Functions of Punctuation

Punctuation is itself a form of markup, historically introduced to provide the reader with an indication about how the text should be read. As such, it is unsurprising that encoders will often wish to encode directly the purpose for which punctuation was provided, as well as, or even instead of, the punctuation itself. We discuss some typical cases below.

The Full stop (period) may mark (orthographic) sentence boundaries, abbreviations, decimal points, or serve as a visual aid in printing numbers. These usages can be distinguished by tagging S-units, abbreviations, and numbers, as described in sections , , and respectively. However, there are independent reasons for tagging these, whether or not they are marked by full stops, and the polysemy of the full stop itself is perhaps no different from that of any other character in the writing system.

The Question mark and exclamation mark usually mark the end of orthographic sentences, but may also be used as a mid-sentence comment by the author (! to express surprise or some other strong feeling, ? to query a word or expression or mark a sentence as dubious in linguistic discussion). Such usages may be distinguished by marking S-units, in which case the mid-sentence uses of these punctuation marks may be left unmarked, or tagged using the pc element discussed in .

Dashes are used for a variety of purposes: as a mark of omission, insertion, or interruption; to show where a new speaker takes over (in dialogue); or to introduce a list item. In the latter two cases particularly, it is clearly desirable to mark the function as well as its rendition using the elements q or item, on which see section , and section , respectively.

Quotation marks may be removed from text contained by q or quote elements on editorial grounds, or they may be marked in a variety of ways; see the discussion of quotation and related features in section .

Apostrophes must be distinguished from single quote marks. As with hyphens, this disambiguation is best performed by selecting the appropriate Unicode character, though it may also be represented by using appropriate XML markup for quotations as suggested above. However, apostrophes have a variety of uses. In English they mark contractions, genitive forms, and (occasionally) plural forms. Full disambiguation of these uses belongs to the level of linguistic analysis and interpretation.

Parentheses and other marks of suspension such as dashes or ellipses are often used to signal information about the syntactic structure of a text fragment. Full disambiguation of their uses also belongs to the level of linguistic analysis and interpretation, and will therefore need to use the mechanisms discussed in chapter .

Where punctuation marks are disambiguated by tagging their assumed function in the text (for example, quotation), it may be debated whether they should be excluded or left as part of the text. In the case of quotation marks, it may be more convenient to distinguish opening from closing marks simply by using the appropriate Unicode character than to use the q element, with or without an indication of rendition.

Where segmentation of a text is performed automatically, the accuracy of the result may be considerably enhanced by a first pass in which the function of different punctuation characters is explicitly marked. This need not be done for all cases, but only where the structural function of the punctuation markup (for example as a word or phrase delimiter) is ambiguous. Thus, dots indicating abbreviation might be distinguished from dots indicating sentence end, and exclamation or question marks internal to a sentence distinguished from those which terminate one. Furthermore, when encoding historical materials, it may be considered essential to retain the original punctuation, whether by using an appropriate character code, if this is available (or using the g element where it is not) or by an explicit encoding using pc. The particular method adopted will vary depending upon the feature concerned and upon the purpose of the project.

Hyphenation

Hyphenation as a phenomenon is generally of most concern when producing formatted text for display in print or on screen: different languages and systems have developed quite sophisticated sets of rules about where hyphens may be introduced and for what reason. These generally do not concern the text encoder, since they belong to the domain of formatting and will generally be handled by the rendition software in use. In this section, we discuss issues arising from the appearance of hyphens in pre-existing formatted texts which are being re-encoded for analysis or other processing. Unicode distinguishes four characters visually similar to the hyphen, including the undifferentiated hyphen-minus (U+002D) which is retained for compatibility reasons. The hard hyphen (U+2010) is distinguished from the minus sign (U+2212) which is for use in mathematical expressions, and also from the soft hyphen (U+00AD) which may appear in born digital documents to indicate places where it is acceptable to insert a hyphen when the document is formatted.

Historically, the hard hyphen has been used in printed or manuscript documents for two distinct purposes. In many languages, it is used between words to show that they function as a single syntactic or lexical unit. For example, in French, est-ce que; in English body-snatcher, tea-party etc. It may also have an important role in disambiguation (for example, by distinguishing say a man-eating fish from a man eating fish). Such usages, although possibly problematic when a linguistic analysis is undertaken, are not generally of concern to text encoders: the hyphen character is usually retained in the text, because it may be regarded as part of the way a compound or other lexical item is spelled. Deciding whether a compound is to be decomposed into its constituent parts, and if so how, is a different question, involving consideration of many other phenomena in addition to the simple presence of a hyphen.

When it appears at the end of a printed or written line however, the hard hyphen generally indicates that—contrary to what might be expected—a word is not yet complete, but continues on the next line (or over the next page or column or other boundary). The hyphen character is not, in this case, part of the word, but just a signal that the word continues over the break. Unfortunately, few languages distinguish these two cases visually, which necessarily poses a problem for text encoders. Suppose, for example, that we wish to investigate a diachronic English corpus for occurrences of "tea-pot" and "teapot", to find evidence for the point at which this compound becomes lexicalized. Any case where the word is hyphenated across a linebreak, like this: is entirely ambiguous: there is simply no way of deciding which of the two spellings was intended.

As elsewhere, therefore, the encoder has a range of choices: They may decide simply to remove any end-of-line hyphenation from the encoded text, on the grounds that its presence is purely a secondary matter of formatting. This will obviously apply also if line endings are themselves regarded as unimportant. Alternatively, they may decide to record the presence of the hyphen, perhaps on the grounds that it provides useful morphological information; perhaps in order to retain information about the visual appearance of the original source. In either case, they need to decide whether to record it explicitly, by including an appropriate punctuation character in the text data, or implicitly by supplying an appropriate symbolic value for one or more of the attributes on the lb or other milestone element used to record the fact of the line division. If the hyphen is included in the character data of the TEI document, it might be marked up using the pc (punctuation character) tag, which allows the encoder to express information about its function as a separator, through the force attribute (see ). A similar range of possibilities applies equally to the representation of other common punctuation marks, notably quotation marks, as discussed in .

The text data of which XML documents are composed is decomposable into smaller units, here called orthographic tokens, even if those units are not explicitly indicated by the XML markup. The ambiguity of the end-of-line hyphen also causes problems in the way a processor identifies such tokens in the absence of explicit markup. If token boundaries are not explicitly marked (for example using the seg or w elements), for most languages a processor will rely on character class information to determine where they are to be found: some punctuation characters are considered to be word-breaking, while others are not. In XML, the newline character in text data is a kind of whitespace, and is therefore word breaking. However, it is generally unsafe to assume that whitespace adjacent to markup tags will always be preserved, and it is decidedly unsafe to assume that markup tags themselves are equivalent to whitespace.

The lb, pb, and cb elements are notable exceptions to this general rule, since their function is precisely to represent (or replace) line, page, or column breaks, which, as noted above, are generally considered to be equivalent to whitespace. These elements provide a more reliable way of preserving the lineation, pagination, etc of a source document, since the encoder should not assume that (untagged) line breaks etc. in an XML source file will necessarily be preserved.

To control the intended tokenization, the encoder may use the break attribute on such elements to indicate whether or not the element is to be regarded as equivalent to whitespace. This attribute can take the values yes or no to indicate whether or not the element corresponds with a token boundary. The value maybe is also available, for cases where the encoder does not wish (or is unable) to determine whether the orthographic token concerned is broken by the line ending.

As a final complication, it should be noted that in some languages, particularly German and Dutch, the spelling of a word may be altered in the presence of end of line hyphenation. For example, in Dutch, the word opaatje (granddad), occurring at the end of a line may be hyphenated as opa-tje, with a single letter a. An encoder wishing to preserve the original form of this orthographic token in a printed text while at the same time facilitating its recognition as the word opaatje will therefore need to rely on a more sophisticated process than simply removing the hyphen. This is however essentially the same as any other form of normalization accompanying the recognition of variations in spelling or morphology: as such it may be encoded using the choice element discussed in , or the more sophisticated mechanisms for linguistic analysis discussed in chapter .

What Is Highlighting?

By highlighting we mean the use of any combination ofhighlighting typographic features (font, size, hue, etc.) in a printed or written text in order to distinguish some passage of a text from its surroundings.Although the way in which a spoken text is performed, (for example, the voice quality, loudness, etc.) might be regarded as analogous to highlighting in this sense, these Guidelines recommend distinct elements for the encoding of such highlighting in spoken texts. See further section . The purpose of highlighting is generally to draw the reader's attention to some feature or characteristic of the passage highlighted; this section describes the elements recommended by these Guidelines for the encoding of such textual features.

In conventionally printed modern texts, highlighting is often employed to identify words or phrases which are regarded as being one or more of the following: distinct in some way—as foreign, dialectal, archaic, technical, etc. emphatic, and which would for example be stressed when spoken not part of the body of the text, for example cross-references, titles, headings, labels, etc. identified with a distinct narrative stream, for example an internal monologue or commentary. attributed by the narrator to some other agency, either within the text or outside it: for example, direct speech or quotation. set apart from the text in some other way: for example, proverbial phrases, words mentioned but not used, names of persons and places in older texts, editorial corrections or additions, etc.

