | "VIAF" | "ESTC" | "OCLC"
   }?,
   ( text | model.gLike | idno )*
}

<content>
 <alternate minOccurs="1" maxOccurs="1">
  <elementRef key="term" minOccurs="1"
   maxOccurs="unbounded"/>
  <elementRef key="list"/>
 </alternate>
</content>
    

element keywords
{
   att.global.attributes,
   attribute scheme { text }?,
   ( term+ | list )
}

<content>
 <alternate minOccurs="0"
  maxOccurs="unbounded">
  <textNode/>
  <classRef key="model.gLike"/>
  <classRef key="model.phrase"/>
  <classRef key="model.inter"/>
  <classRef key="model.global"/>
 </alternate>
</content>
    

element l
{
   att.global.attributes,
   ( text | model.gLike | model.phrase | model.inter | model.global )*
}

<content>
 <macroRef key="macro.phraseSeq"/>
</content>
    

element label { att.global.attributes, att.typed.attributes, macro.phraseSeq }

<content>
 <alternate minOccurs="1" maxOccurs="1">
  <classRef key="model.pLike" minOccurs="1"
   maxOccurs="unbounded"/>
  <elementRef key="language" minOccurs="1"
   maxOccurs="unbounded"/>
 </alternate>
</content>
    

element langUsage { att.global.attributes, ( model.pLike+ | language+ ) }

<content>
 <macroRef key="macro.phraseSeq.limited"/>
</content>
    

element language
{
   att.global.attributes,
   attribute ident { text },
   attribute usage { text }?,
   macro.phraseSeq.limited
}

<content>
 <elementRef key="p" minOccurs="0"/>
</content>
    

element licence { att.global.attributes, attribute target { list { + } }, p? }

<content>
 <dataRef key="teidata.numeric"/>
</content>
    

element measure
{
   att.global.attributes,
   att.measurement.attribute.unitRef,
   att.measurement.attribute.quantity,
   att.measurement.attribute.commodity,
   attribute unit { "pages" | "words" | "vols" },
   teidata.numeric
}

<content>
 <empty/>
</content>
    

element milestone
{
   att.global.attributes,
   att.milestoneUnit.attributes,
   att.typed.attributes,
   empty
}

<content>
 <macroRef key="macro.phraseSeq"/>
</content>
    

element name
{
   att.global.attributes,
   att.datable.attributes,
   att.typed.attributes,
   macro.phraseSeq
}

<content>
 <macroRef key="macro.specialPara"/>
</content>
    

element note
{
   att.global.attributes,
   att.pointing.attributes,
   att.typed.attributes,
   macro.specialPara
}

<content>
 <macroRef key="macro.paraContent"/>
</content>
    

element p { att.global.attributes, macro.paraContent }

Appendix A.1.35 <pb>

<pb> (page beginning) marks the beginning of a new page in a paginated document. [3.11.3. Milestone Elements]
Module	core
Attributes	Attributes att.global (@xml:id, @n, @xml:lang, @xml:base, @xml:space) (att.global.rendition (@rend)) att.typed (@type) facs link to a graphic image of the page beginning here Status Optional Datatype anyURI
Member of	model.milestoneLike
Contained by	core: bibl corr date emph foreign head hi l label name note p pubPlace publisher quote ref resp term title header: change distributor language textstructure: back body div front text trailer
May contain	Empty element
Note	A <pb> element should appear at the start of the page which it identifies. The global n attribute indicates the number or other value associated with this page. This will normally be the page number or signature printed on it, since the physical sequence number is implicit in the presence of the <pb> element itself.
Example	A page break may be associated with a facsimile image of the page it introduces by means of the facs attribute <body>  <pb n="1" facs="page1.png"/> <!-- page1.png contains an image of the page; the text it contains is encoded here -->  <p> <!-- ... -->  </p>  <pb n="2" facs="page2.png"/> <!-- similarly, for page 2 -->  <p> <!-- ... -->  </p> </body>
Example	If a page break interrupts a word the word fragments should be reassembled following it. <p>My own relations too were nobly generous and by their kindness I have been <pb n="100"/> established in this shop, and for the last year have carried on this little business.... </p>
Content model	<content>  <empty/> </content>
Schema Declaration	element pb { att.global.attributes, att.typed.attributes, attribute facs { text }?, empty }

<content>
 <elementRef key="langUsage" minOccurs="1"
  maxOccurs="1"/>
 <elementRef key="textClass" minOccurs="0"
  maxOccurs="1"/>
 <elementRef key="textDesc" minOccurs="1"
  maxOccurs="1"/>
</content>
    

element profileDesc { att.global.attributes, langUsage, textClass?, textDesc }

<content>
 <macroRef key="macro.phraseSeq"/>
</content>
    

element pubPlace { att.global.attributes, macro.phraseSeq }

<content>
 <alternate minOccurs="1" maxOccurs="1">
  <elementRef key="p"/>
  <sequence minOccurs="1" maxOccurs="1">
   <elementRef key="publisher"/>
   <elementRef key="distributor"/>
   <elementRef key="date"/>
   <elementRef key="availability"/>
   <elementRef key="ref"
    maxOccurs="unbounded" minOccurs="0"/>
  </sequence>
 </alternate>
</content>
    

element publicationStmt
{
   att.global.attributes,
   ( p | ( publisher, distributor, date, availability, ref* ) )
}

<content>
 <macroRef key="macro.phraseSeq"/>
</content>
    

element publisher
{
   att.global.attributes,
   att.canonical.attributes,
   macro.phraseSeq
}