The textual functions indicated by highlighting may not be rendered consistently in different parts of a text or in different texts. (For example, a foreign word may appear in italics if the surrounding text is in roman, but in roman if the surrounding text is in italics.) For this reason, these Guidelines distinguish between the encoding of rendering itself and the encoding of the underlying feature expressed by it.

Highlighting as such may be encoded by using one of the global attributes rend, rendition, or style (see further ). This allows the encoder both to specify the function of a highlighted phrase or word, by selecting the appropriate element described here or elsewhere in the Guidelines, and to further describe the way in which it is highlighted, by means of an attribute. If the encoder wishes to offer no interpretation of the feature underlying the use of highlighting in the source text, then the hi element may be used, which indicates only that the text so tagged was highlighted in some way. The hi element is provided by the model.hiLike class.

The possible values carried by the rend attribute are not formally defined in this version of the Guidelines. It may be used to document any peculiarity of the way a given segment of text was rendered in the original source text, and may thus express a very large range of typographic or other features, by no means restricted to typeface, type size, etc. The style attribute, by contrast, defines the way the source text was rendered using a formally defined style language, such as the W3C standard Cascading Stylesheet Language (). The complementary rendition attribute is used to point to one or more fragments expressed using such a language which have been predefined in the TEI header using the rendition element discussed in section .

Where it is both appropriate and feasible, these Guidelines recommend that the textual feature marked by the highlighting should be encoded, rather than just the simple fact of the highlighting. This is for the following reasons: the same kind of highlighting may be used for different purposes in different contexts the same textual function may be highlighted in different ways in different contexts for analytic purposes, it is in general more useful to know the intended function of a highlighted phrase than simply that it is distinct.

In many, if not most, cases the underlying function of a highlighted phrase will be obvious and non-controversial, since the distinctions indicated by a change of highlighting correspond with distinctions discussed elsewhere in these Guidelines. The elements available to record such distinctions are, for the most part, members of the model.emphLike class. This and the model.hiLike class mentioned above constitute the model.highlighted class, which is a phrase level class. Members of this class may appear anywhere within paragraph level elements.

The distinction between the two classes is simple, and typified by the two elements hi and emph: the former marks simply that a passage is typographically distinct in some way, while the latter asserts that a passage is linguistically emphasized for some purpose. These two properties, though often combined, are not identical. It should however be recognized, however, that cases do exist in which it is not economically feasible to mark the underlying function (e.g. in the preparation of large text corpora), as well as cases in which it is not intellectually appropriate (as in the transcription of some older materials, or in the preparation of material for the study of typographic practice). In such cases, the hi element or some other element from the model.hiLike class should be used.

Elements which are sometimes realized by typographic distinction but which are not discussed in this section include title (discussed in section ) and name (discussed in section ).

Emphasis, Foreign Words, and Unusual Language

This subsection discusses the following elements: These elements are all members of the model.emphLike class.

Quotation

One form of presentational variation found particularly frequently in written and printed texts is the use of quotation marks. As with the typographic variations discussed in the preceding section, it is generally helpful to separate the encoding of the underlying textual feature (for example, a quotation or a piece of direct speech) from the encoding of its rendering (for example, the use of a particular style of quotation marks).

This section discusses the following elements, all of which are often rendered by the use of quotation marks: The elements mentioned and soCalled are members of the class model.emphLike; the q and said are members of the class model.qLike in their own right, while cit and quote are members of model.quoteLike, a subclass of model.qLike. This class is a subclass of model.inter; hence all of these elements are permitted both within and between paragraph-level elements.

The most common and important use of quotation marks is, of course, to mark quotation, by which we mean simply any part of the text which the author or narrator wishes to attribute to some agency other than the narrative voice. The q element may be used if no further distinction beyond this is judged necessary. If it is felt necessary to distinguish such passages further, for example to indicate whether they are regarded as speech, writing, or thought, either the type attribute or one of the more specialized elements discussed in this section may be used. For example, the element quote may be used for written passages cited from other works, or the element said for words or phrases represented as being spoken or thought by people or characters within the current work. The soCalled element is used for cases where the author or narrator distances him or herself from the words in question without however attributing them to any other voice in particular. The mentioned element is appropriate for a case where a word or phrase is being discussed in the body of a text rather than forming part of the text directly.

As noted above, if the distinction among these various reasons why a passage is offset from surrounding text cannot be made reliably, or is not of interest, then any representation of speech, thought, or writing may simply be marked using the q element.

Quotation may be indicated in a printed source by changes in type face, by special punctuation marks (single or double or angled quotes, dashes, etc.) and by layout (indented paragraphs, etc.), or it may not be explicitly represented at all. If these characteristics are of interest, one or other of the global rend or rendition attributes discussed in section may be used to record them.

Quotation marks themselves may, like other punctuation marks, be felt for some purposes to be worth retaining within a text, quite independently of their description by the rend attribute. This should generally be done using the appropriate Unicode character, or, if this is not possible, a numeric character reference (see ). If the encoder decides both to retain the quotation marks and to represent their function by means of an explicit tag such as quote, the quotation marks should be included within the element, rather than outside it, as in the first example below: Adolphe se tourna vers lui : — Alors, Albert, quoi de neuf? — Pas grand-chose. — Il fait beau, dit Robert. Alternatively, since this use of the leading mdash is very common typographic practice, it may be considered unnecessary to retain it in the encoding. Its presence in the source might instead be signalled using one of the attributes rend, style, or rendition. This kind of rendering might be predefined using a rendition element, which can then be referenced using the rendition attribute as follows: Adolphe se tourna vers lui : Alors, Albert, quoi de neuf ? Pas grand-chose. Il fait beau, dit Robert. content: '— '

Whatever policy is adopted, the encoder should document the decision in some way, for example by using the quotation element provided in the TEI header (see ) to indicate that quotation marks have not been retained in the encoding; their presence in the source is implied by the rendition attribute values supplied.

Whether or not the quotation marks are suppressed, their presence and nature may be described using some appropriate set of conventions in the rend attribute. These conventions may be entirely idiosyncratic, and hence not necessarily useful for interchange, as in the following example: Who-e debel you? — he at last said — you no speak-e, damme, I kill-e. And so saying, the lighted tomahawk began flourishing about me in the dark.

Such passages might more effectively be encoded without loss of rendering information by using the rendition attribute and its associated rendition element as described in section . If the rendition of passages tagged as said is uniform throughout a text, then the render attribute of the tagUsage element in the header may be used to specify a default rendering, in which case the same section might simply be tagged: Who-e debel you? — he at last said — you no speak-e, damme, I kill-e. And so saying, the lighted tomahawk began flourishing about me in the dark. content:"‘"; content:"’";

As members of the att.ascribed class, elements said and q share the following attribute: This may be used to make explicit who is speaking: Adolphe se tourna vers lui : — Alors, Albert, quoi de neuf? — Pas grand-chose. — Il fait beau, dit Robert. The who attribute may be supplied whether or not an indication of the speaker is given explicitly in the text. It may take the form (as above) of a normalized form of the speaker's name, but its role is to act as a pointer to a location elsewhere in the text, or another document, where data about each speaker may be supplied. While this attribute could point to any source of information about the speaker available by a URI, the most appropriate place to place such information is within the participant description component of the TEI header, as further discussed in but for simple cases like the above, a simple list of speakers located in the front or back matter of the text may suffice.

It may also be useful to distinguish representations of speech from representations of thought, in modern printed texts often indicated by a change of typeface. The aloud attribute is provided for this purpose, as in this example: Oh yes, said Henry, I mean Gordon Macrae, for example… Jungian Analyst with Winebox! That's what you called him, you callous bastard, didn't you? Eh? Eh?

Quoted matter may be embedded within quoted matter, as when one speaker reports the speech of another: Spaulding, he came down into the office just this day eight weeks with this very paper in his hand, and he says:— I wish to the Lord, Mr. Wilson, that I was a red-headed man. Wilson Spaulding reported by Wilson

Direct speech nested in this way is treated in the same way as elsewhere: a change of rendition may occur, but the same element should be used. An encoder may however choose to distinguish between direct speech which contains quotations from extra-textual matter and direct speech itself, as in the following example:

The Lord! The Lord! It is Sakya Muni himself, the lama half sobbed; and under his breath began the wonderful Buddhist invocation:- To Him the Way — the Law — Apart — Whom Maya held beneath her heart Ananda's Lord — the Bodhisat And He is here! The Most Excellent Law is here also. My pilgrimage is well begun. And what work! What work!

Quotations from other works are often accompanied by a reference to their source. The cit element may be used to group together the quotation and its associated bibliographic reference, which should be encoded using the elements for bibliographic references discussed in section , as in the following example.

Like other bibliographic references, the citation associated with a quotation may be represented simply by a cross-reference, as in this example: Lexicography has shown little sign of being affected by the work of followers of J.R. Firth, probably best summarized in his slogan, You shall know a word by the company it keeps. (Firth, 1957) It is also common for quotations to be separated from their bibliographic reference by intervening text, which makes the use of cit impractical. In such circumstances, the quotation can be linked to a bibliographical reference using source: Tolkien (1936) tells us that is in fact so interesting as poetry, in places poetry so powerful, that this quite overshadows the historical content. source could also be used to point to a complete external bibliographic reference in a listBibl elsewhere in the document, or external to it.

Unlike most of the other elements discussed in this chapter, direct speech and quotations may frequently contain other high-level elements such as paragraphs or verse lines, as well as being themselves contained by such elements. Three possible solutions exist for this well-known structural problem: the quotation is broken into segments, each of which is entirely contained within a paragraph the quotation is marked up using stand-off markup the quotation boundaries are represented by empty segment boundary delimiter elements For further discussion and several examples, see chapter .

Finally, in this section, the element soCalled is provided for all cases in which quotation marks are used to distance the quoted text from the narrator or speaker. Common examples include the scare quotes often found in newspaper headlines and advertising copy, where the effect is to cast doubts on the veracity of an assertion: PM dodges election threat in interview

The same element should be used to mark a variety of special ironic usages. Some further examples follow: He hated good books. Croissants indeed! toast not good enough for you? Although Chomsky's decision that all NL sentences are finite objects was never justified by arguments from the attested properties of NLs, it did have a certain social justification. It was commonly assumed in works on logic until fairly recently that the notion language is necessarily restricted to finite strings.

Terms, Glosses, Equivalents, and Descriptions

This section describes a set of textual elements which are used to provide a gloss, alternate identification, or description of something.

Technical terms are often italicized or emboldened upon first mention in printed texts; an explanation or gloss is sometimes given in quotation marks. Linguistic analyses conventionally cite words in languages under discussion in italics, providing a gloss immediately following marked with single quotation marks. Other texts in which individual words or phrases are mentioned (formentionvs. useusevs. mention example, as examples) rather than used may mark them either with italics or with quotation marks, and will gloss them less regularly. These elements are also members of the class model.emphLike.

A term may appear with or without a gloss, as may a mentioned element. Where the gloss is present, it may be linked to the term it is glossing by means of its target attribute. To establish such a link, the encoder should give an xml:id value to the term or mentioned element and provide that id as the value of the target attribute on the gloss element. The following examples demonstrate this facility:

Examples: We may define discoursal point of view as the relationship, expressed through discourse structure, between the implied author or some other addresser, and the fiction. A computational device that infers structure from grammatical strings of words is known as a parser, and much of the history of NLP over the last 20 years has been occupied with the design of parsers.

Note that the element term is intended for use with words or phrases identified as terminological in nature; where words or phrases are simply being cited, discussed, or glossed in a text, it will often be more appropriate to use the mentioned element, as in the following example: There is thus a striking accentual difference between a verbal form like eluthemen we were released, accented on the second syllable of the word, and its participial derivative lutheis released, accented on the last.

For technical terminology in particular, and generally in terminological studies, it may be useful to associate an instance of a term within a text with a canonical definition for it, which is stored either elsewhere in the same text (for example in a glossary of terms) or externally, for example in a database, authority file, or published standard. The attributes key and ref discussed in section below are available on the term element for this purpose.

Another group of elements is used to supply different kinds of names for objects described by the TEI. Examples of this are documentation of elements, attributes, classes (and also attribute values where appropriate), and description of glyphs. Along with the gloss element mentioned above, these elements constitute the model.glossLike class. They are described in more detail in .

Some Further Examples

As a simple example of the elements discussed here, consider the following sentence: On the one hand the is associated with the new rise of romance of twelfth-century France, the romans d'antiquité, the romances of Chrétien de Troyes, and the German adaptations of these works by Heinrich van Veldeke, Hartmann von Aue, and Wolfram von Eschenbach. A first approximation to the encoding of this sentence might be simply to record the fact that the phrases printed above in italics are highlighted, as follows: On the one hand the Nibelungenlied is associated with the new rise of romance of twelfth-century France, the romans d'antiquité, the romances of Chrétien de Troyes, ... This encoding would, however, lose the important distinction between an italicized title and an italicized foreign phrase. Many other phrases might also be italicized in the text, and a retrieval program seeking to identify foreign terms (for example) would not be able to produce reliable results by simply looking for italicized words. Where economic and intellectual constraints permit, therefore, it would be preferable to encode both the function of the highlighted phrases and their appearance, as follows: On the one hand the is associated with the new rise of romance of twelfth-century France, the romans d'antiquité, the romances of Chrétien de Troyes, ...

In this example, the decision as to which textual features are distinguished by the highlighting is relatively uncontroversial. As a less straightforward example, consider the use of italic font in the following passage: A pretty common case, I believe; in all vehement debatings. She says I am too witty; Anglicé, too pert; I, that she is too wise; that is to say, being likewise put into English, not so young as she has been: in short, she is grown so much into a mother, that she had forgotten she ever was a daughter. ...

Clearly, the word vehement is not italicized for the same reason as the phrase not so young as she has been; the former is emphasized, while the latter is proverbial. It also provides an ironic gloss for the words too wise, in the same way as too pert glosses too witty. The glossed phrases are not, however, technical terms or cited words, but quoted phrases, as if the writer were putting words into her own and her mother's mouths. Finally, the words mother and daughter are apparently italicized simply to oppose them in the sentence; certainly they do not fit into any of the categories so far proposed as reasons for italicizing. Note also that the word Anglicé is not italicized although it is not generally considered an English word.

The following sample encoding for the above passage attempts to take into account all the above points: A pretty common case, I believe; in all vehement debatings. She says I am too witty; Anglicé, too pert; I, that she is too wise; that is to say, being likewise put into English, not so young as she has been: in short, she is grown so much into a mother, that she had forgotten she ever was a daughter.

Apparent Errors

When the copy text is manifestly faulty, an encoder or transcriber may elect simply to correct it without comment, although for scholarly purposes it will often be more generally useful to record both the correction and the original state of the text. The elements described here enable all three approaches, and allows the last to be done in such a way as make it easy for software to present either the original or the correction.

The following examples show alternative treatment of the same material. The copy text reads: Another property of computer-assisted historical research is that data modelling must permit any one textual feature or part of a textual feature to be a part of more than one information model and to allow the researcher to draw on several such models simultaneously, for example, to select from a machine-readable text those marginal comments which indicate that the date's mentioned in the main body of the text are incorrect.

An encoder may choose to correct the typographic error, either silently or with an indication that a correction has been made, as follows: … marginal comments which indicate that the dates mentioned in the main body of the text are incorrect.

Alternatively, the encoder may simply record the typographic error without correcting it, either without comment or with a sic element to indicate the error is not a transcription error in the encoding: … marginal comments which indicate that the date's mentioned in the main body of the text are incorrect.

If the encoder elects both to record the original source text and to provide a correction for the sake of word-search and other programs, both sic and corr are used, wrapped in a choice: … marginal comments which indicate that the dates date's mentioned in the main body of the text are incorrect.The sic and corr elements can appear in either order.

If it is desired to indicate the person or edition responsible for the emendation, this might be done as follows: … marginal comments which indicate that the dates date's mentioned in the main body of the text are incorrect. editor C.M. Sperberg-McQueen Here the resp attribute has been used to indicate responsibility for the correction. Its value (#msm) is an example of the pointer values discussed in section ; in this case, it points to a respStmt element within the TEI header, but any element might be indicated in this way, including for example a name element, or (if the module described in has been included) a person element. The resp attribute is available for all elements which are members of the att.global.responsibility class. The same class makes available a cert attribute, which may be used to indicate the degree of editorial confidence in a particular correction, as in the following example: An AutumnAntony it was, That grew the more by reaping See further the discussion in section .

Where, as here, the correction takes the form of adding text not otherwise present in the text being encoded, the encoder should use the corr element. Where the correction is present in the text being encoded, and consists of some combination of visible additions and deletions, the elements add or del should be used: see further section below. Where the correction takes the form of addition of material not present in the original because of physical damage or illegibility, the supplied element may be used. Where the correction is simply a matter of expanding an abbreviation the ex element may be used. These and other elements to support the detailed encoding of authorial or scribal interventions of this kind are all provided by the module described in chapter .

Regularization and Normalization

When the source text makes extensive use of variant forms or non-standard spellings, it may be desirable for a number of reasons to regularize it: that is, to provide standard or regularized forms equivalent to the non-standard forms.In some contexts, the term regularization has a narrower and more specific significance than that proposed here: the reg element may be used for any kind of regularization, including normalization, standardization, and modernization.

As with other such changes to the copy text, the changes may be made silently (in which case the TEI header should specify the types of silent changes made) or may be explicitly marked using the following elements:

Typical applications for these elements include the production of editions intended for student or lay readers, linguistic research in which spelling or usage variation is not the main question at issue, production of spelling dictionaries, etc.

Consider this 16th-century text: how godly a dede it is to overthrowe so wicked a race the world may judge: for my part I thinke there canot be a greater sacryfice to God.