<content>
 <macroRef key="macro.specialPara"/>
</content>
    

element quote
{
   att.global.attributes,
   att.typed.attributes,
   att.notated.attributes,
   macro.specialPara
}

<content>
 <macroRef key="macro.paraContent"/>
</content>
    

element ref
{
   att.global.attributes,
   att.pointing.attributes,
   att.typed.attributes,
   macro.paraContent
}

<content>
 <empty/>
</content>
    

element reprintCount
{
   attribute key { "high" | "low" | "unspecified" },
   empty
}

<content>
 <macroRef key="macro.phraseSeq.limited"/>
</content>
    

element resp
{
   att.global.attributes,
   att.canonical.attributes,
   att.datable.attributes,
   macro.phraseSeq.limited
}

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <alternate minOccurs="1" maxOccurs="1">
   <sequence minOccurs="1" maxOccurs="1">
    <elementRef key="resp" minOccurs="1"
     maxOccurs="unbounded"/>
    <classRef key="model.nameLike.agent"
     minOccurs="1" maxOccurs="unbounded"/>
   </sequence>
   <sequence minOccurs="1" maxOccurs="1">
    <classRef key="model.nameLike.agent"
     minOccurs="1" maxOccurs="unbounded"/>
    <elementRef key="resp" minOccurs="1"
     maxOccurs="unbounded"/>
   </sequence>
  </alternate>
  <elementRef key="note" minOccurs="0"
   maxOccurs="unbounded"/>
 </sequence>
</content>
    

element respStmt
{
   att.global.attributes,
   att.canonical.attributes,
   (
      ( ( resp+, model.nameLike.agent+ ) | ( model.nameLike.agent+, resp+ ) ),
      note*
   )
}

<content>
 <alternate minOccurs="1" maxOccurs="1">
  <elementRef key="list"/>
  <elementRef key="listChange"/>
  <elementRef key="change" minOccurs="1"
   maxOccurs="unbounded"/>
 </alternate>
</content>
    

element revisionDesc
{
   att.global.attributes,
   ( list | listChange | change+ )
}

<content>
 <empty/>
</content>
    

element size { attribute key { "long" | "medium" | "short" }, empty }

<content>
 <alternate minOccurs="1" maxOccurs="1">
  <classRef key="model.pLike" minOccurs="1"
   maxOccurs="unbounded"/>
  <alternate minOccurs="1"
   maxOccurs="unbounded">
   <classRef key="model.biblLike"/>
   <classRef key="model.sourceDescPart"/>
   <classRef key="model.listLike"/>
  </alternate>
 </alternate>
</content>
    

element sourceDesc
{
   att.global.attributes,
   (
      model.pLike+
    | ( model.biblLike | model.sourceDescPart | model.listLike )+
   )
}

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <elementRef key="fileDesc"/>
  <elementRef key="encodingDesc"/>
  <elementRef key="profileDesc"/>
  <elementRef key="revisionDesc"/>
 </sequence>
</content>
    

element teiHeader
{
   att.global.attributes,
   ( fileDesc, encodingDesc, profileDesc, revisionDesc )
}

<content>
 <macroRef key="macro.phraseSeq"/>
</content>
    

element term
{
   att.global.attributes,
   att.typed.attributes,
   att.canonical.attributes,
   macro.phraseSeq
}

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <classRef key="model.global"
   minOccurs="0" maxOccurs="unbounded"/>
  <sequence minOccurs="0" maxOccurs="1">
   <elementRef key="front"/>
   <classRef key="model.global"
    minOccurs="0" maxOccurs="unbounded"/>
  </sequence>
  <alternate minOccurs="1" maxOccurs="1">
   <elementRef key="body"/>
   <elementRef key="group"/>
  </alternate>
  <classRef key="model.global"
   minOccurs="0" maxOccurs="unbounded"/>
  <sequence minOccurs="0" maxOccurs="1">
   <elementRef key="back"/>
   <classRef key="model.global"
    minOccurs="0" maxOccurs="unbounded"/>
  </sequence>
 </sequence>
</content>
    

element text
{
   att.global.attributes,
   att.typed.attributes,
   (
      model.global*,
      ( front, model.global* )?,
      ( body | group ),
      model.global*,
      ( back, model.global* )?
   )
}

<content>
 <elementRef key="keywords"/>
</content>
    

element textClass { att.global.attributes, keywords }

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <elementRef key="authorGender"/>
  <elementRef key="size"/>
  <alternate minOccurs="1" maxOccurs="1">
   <elementRef key="canonicity"/>
   <elementRef key="reprintCount"/>
  </alternate>
  <elementRef key="timeSlot"/>
 </sequence>
</content>
    

element textDesc
{
   att.global.attributes,
   ( authorGender, size, ( canonicity | reprintCount ), timeSlot )
}

<content>
 <empty/>
</content>
    

element timeSlot { attribute key { "T1" | "T2" | "T3" | "T4" }, empty }

<content>
 <macroRef key="macro.paraContent"/>
</content>
    

element title
{
   att.global.attributes,
   att.canonical.attributes,
   attribute level { "a" | "m" | "j" | "s" | "u" }?,
   macro.paraContent
}

<content>
 <sequence minOccurs="1" maxOccurs="1">
  <elementRef key="title" minOccurs="1"
   maxOccurs="unbounded"/>
  <classRef key="model.respLike"
   minOccurs="0" maxOccurs="unbounded"/>
 </sequence>
</content>
    

element titleStmt { att.global.attributes, ( title+, model.respLike* ) }