An encoder may choose to preserve the original spelling of this text, but simply flag it as nonstandard by using the orig element with no attributes specified, as follows:

...how godly a dede it is to overthrowe so wicked a race the world may judge: for my part I thinke there canot be a greater sacryfice to God

Alternatively, the encoder may simply indicate that certain words have been modernized by using the reg element with no attributes specified, as follows:

...how godly a deed it is to overthrow so wicked a race the world may judge: for my part I think there cannot be a greater sacrifice to God.

Alternatively, the encoder may elect to record both old and new spellings, so that (for example) the same electronic text may serve as the basis of an old- or new-spelling edition:

...how godly a dededeed it is to overthroweoverthrow so wicked a race the world may judge: for my part I thinkethink there canotcannot be a greater sacryficesacrifice to God.

As elsewhere, the resp attribute may be used to specify the agency responsible for the regularization.

Additions, Deletions, and Omissions

The following elements are used to indicate when words or phrases have been omitted from, added to, or marked for deletion from, a text. Like the other editorial elements, they allow for a wide range of editorial practices:

Encoders may choose to omit parts of the copy text for reasons ranging from illegibility of the source or impossibility of transcribing it, to editorial policy, e.g. a systematic exclusion of poetry or prose from an encoding. The full details of the policy decisions concerned should be documented in the TEI header (see section ). Each place in the text at which omission has taken place should be marked with a gap element, with optionally further information about the reason for the omission, its extent, and the person or agency responsible for it, as in the following examples: Note that the extent of the gap may be marked precisely using attributes unit and quantity, or more descriptively using the extent attribute. Other, more detailed, options are also available for representing dimensions of any kind; see further .

The desc element may be used to supply a description of the material omitted, where that is considered useful: irrelevant commentary … Their arrangement with respect to Jupiter and to each other was as follows: astrological figure That is, there were two stars on the easterly side and one to the west; …

The add and del elements may be used to record where words or phrases have been added or deleted in the copy text. They are not appropriate where longer passages have been added or deleted, which span several elements; for these, the elements addSpan and delSpan described in chapter must be used.

Additions to a text may be recorded for a number of reasons. Sometimes they are marked in a distinctive way in the source text, for example by brackets or insertion above the line (supralinear insertion),additionssupralinearinsertionssupralinearsupralinear insertions as in the following example, taken from a 19th century manuscript: The story I am going to relate is true as to its main facts, and as to the consequences of these facts from which this tale takes its title.

The add element should not be used to mark editorial changes, such as supplying a word omitted by mistake from the source text or a passage present in another version. In these cases, either the corr or supplied tags should be used, as discussed above in section , and in section , respectively.

The unclear element is used to mark passages in the original which cannot be read with confidence, or about which the transcriber is uncertain for other reasons, as for example when transcribing a partially inaudible or illegible source. Its reason and resp attributes are used, as with the gap element, to indicate the cause of uncertainty and the person responsible for the conjectured reading.

For example: And where the sandy mountain Fenwick scald The sea between yet hence his pray'r prevail'd or from a spoken text:

... and then marbled queen...

Where the material affected is entirely illegible or inaudible, the gap element discussed above should be used in preference.

The del element is used to mark material which is deleted in the source but which can still be read with some degree of confidence, as opposed to material which has been omitted by the encoder or transcriber either because it is entirely illegible or for some other reason. This is of particular importance in transcribing manuscript material, though deletion is also found in printed texts, sometimes for humorous purposes: One day I will sojourn to your shores I live in the middle of England But! Norway! My soul resides in your watery fiords fyords fiiords Inlets.

The rend attribute may be used to distinguish different methods of deletion in manuscript or typescript material, as in this line from the typescript of Eliot's : Mein Frisch schwebt weht der Wind

Deletion in manuscript or typescript is often associated with addition: Inviolable Inexplicable splendour of Corinthian white and gold The subst element discussed in provides a way of grouping additions and deletions of this kind.

The del element should not be used where the deletion is such that material cannot be read with confidence, or read at all, or where the material has been omitted by the transcriber or editor for some other reason. Where the material deleted cannot be read with confidence, the unclear tag should be used with the reason attribute indicating that the difficulty of transcription is due to deletion. Where material has been omitted by the transcriber or editor, this may be indicated by use of the gap element. A deletion in which some parts may be read but not others may thus be represented by one or more gap elements intermingled with text, all contained by a del element. Text supplied or marked as unneccessary by an editor should be marked with the supplied and surplus elements (discussed in ) rather than add and del. These two sets of elements allow the encoder to distinguish editorial changes from those visible in the source text.

Referring Strings

A referring string is a phrase which refers to some person, place, object, etc. Two elements are provided to mark such strings: The name element is a member of the att.typed class, from which it inherits the following attributes: which may be used to further categorize the kind of object referred to. The rs element defines the type attribute locally.

Examples include:

My dear Mr. Bennet, said his lady to him one day, have you heard that Netherfield Park is let at last?

Collectors of water-rents were appointed by the Watering Committee. They were paid a commission not exceeding four per cent, and gave bond.

It being one of the principles of the Circumlocution Office never, on any account whatsoever, to give a straightforward answer, Mr Barnacle said, Possibly.

As the following example shows, the rs element may be used for any reference to a person, place, etc., not only to references in the form of a proper noun or noun phrase.

My dear Mr. Bennet, said his lady to him one day ...

The name element by contrast is provided for the special case of referencing strings which consist only of proper nouns; it may be used synonymously with the rs element, or nested within it if a referring string contains a mixture of common and proper nouns. The following example shows an alternative way of encoding the short sentence from quoted above:

My dear Mr. Bennet, said his lady to him one day, have you heard that Netherfield Park is let at last?

As the following example shows, a proper name may be nested within a referring string: His Excellency the Life President, Ngwazi Dr H. Kamuzu Banda

Simply tagging something as a name is generally not enough to enable automatic processing of personal names into the canonical forms usually required for reference purposes. The name as it appears in the text may be inconsistently spelled, partial, or vague. Moreover, name prefixes such as van or de la may or may not be included as part of the reference form of a name, depending on the language and country of origin of the bearer.

Two issues arise in this context: firstly, there may be a need to encode a regularized form of a name, distinct from the actual form in the source to hand; secondly, there may be a need to identify the particular person, place, etc. referred to by the name, irrespective of whether the name itself is normalized or not. The element reg, introduced in is provided for the former purpose; the attributes key or ref for the latter.

The key and ref attributes are common to all members of the att.canonical class and are defined as follows:

A very useful application for them is as a means of gathering together all references to the same individual or location scattered throughout a document:

My dear Mr. Bennet, said his lady to him one day, have you heard that Netherfield Park is let at last?

Mme. de Volanges marie sa fille: c'est encore un secret; mais elle m'en a fait part hier.

The value of the key attribute may be an unexpanded code, as in the examples above, with no particular significance. More usually however, it will be an externally defined code of some kind, as provided by a standard reference source.

Heathrow

The standard reference source should be documented using a taxonomy element in the TEI header.

The ref attribute can be used to point directly to some other resource providing more information about the entity named by the element, such as an authority record in a database, an encylopaedia entry, another element in the same or a different document etc.

Heathrow

This use should be distinguished from the use of a nested reg (regularization) element to provide the standard form of a referring string, as in this example:

My personal life during the administration of Col. Polk (Polk, James K.) has but poorly compensated me for the suspended enjoyments and pursuits of private and professional spheres

No particular syntax is proposed for the values of the key attribute, since its form will depend entirely on practice within a given project. For the same reason, this attribute is not recommended in data interchange, since there is no way of ensuring that the values used by one project are distinct from those used by another. In such a situation, a preferable approach for magic tokens which follows standard practice on the Web is to use a ref attribute whose value is a tag URI as defined in RFC 4151. For example:

Mme. de Volanges marie sa fille: c'est encore un secret; mais elle m'en a fait part hier.

The inclusion of the domain name of the party responsible for tagging (theworksoflaclos.org), as specified in RFC 4151, helps ensure uniqueness of magic token values across TEI encoding projects, allowing for improved interchange of TEI documents.

The choice element discussed in may be used if it is desired to record both a normalized form of a name and the name used in the source being encoded:

Walter de la Mare de la Mare, Walter was born at Charlton, in Kent, in 1873.

The index element discussed in may be more appropriate if the function of the regularization is to provide a consistent index:

Montaillou is not a large parish. At the time of the events which led to FournierBenedict XII, Pope of Avignon (Jacques Fournier)'s investigations, the local population consisted of between 200 and 250 inhabitants.

Although adequate for many simple applications, these methods have two inconveniences: if the name occurs many times, then its regularized form must be repeated many times; and the burden of additional XML markup in the body of the text may be inconvenient to maintain and complex to process. For applications such as onomastics, relating to persons or places named rather than the name itself, or wherever a detailed analysis of the component parts of a name is needed, the specialized elements described in chapter or the analytical tools described in chapter should be used.

Addresses

These Guidelines propose the following elements to distinguish postal and electronic addresses: These two elements constitute the class of model.addressLike elements; for other kinds of address this class may be extended by adding new elements if necessary.

These Guidelines provide no particular means for encoding the substructure of an email address (for example, distinguishing the local part from the domain part), nor of distinguishing personal email addresses from generic or fictitious ones. info@tei-c.org

The simplest way of encoding a postal address is to regard it as a series of distinct lines, just as they might be written on an envelope. The following element supports this view: Here is an example of a postal address encoded using this approach:

110 Southmoor Road, Oxford OX2 6RB, UK

Alternatively, an address may be encoded as a structure of more semantically rich elements. The class model.addrPart element class identifies a number of such possible components: Any number of elements from the model.addrPart class may appear within an address and in any order. None of them is required.

Where code letters are commonly used in addresses (for example, to identify regions or countries) a useful practice is to supply the full name of the region or country as the content of the element, but to supply the abbreviatory code as the value of the global n attribute, so that (for example) an application preparing formatted labels can readily find the required information. Other components of addresses may be represented using the general-purpose name element or (when the additional module for names and dates is included) the more specialized elements provided for that purpose.

Using just the elements defined by the core module, the above address could thus be represented as follows:

110 Southmoor Road Oxford OX2 6RB United Kingdom

The order of elements within an address is highly culture-specific, and is therefore unconstrained:

Università di Bologna Italy 40126 Bologna via Marsala 24

A telephone number (normally outside of the address element) might be tagged with an addrLine and ref, with the number itself appearing in the tel: namespace: (201) 555 0123

For further discussion of ways of regularizing the names of places, see section . A full postal address may also include the name of the addressee, tagged as above using the general purpose name element.

When a schema includes the names and dates module discussed in chapter , a large number of more specific elements such as country or settlement will be available from the class model.addrPart. The above example might then be encoded as follows:

110 Southmoor Road Oxford OX2 6RB United Kingdom

Numbers and Measures

This section describes elements provided for the simple encoding of numbers and measurements and gives some indication of circumstances in which this may usefully be done. The following phrase level elements are provided for this purpose:

Like names or abbreviations, numbers can occur virtually anywhere in a text. Numbers are special in that they can be written with either letters or digits (twenty-one, xxi, and 21) and their presentation is language-dependent (e.g. English 5th becomes Greek 5.; English 123,456.78 equals French 123.456,78).

For many kinds of application, e.g. natural-language processing or machine translation, numbers are not regarded as lexical in the same way as other parts of a text. For these and other applications, the num element provides a convenient method of distinguishing numbers from the surrounding text. For other kinds of application, numbers are only useful if normalized: here the num element is useful precisely because it provides a standardized way of representing a numerical value.

For example: xxxiii twenty-one ten percent 10% 5th one half 1/2

Sometimes it may be desired to mark something as numerical which cannot be accurately normalized, for example an expression such as dozens; less frequently the number may be recognisable linguistically as such but may use a notation with which the encoder is unfamiliar. To help in these situations, the num element may also bear either or both of the following attributes from the att.ranging class:

In its fullest form, a measure consists of a number, a phrase expressing units of measure and a phrase expressing the commodity being measured, though not all of these components need be present in every case. It may be helpful to distinguish measures from surrounding text for two reasons. Firstly, a measure may be expressed using a particular notation or system of abbreviations which the encoder does not wish to regard as lexical. Secondly, a quantitative application may wish to distinguish and normalize the internal components of a measure, in order to perform calculations on them.

Consider, as an example of the first case, the following list of Celia's charms, in which the encoder has chosen to make explicit the measurements:

In the same way, it may be convenient to mark representations of currency which might otherwise be misinterpreted as lexical:

...the sum of 12s 6d...

In general, normalization of a measure will require specification of one or more of its three parts: the quantity, the units, and possibly also the commodity being measured. This is accomplished by supplying values for the three attributes quantity, unit, and commodity, which are supplied by the att.measurement class: With these attributes, the measurement of Celia's neck may be specified in a normalized form: 13¾" Such techniques are particularly useful when representing historical data such as inventories: ii bags hops six trusses Woolen and linen goods 5 tonnes coale

The measureGrp element is provided as a means of grouping several related measurements together, either because the measurement involves several dimensions (for example height and width) or to avoid the need to repeat all the normalizing attributes: xiv v x

Dates and Times

Dates and times, like numbers, can appear in widely varying culture- and language-dependent forms, and can pose similar problems in automatic language processing. Such elements constitute the model.dateLike class, of which the default members are: These elements have some additional attributes by virtue of being members of the att.datable and att.duration classes which, in turn, are members of the att.datable.w3c and att.duration.w3c classes. In particular, the when and calendar attributes will be discussed here:

Dates can occur virtually anywhere in a text, but in some contexts (e.g. bibliographic citations) their encoding is recommended or required rather than optional. Times can also appear anywhere but are generally optional.

Partial dates or times (e.g. 1990, September 1990, twelvish) can be expressed in the when attribute by simply omitting a part of the value supplied. Imprecise dates or times (for example early August, some time after ten and before twelve) may be expressed as date or time ranges.

These mechanisms are useful primarily for fully specified dates or times known with certainty. If component parts of dates or times are to be marked up, or if a more complex analysis of the meaning of a temporal expression is required, the techniques described in chapter should be used in preference to the simple method outlined here.

Where the certainty (i.e. reliability) of the date or time is in question, the encoder should record this fact using the mechanisms discussed in chapter . The same chapter also discusses various methods of recording the precision of numerical or temporal assertions.

The when attribute is a useful way of normalizing or disambiguating dates and times which can appear in many formats, as the following examples show: 12/2/1980 Given on the Twelfth Day of June in the Year of Our Lord One Thousand Nine Hundred and Seventy-seven of the Republic the Two Hundredth and first and of the University the Eighty-Sixth.

The when attribute always supplies a normalized representation of the date given as content of the date element. The format used should be a valid W3C schema datatype.The datatypes are taken from the W3C Recommendation . The permitted datatypes are: date gYear gMonth gDay gYearMonth gMonthDay time dateTime There is one exception: these Guidelines permit a time to be expressed as only a number of hours, or as a number of hours and minutes, as per ISO 8601:2004 section 4.2.2.3 and 4.3.3. The W3C time and dateTime datatypes require that the minutes and seconds be included in the normalized value if they are to be correctly processed for example when sorting. Some typical examples follow: The year 2001 September 2001 11 Sep 01 9/11 September Eleventh of the month 8:48 Sept 11th, 12 minutes before 9 am Note in the last example the use of a normalized representation for the date string which includes a time: this example could thus equally well be tagged using the time element.

The following examples demonstrate the use of the date element to mark a period of time:

Those five years — 1918 to 1923 — had been, he suspected, somehow very important.

The Eddic poems are preserved in a unique manuscript (Codex Regius 2365) from the second half of the thirteenth century, and dates from around 1300.

The calendar attribute may be used to specify a date in any calendar system; if the when attribute is also supplied, it should specify the equivalent date in the Gregorian calendar.

Abbreviations and Their Expansions

It is sometimes desirable to mark abbreviations in the copy text, whether to trigger special processing for them, to provide the full form of the word or phrase abbreviated, or to allow for different possible expansions of the abbreviation. Abbreviations may be transcribed as they stand, or expanded; they may be left unmarked, or marked using these tags:

The abbr element is useful as a means of distinguishing semi-lexical items such as acronyms or jargon: We can sum up the above discussion as follows: the identity of a CC is defined by that calibration of values which motivates the elements of its GSP; ... Every manufacturer of 3GL or 4GL languages is currently nailing on OOP extensions.

The type attribute may be used to distinguish types of abbreviation by their function:Dr. M. Deegan is the Director of the CTI Centre for Textual Studies.

Abbreviations such as Dr. M. above may be treated as two abbreviations, as above, or as one: Dr. M. Deegan is the Director of the CTI Centre for Textual Studies.

The expan element may be used simply to record that an abbreviation has been silently expanded by the encoder, perhaps for reasons of house style or editorial policy. It should always include the whole of an abbreviated phrase or word. More usually however this will be combined with the abbr element inside a choice element to record both the abbreviation and its expansion: the World Wide Web Consortium W3C Nested abbreviations may also be handled in this way: RELAXNGregular language for XMLextensible markup language, next generation

Abbreviation is a particularly important feature of manuscript and other source materials, the transcription of which needs more detailed treatment than is possible using these simple elements. A more detailed set of recommendations is discussed in , which includes additional elements made available for the purpose by the transcr module.

Notes and Simple Annotation

The following element is provided for the encoding of discursive notes, whether already present in the copy text or supplied by the encoder:

A note is any additional comment found in a text, marked in some way as being out of the main textual stream. All notes should be marked using the same tag, note, whether they appear as block notes in the main text area, at the foot of the page, at the end of the chapter or volume, in the margin, or in some other place.

Notes may be in a different hand or typeface, may be authorial or editorial, and may have been added later. Attributes may be used to specify these and other characteristics of notes, as detailed below.

A note is usually attached to a specific point or span within a text, which we term here its point of attachment. In conventional printed text, the point of attachment is represented by some siglum such as a star or cross, or a superscript digit.

When encoding such a text, it is conventional to replace this siglum by the content of the annotation, duly marked up with a note element. This may not always be possible for example with marginal notes, which may not be anchored to an exact location. For ease of processing, it may be adequate to position marginal notes before the relevant paragraph or other element. In printed texts, it is sometimes conventional to group notes together at the foot of the page on which their points of attachment appear. This practice is not generally recommended for TEI-encoded texts, since the pagination of a particular printed text is unlikely to be of structural significance. In some cases, however, it may be desirable to transcribe notes not at their point of attachment to the text but at their point of appearance, typically at the end of the volume, or the end of the chapter. In such cases, the target attribute of the note may be used to indicate the point of attachment. It is also possible to encode the point of attachment itself, using the ptr or ref element, pointing from that to the body of the note placed elsewhere.

In cases where the note is applied not to a point but to a span of text, not itself represented as a TEI element, the target attribute may use an appropriate pointer expression, for example using the range() function to specify the span of attachment.

For further discussion of pointing to points and spans in the text, see section .

In the following example, the type attribute is used to categorise the note as a gloss: The self-same moment I could pray And from my neck so free The albatross fell off, and sank Like lead into the sea. The spell begins to break As the note appears within an l element, we may infer that its point of attachment is in the margin adjacent to the line in question. In the following version of the same text, however, it may be inferred that the note applies to the whole of the stanza: The self-same moment I could pray And from my neck so free The albatross fell off, and sank Like lead into the sea. The spell begins to break

This type of annotation, very common in the early printed texts which Coleridge may be presumed to be imitating in this case, may also be regarded as providing a heading or descriptive label for the passage concerned. The encoder may therefore prefer to use the label element to represent it, as in the following case: The self-same moment I could pray And from my neck so free The albatross fell off, and sank Like lead into the sea. The spell begins to break

In the following example, a note which appears at the foot of the page in the printed source is given at its point of attachment within the text. The global n attribute is used to indicate the note number: Collections are ensembles of distinct entities or objects of any sort.We explain below why we use the uncommon term collection instead of the expected set. Our usage corresponds to the aggregate of many mathematical writings and to the sense of class found in older logical writings. The elements ...

In addition to transcribing notes already present in the copy text, researchers may wish to add their own notes or comments to it. The note element may be used for either purpose, but it will usually be advisable to distinguish the two categories. One way might be to use the type attribute shown above, categorizing notes as authorial, editorial, etc. Where notes derive from many sources, or where a more precise attribution is required, the resp attribute may be used to point to a definition of the person or other agency responsible for the content of the note.

As a simple example, an edition of the might include both Coleridge's original glosses and those of a modern commentator: The self-same moment I could pray; The spell begins to break The turning point of the poem... For this to be valid, the codes #JLL and #STC must point to some more information identifying the agency concerned. The syntax used is identical to that used for other cross-references, as discussed in ; thus in this case, the TEI header for this text might contain a title statement like the following: Samuel Taylor Coleridge John Livingston Lowes

When annotating the electronic text by means of analytic notes in some structured vocabulary, e.g. to specify the topics or themes of a text, the span and interp elements may be more effective than the free form note element; these elements are available when the module for simple analysis is selected (see section ).

Index Entries

The indexing of scholarly texts is a skilled activity, involving substantial amounts of human judgment and analysis. It should not therefore be assumed that simple searching and information retrieval software will be able to meet all the needs addressed by a well-crafted manual index, although it may complement them for example by providing free text search. The role of an index is to provide access via keywords and phrases which are not necessarily present in the text itself, but must be added by the skill of the indexer.

Using the xml:id and n Attributes

When traditional reference schemes represent a hierarchical structuring of the text which mirrors that of the marked-up document, the n attribute defined for all elements may be used to indicate the traditional identifier of the relevant structural units. The n attribute may also be used to record the numbering of sections or list items in the copy text if the copy-text numbering is important for some reason, for example because the numbers are out of sequence.

For example, a traditional reference to Ovid's might be Amores 2.10.7—book 2, poem 10, line 7. Book, poem, and line are structural units of the work and will therefore be tagged in any case. (See chapter for a discussion of structural units in verse collections.) In such cases, it is convenient to record traditional reference numbers of the structural units using the n attribute. The relevant tags for our example would be: ... ... ...

One may also place the entire standard reference for each portion of the text into the appropriate value for the n attribute, though for obvious reasons this takes more space in the file: ...

If the names used by the traditional reference system can be formulated as identifiers, then the references can be given as values for the xml:id attribute; this requires that the reference be given without internal spaces, begin with a letter or underscore, and contain no characters other than letters, digits, hyphens, underscores, full stops, and the various combining and extender characters, as defined by the XML specification. Unlike values for the n attribute, values for the xml:id attribute must be unique throughout the document. Our example then looks like this: ...

To document the usage and to allow automatic processing of these standard references, it is recommended that the TEI header be used to declare whether standard references are recorded in the n or xml:id attributes and which elements may carry standard references or portions of them. For examples of declarations for the reference systems just shown, see section .

Using the n attribute one can specify only a single standard referencing system, a limitation not without problems, since some editions may define structural units differently and thus create alternative reference systems. For example, another edition of the considers poem 10 a continuation of poem 9, and therefore would specify the same line as Amores 2.9.31. In order to record both of these reference systems one could employ any of a variety of methods discussed in chapter .

Creating New Reference Systems

If a text has no canonical reference system of its own, a new custom reference system may be used.

The global attributes n and xml:id may be used to assign reference identifiers to segments of the text. Identifiers specified by either attribute apply to the entire element for which they are given. xml:id attributes must be unique within a single document, and xml:id values must begin with a letter. No such restrictions are made on the values of n attributes.

Determining a referencing system for a TEI encoding depends on many factors that may either be derived from textual structure, or influenced by extra-textual contingencies such as project and file management concerns. It is important, therefore, that the attribute used, the elements which can bear standard reference identifiers, and the method for constructing standard reference identifiers, should all be declared in the header as described in section .

The Guidelines do not recommend one specific method for creating new referencing systems; however, the rest of this section lists some possibly useful strategies.

Milestone Elements

Where the desired reference system does not correspond to any particular structural hierarchy, or the document combines multiple structural hierarchies (as further discussed in ), simpler though less expressive methods may be necessary. In such cases the simplest solution may be just to mark up changes in the reference system where they occur, by using one or more of the following milestone elements:

These elements simply mark the points in a text at which some category in a reference system changes. They have no content but subdivide the text into regions, rather in the same way as milestones mark points along a road, thus implicitly dividing it into segments. The elements gb, pb, cb, and lb are specialized types of milestone, marking gathering, page, column, and line boundaries respectively. The global n attribute is used in each case to provide a value for the particular unit associated with this milestone (for example, the page or line number). Since it is not structural, validation of a reference system based on milestones cannot readily be checked by an XML parser, so it will be the responsibility of the encoder or the application software to ensure that they are given in the correct order.

Milestone elements are often used as a simple means of capturing the original appearance of an early printed text, which will rarely coincide exactly with structural units, but they are generally useful wherever a text has two or more competing structures. For example, many English novels were first published as serial works, individual parts of which do not always contain a whole number of chapters. An encoder might decide to represent the chapter-based structure using div1 elements, with milestone elements to mark the points at which individual parts end; or the reverse. Thus, an encoding in which chapters are regarded as more important than parts might encode some work in which chapter three begins in part one and is concluded in part two as follows:

An encoding of the same work in which parts are regarded as more important than chapters might begin as follows:

Similarly, when tagging dramatic verse one may wish to privilege stanzas and lines over speeches and speakers, particularly where speeches cross line and line group boundaries. One might also wish to mark changes in narrative voice in a prose text. In either case, a milestone tag may be used to indicate change of speaker: Oh what is this I cannot see With icy hands gets a hold on me Oh I am Death, none can excel I open the doors of heaven and hell

Milestone tags also make it possible to record the reference systems used in a number of different editions of the same work. The reference system of any one edition can be recreated from a text in which all are marked by simply ignoring all elements that do not specify that edition on their ed attribute.

As a simple example, assuming that edition E1 of some collection of poems regards the first two poems as constituting the first book, while edition E2 regards the first poem as prefatory, a markup scheme like the following might be adopted:

In this case no n value is specified, since the numbers rise predictably and the application can keep a count from the start of the document, if desired.

The value of the n attribute may but need not include the identifiers used for any larger sections. That is, either of the following styles is legitimate: or

When using milestone tags, line numbers may be supplied for every line or only periodically (every fifth, every tenth line). The latter may be simpler; the former is more reliable.

The style of numbering used in the values of n is unrestricted: for the example above, I.i, I.ii, and I.iii could have been used equally well if preferred. The special value unnumbered should be reserved for marking sections of text which fall outside the normal numbering system (e.g. chapter heads, poem numbers, titles, or speaker attributions in a verse drama).

By default, there are no constraints on the values supplied for the ed attribute. If it is felt appropriate to enforce such a restriction, the techniques described in may be used, for example to specify that the attribute must specify one of a predefined set of values.

See below, section , for examples of declarations for the reference systems just shown.

Milestone elements may be used to mark any kind of shift in the properties associated with a piece of text, whether or not would normally be considered a reference system. For example, they may be used to mark changes in narrative voice in a prose text, or changes of speaker in a dramatic text, where these are not marked using structural elements such as sp, perhaps in order to avoid a clash of hierarchies.

As noted in above, milestone elements such as lb or pb represent whitespace and are therefore by default assumed to occur between orthographic tokens in the text, where these are not otherwise indicated. By default it is reasonable to assume that words are not broken across page or line boundaries, and that therefore a sequence such as ...sed imperator dixit... should be tokenized as four words (sed, imp, erator, and dixit). The break attribute is provided to change the default assumption. To make explicit that imperator in the above example should be treated as a single word, a tagging such as the following is recommended: ...sed imperator dixit... Where hyphenation appears before a line or page break, the encoder may or may not choose to record the fact, either explicitly using an appropriate Unicode character, or descriptively for example by means of the rend attribute; see further .

Declaring Reference Systems

Whatever kind of reference system is used in an electronic text, it is recommended that the TEI header contain a description of its construction in the refsDecl element described in section . As described there, the declaration may consist either of a formal declaration using the cRefPattern element or an informal description in prose. The former is recommended because unlike prose it can be processed by software.

The three examples given in section would be declared as follows. The first example encodes the standard references for Ovid's one level at a time, using the n attribute on the div1, div2, div3, and l tags. The header section for such an encoding should look something like this:

A canonical reference is assembled with the name of the work: the n of a div1, a space, the number of the book: the n of a child div2, a full stop the number of the poem: the n of a child div3, the line number: the n value of a child l

Same as above, but without the last component (full stop followed by the l's n.

Same as above, but without the poem component (full stop followed by the div3's n.

The second example encodes the same reference system, again using the n attribute on the div1, div2, div3, and l tags, but giving the reference string in full on each tag. If canonical references are made only to lines, the reference system could be declared as follows: Since the entire regular expression is enclosed as a parenthetical subgroup, the entire canonical reference string is sought as the value of the n attribute on an l element.

In order to handle references to poems as well as to individual lines, the declaration for the reference system must be more complicated: This declaration indicates that the entire reference string must be sought as the value of the n attribute on a div1, div2, div3, or l element.

The third example encodes the same reference system, this time giving the entire reference string as the value of the xml:id attribute on the relevant tags. The reference system declaration for such an encoding could be: although in general there seems to be little advantage in this case: it is no more difficult to use a standard relative URI reference as the value of target.

Reference systems recorded by means of milestone tags can also be declared; the following prose description could be used to declare the example given in section .

Standard references to work, book, poem, and line may be constructed from the milestone tags in the text.

Or in this way, using a formal declaration for this reference scheme derived from edition E1.

Methods of Encoding Bibliographic References and Lists of References

The members of the model.biblLike class all share a number of possible component sub-elements. For the bibl and biblStruct elements, exactly the same sub-elements are concerned, and they are described together in section ; for the biblFull element, the sub-elements concerned are fully described in section .

Different levels of specific tagging may be appropriate in different situations. In some cases, it may be felt necessary to mark just the extent of the reference itself, with perhaps a few distinctions being made within it (for example, between the part of the reference which identifies a title or author and the rest). Such references, containing a mixture of text with specialized bibliographic elements, are regarded as bibl elements, and tagged accordingly. For example:

A book which had a great influence on him was Tufte's , although he may never have actually read it.

Indeed, some encoders may find it unnecessary to mark the bibliographic reference at all:

A book which had a great influence on him was Tufte's , although he may never have actually read it.

Some bibliographic references are extremely elliptical, often only a string of the form Baxter, 1983. If no further details of Baxter's book are given in the source text and none is supplied by the encoder, then the reference thus given should be tagged as a bibl: All of this is of course much more fully treated in Baxter, 1983. In general, however, normal modern bibliographic practice, and these Guidelines, distinguish between a bibliographic reference,referencesbibliographicbibliographic references which is a self-sufficient description of a bibliographic item, and a bibliographic pointer,bibliographic pointerspointersbibliographic which is a short-form citation (e.g. Baxter, 1983) which serves usually as a place-holder or pointer to a full long-form reference found elsewhere in the text. The usual encoding of short-form references such as Baxter, 1983 is not as bibl elements but as cross-references to such elements; see section below.

In cases where the encoder wishes to impose more structure on the bibliographic information, for example to make sure it conforms to a particular stylesheet or retrieval processor, the biblStruct element should be used. Note that several of the features in this and later examples are explained later in the current section. Edward R. Tufte 6506403994 http://id.loc.gov/authorites/names/n50012763.html Cheshire, Conn. Graphics Press

A more complex and detailed bibliographic structure is provided by the biblFull element defined in the TEI header module. This element is provided as a means of embedding the file description of one existing digital text within that of another (see further section ); however, its use is not confined to digital texts, and it may be used in the same way as any other bibliographic element, as in this example: Tufte, Edward R[olf] 126 pp. Graphics Press Cheshire, Conn. USA 1990

A list of bibliographic items, of whatever kind, may be treated in the same way as any other list (see section ). Alternatively, the specialized listBibl element may be used. The difference between the two is that a list contains item elements, within which bibliographic elements (bibl, biblStruct, or biblFull) may appear, as well as other phrase- and paragraph-level elements, whereas the listBibl may contain only bibliographic elements, optionally preceded by a heading and a series of introductory paragraphs. For most purposes, good practice would usually require that a listBibl contain only one kind of bibliographic element, though the following example combines both fully structured biblStruct and informal bibl elements: Bibliography Nelson Theodore Holm Simon H. Lavington North-Holland Amsterdam 1013–23 Apparently a draft of section 4 of . Ted Nelson: (privately published, 1987). Baxter, Glen: London: Thames and Hudson, 1988. This example also demonstrates the way that bibliographical markup of authors, titles, dates etc. can be handled differently in biblStructs and bibls. In the two bibl items, the key information is marked up, but it is presented in an order which makes it suitable for direct rendering, with the punctuation included.

The listBibl element is most appropriate for a more formal bibliography. The same bibl or biblStruct elements may however be embedded within an ordinary list, thus allowing them to be mixed with running prose or presented informally, as in the following version of the same example: Bibliography Nelson, T. H. . . Simon H. Lavington North-Holland: Amsterdam, 1980. pp 1013–23 Apparently a draft of section 4 of . Ted Nelson: (privately published, 1987) Baxter, Glen London: Thames and Hudson, 1988.

Components of Bibliographic References

This section discusses commonly occurring components of bibliographic references and elements used for encoding them. They fall into four groups: elements for grouping components of the analytic, monographic, and series levels in a structured bibliographic reference titles of various kinds, and statements of intellectual responsibility (authorship, etc.) information relating to the publication, pagination, etc. of an item (most of these constitute the default members of the model.biblPart class) annotation, commentary, and further detail The following sections describe the elements which may be used to represent such information within a bibl or biblStruct element. Within the former, elements from the model.biblPart class, other phrase-level elements, and plain text may be combined without other constraint; within the latter, such of these elements as exist for a given reference must be distinguished, and must also be presented in a specific order, discussed further below (section ).

Bibliographic Pointers

References which are pointers to bibliographic items, of whatever kind, should be treated in the same way as other cross-references (see section ). As discussed in that section, cross-referencing within TEI texts is in general represented by means of ptr or ref elements. A target attribute on these elements is used to supply an identifying value for the target of the cross-reference, which should be, in the case of bibliographic elements, a bibliographic reference of some kind. Where the form of the reference itself is unimportant, or may be reconstructed mechanically, or is not to be encoded, the ptr element is used, as in the following example: As shown above () ...

Where the form of the reference is important, or contains additional qualifying information which is to be kept but distinguished from the surrounding text, the ref element should be used, as in the following example: Nelson claims (ibid, passim) ... It may be important to distinguish between the short form of a bibliographic reference and some qualifying or additional information. The latter should not appear within the scope of the ref element when this is the case, as for example in an application concerned to normalize bibliographic references: Nelson claims (Nelson [1980] pages 13–37) ...

The ref element may also be used to provide a reference to a copy of the bibliographic item itself, particularly if this is available online, as in the following example: Suzana Sukovic http://www.dlib.org/dlib/january02/sukovic/01sukovic.html 8 1 2002

Relationship to Other Bibliographic Schemes

The bibliographic tagging defined here can capture the distinctions required by most bibliographic encoding systems; for the benefit of users of some commonly used systems, the following lists of equivalences are offered, showing the relationship of the markup defined here to the fields defined for bibliographic records in the Scribe, BibTeX, and ProCite systems.

Listed below are the equivalences between the various bibliographic fields defined for use in the Scribe and BibTeX systems of bibliographic databases and the elements defined in this module.The BibTeX scheme is intentionally compatible with that of Scribe, although it omits some fields used by Scribe. Hence only one list of fields is given here. Elements and structures available in the module defined here which have no analogues in Scribe and BibTeX are not noted. address tag as placeName or address annote tag as note author tag as author booktitle tag as title level="m" or title within monogr chapter tag as biblScope unit="chap" date used only to record date entry was made in the bibliographic database; not supported edition tag as edition editor tag as editor or respStmt editors tag as multiple editor or respStmt elements fullauthor use the reg element, possibly inside a choice element, inside either an author or name fullorganization use the reg element, possibly inside a choice element, inside a name type="org" howpublished tag as note, possibly using the form note place="inline" institution used only for issuer of technical reports; tag as publisher journal tag as title level="j" or title within monogr key used to specify an alternate sort key for the bibliographic item, for use instead of author's or editor's name; not supported meeting tag as meeting or as note month use date; if the date is not in a trivially parseable form, use the when attribute to provide a normalized equivalent in one of the format from note tag as note number tag as biblScope unit="issue" or biblScope unit="number"; for technical report numbers, use idno type="docno" organization used only for sponsor of conference; use name type="org" within respStmt within meeting element pages tag as biblScope unit="pp" publisher tag as publisher school used only for institutions at which thesis work is done; tag as publisher series tag as title level="s" or title within series title tag as title in appropriate context or with appropriate level value volume tag as biblScope unit="vol" year tag as date; if the date is not in a trivially parseable form, use the when attribute to provide an ISO-format equivalent

Core Tags for Verse

Like other written texts, verse texts or poems may be hierarchically subdivided, for example into books or cantos. These structural subdivisions should be encoded using the general purpose div or div1 (etc.) elements described below in chapters and . The fundamental unit of a verse text is the verse line rather than the paragraph, however.

The l element is used to mark up verse lines, that is metrical rather than typographic lines. In some modern or free verse, it may be hard to decide whether the typographic line is to be regarded as a verse line or not, but the distinction is quite clear for verse following regular metrical patterns. Where a metrical line is interrupted by a typographic line break, the encoder may choose to ignore the fact entirely or to use the empty lb (line break) element discussed in . By convention, the start of a metrical line implies the start of a typographic line; hence there is no need to introduce an lb tag at the start of every l element, but only at places where a new typographic line starts within a metrical line, as in the following example: Of Mans First Disobedience, and the Fruit Of that Forbidden Tree, whose mortal tast Brought Death into the World, and all our woe, With loss of Eden, till one greater Man Restore us, and regain the blissful Seat... In the original copy text, the presence of an ornamental capital at the start of the poem means that the measure is not wide enough to print the first four lines on four lines; instead each metrical line occupies two typographic lines, with a break at the point indicated. Note that this encoding makes no attempt to preserve information about the whitespace or indentation associated with either kind of line; if regarded as essential, this information would be recorded using the rend or rendition attributes discussed in .

The l element should not be used to represent typographic lines in non-verse materials: if the line-breaking points in a prose text are considered important for analysis, they should be marked with the lb element. Alternatively, a neutral segmentation element such as seg or ab may be used; see further discussion of these elements in chapter . The l element is a member of the model.lLike class, which is a subclass of the model.divPart class, along with elements from the model.pLike (paragraph-like) class.

In some verse forms, regular groupings of lines are regarded as units of some kind, often identified by a regular verse scheme. In stichic verse and couplets, groups of lines analogous to paragraphs are often indicated by indentation. In other verse forms, lines are grouped into irregular sequences indicated simply by whitespace. The lg or line group element may be used to mark any such grouping of elements from the model.lLike class. As a member of the att.typed class, the lg element bears the following attributes: which may be used to further categorize the line group where this is felt desirable, as in the following example. This example also demonstrates the rend attribute to indicate whether or not a line is indented. Come fill up the Glass, Round, round let it pass, 'Till our Reason be lost in our Wine: Leave Conscience's Rules To Women and Fools, This only can make us divine. Then a Mohock, a Mohock I'll be, No Laws shall restrain Our Libertine Reign, We'll riot, drink on, and be free.

For some kinds of analysis, it may be useful to identify different kinds of line group within the same piece of verse. Such line groups may self-nest, in much the same way as the un-numbered div element described in chapter . For example: Thus speaks the Muse, and bends her brow severe:— “Did I, Lætitia, lend my choicest lays, And crown thy youthful head with freshest bays, That all the' expectance of thy full-grown year Should lie inert and fruitless? O revere Those sacred gifts whose meed is deathless praise, Whose potent charms the' enraptured soul can raise Far from the vapours of this earthly sphere! Seize, seize the lyre! resume the lofty strain! 'T is time, 't is time! hark how the nations round With jocund notes of liberty resound,— And thy own Corsica has burst her chain! O let the song to Britain's shores rebound, Where Freedom's once-loved voice is heard, alas! in vain.”

It is often the case that verse line boundaries conflict with the boundaries of other structural elements. In the following example, the single verse line A Workeman in't... welcome is interrupted by a stage direction: Thou fumblest Eros, and my Queenes a Squire More tight at this, then thou: Dispatch. O Loue, That thou couldst see my Warres to day, and knew'st The Royall Occupation, thou should'st see A Workeman in't. Enter an Armed Soldier. Good morrow to thee, welcome. In this encoding, the part attribute is used, as with div, to indicate that the last two l elements should be regarded as the initial and final parts of a single line, rather than as two lines.

The same technique may be used where verse lines are collected together into units such as verse paragraphs: Unprofitably travelling toward the grave, Like a false steward who hath much received And renders nothing back. Was it for this That one, the fairest of all rivers, loved To blend his murmurs with my nurse's song,

The part attribute may also be attached to an lg element to indicate that it is incomplete, for example because it forms part of a group that is divided between two speakers, as in the following example: First Voice But why drives on that ship so fast Withouten wave or wind? Second Voice The air is cut away before, And closes from behind.

For alternative methods of aligning groups of lines which do not form simple hierarchic groups, or which are discontinuous, see the more detailed discussion in chapter . For discussion of other elements and attributes specific to the encoding of verse, see chapter .

Core Tags for Drama

Like other written texts, dramatic and other performance texts such as cinema or TV scripts are often hierarchically organized, for example into acts and scenes. These structural subdivisions should be encoded using the general purpose div or div1 (etc.) elements described below in chapters and . Within these divisions, the body of a performance text typically consists of speeches, often prefixed by a phrase indicating who is speaking, and occasionally interspersed with stage directions of various kinds.

In the following simple example, each speech consists of a single paragraph: Scene 2. Peachum, Filch. FILCH.

Sir, Black Moll hath sent word her Trial comes on in the Afternoon, and she hopes you will order Matters so as to bring her off.

PEACHUM.

Why, she may plead her Belly at worst; to my Knowledge she hath taken care of that Security. But, as the Wench is very active and industrious, you may satisfy her that I'll soften the Evidence.

FILCH.

Tom Gagg, sir, is found guilty.

In the following example, each speech consists of a sequence of verse lines, some of them being marked as metrically incomplete: ACT I SCENE I Enter Barnardo and Francisco, two Sentinels, at several doors Barn Who's there? Fran Nay, answer me. Stand and unfold yourself. Barn Long live the King! Fran Barnardo? Barn He. Fran You come most carefully upon your hour. Barn 'Tis now struck twelve. Get thee to bed, Francisco. Fran For this relief much thanks. 'Tis bitter cold, And I am sick at heart.

In some cases, as here in the First Quarto of , the printed speaker attributions need to be supplemented by use of the who attribute; again, the lines are marked as complete or incomplete: Enter two Centinels. Now call'd Bernardo & Francesco. 1. Stand: who is that? 2. Tis I. 1. O you come most carefully vpon your watch, 2. And if you meete Marcellus and Horatio, The partners of my watch, bid them make haste. 1. I will: See who goes there. Enter Horatio and Marcellus.

By contrast with the preceding examples, the following encodes an early printed edition without making any assumption about which parts are prose or verse: Actus primus, Scena prima. A tempestuous noise of Thunder and Lightning heard: Enter a Ship-master, and a Boteswaine. Master.

Bote-swaine.

Botes.

Heere Master: What cheere?

Mast.

Good: Speake to th' Mariners: fall too't, yarely, or we run our selues a ground, bestirre, bestirre. Exit.

Enter Mariners. Botes.

Heigh my hearts, cheerely, cheerely my harts: yare, yare: Take in the toppe-sale: Tend to th' Masters whistle: Blow till thou burst thy winde, if roome e-nough.

The sp and stage elements should also be used to mark parts of a text otherwise in prose which are presented as if they were dialogue in a play. The following example is taken from a 19th century novel in which passages of narrative and passages of dialogue are mixed within the same chapter: The reverend Doctor Opimian

I do not think I have named a single unpresentable fish.

Mr Gryll

Bream, Doctor: there is not much to be said for bream.

The Reverend Doctor Opimian

On the contrary, sir, I think there is much to be said for him. In the first place ...

Fish, Miss Gryll — I could discourse to you on fish by the hour: but for the present I will forbear ...

Lord Curryfin (after a pause).

Mass as the second grave-digger says in , I cannot tell.

A chorus of laughter dissolved the sitting.