This document describes a bare-bones tag set taken from the published in 1994 by the Text Encoding Initiative (TEI). The tags described have been chosen to serve as a simple introduction to the full markup scheme described in the Guidelines; they may suffice in some cases for the creation of simple electronic texts, but serious work will require a larger selection of the TEI tags. The reader is encouraged to use this document as a first introduction to TEI tagging, and to progress, after reading this document and using its tag set for a while, to a study of other TEI documentation, either the document called (document number TEI U5), or the full text of the Guidelines themselves (, document number TEI P3).

This document introduces the tags informally, with examples. As an incentive to learn the full TEI tag set, it mentions, from time to time, tags which are in the full tag set but have been omitted here to keep the bare-bones tag set simple. Such references may be ignored on first reading. Fuller discussion of all tags, and their formal descriptions in terms of the Standard Generalized Markup Language (SGML) may be found in the Guidelines.

SGML, the Standard Generalized Markup Language, is a formal language for representing text in electronic form. The TEI tag set is defined in terms of SGML, and all TEI-conformant documents must also conform to SGML.

In SGML-based encoding schemes, a document is represented by a combination of content (roughly speaking, the characters of the text, what you see on a printed page when the text is printed out) and markup (roughly speaking, information about the structure of the text, or features important for proper processing of the text, such as its division into chapters and sections, or the fact that a given phrase is a technical term and must be italicized). Non-SGML software, such as proprietary word processors, uses a similar division into content and markup. In sophisticated software, markup is usually invisible to the user unless you use a reveal-codes function or the like, to make it visible. SGML differs from proprietary markup systems in several ways: SGML is non-proprietary, and fully documented independent of any SGML software, so you can move your documents from one program to another at any time without losing any information. In contrast, conversion among proprietary systems is notoriously difficult and error prone. All systems use the markup to decide how to process the text, but in SGML systems the markup is typically defined in abstract terms, rather than directly in processing terms. An SGML system is more likely to mark a phrase as a technical term than simply as italic: a separate style sheet is normally used to determine whether technical terms are italicized, bolded on first reference, or displayed in blue, and to specify further processing such as adding an entry for the term to a glossary of technical terms. Proprietary systems nowadays often have style sheet mechanisms, too, but these seldom match SGML software in convenience or power. SGML-based markup schemes, called document type definitions (DTDs), are typically accompanied by system-independent documentation of the markup. The document you are reading provides precisely this type of system-independent documentation — i.e., since you can use these tags with any SGML software, this paper cannot describe what will happen on the screen or what keys you must press, with the particular program you choose to use. Such matters will vary from program to program, and you should consult the software documentation for help. Unlike proprietary systems, SGML systems invariably provide facilities for exporting documents in standard (SGML) form, so they can easily be used by other software. Some enthusiasts phrase this as a sort of slogan: With SGML, you own your documents; without SGML, your documents are owned by people in Orem, Utah, or in Redmond, Washington. Which do you prefer? There are other differences, but these will do for now.

SGML markup takes three forms: declarations, entity references and tags. I cannot tell a lie: actually, there are four forms of markup. The fourth, processing instructions, won't concern us here.

Declarations are used to define the tags and entity references which are legal in a document type. Since the tags and entities we are concerned with here have all already been defined by the TEI, there is no need to discuss declarations further in this document. You will need to learn about them if you want to customize the TEI tag sets, but that won't be covered here. The only form of declaration you need to know about, to follow the examples below, is the comment, which is preceded by &mdo;-- and followed by -->:

]]>

Entities are named portions of documents, which may be stored separately; entity references show where each entity goes. Among other things, entity references are used to embed special characters in the text when, as often happens, the characters in question are not available on the keyboard. Some entities for special characters are defined in international standards. For example, the entity eacute names the character e with an acute accent (é). When the standard entity sets are in use, the following two examples are identical in meaning:

(In case this has been corrupted in transmission, or is being rendered on a device without accented characters, that second one is the same as the first, except that the reference to the entity eacute in characters 3-10 of the first example has been replaced with a real e with an acute accent in the system's native character set in character 3 of the second example.)

Entities are also used to handle graphics and other material in non-SGML notations, and to divide a document up into sections stored in separate files for purposes of simpler maintenance, but we won't discuss such uses here.

Tags mark the beginning and ending of parts of the document; the parts themselves are called elements. Normally, tags are marked in the document by angle brackets; end-tags have a slash after the opening angle bracket. In the following example, the sentence is marked as a quotation by the start-tag and end-tag which surround it; quote is an element type defined by the TEI.

L'étag, c'est moi.]]>

Elements always have a basic type (in the example above, it is quote); they may also have other attributes, which are indicated by special notations inside the start-tag for the element. For example, the TEI defines the attribute lang as applying to every type of element; its value indicates the language of the element's content, using standard two- or three-letter abbreviations (e.g. fra for French).

L'état, c'est moi.]]>

Some attributes may be restricted to certain types of values. Attributes of type id, for example, must provide a unique name or identifier for the element on which they appear; this identifier can be referred to by other attributes, of type idref (id reference). The TEI defines a global attribute named id, of type id, for use in cross-references and other kinds of hypertext links. (TEI attributes are called global when they apply to every type of element.)

Finally, it should be noted that SGML allows some tags to be omitted from documents, in cases when they are logically redundant and their location can be inferred from that of other tags; in the examples given here, we will not exploit this facility, but always give all tags explicitly. Tag omission is generally of interest only to those working without an SGML editor.

In sum: in SGML, everything is delimited. Elements are delimited by start- and end-tags. Start- and end-tags are delimited by angle brackets. Attribute values are delimited by single or double quotes. Entity references are delimited by ampersand and semicolon.

That's all there is to it. If you understand the rules just described, you should have no trouble understanding all the SGML examples in this document.

A TEI-conformant electronic text consists of the text itself (transcribed from some source, or created in electronic form), preceded by a TEI header, which identifies the electronic text and can also document the encoding practices used in creating it. The entire thing is enclosed within a tei.2 element, and preceded by an SGML declaration identifying the document type to be used in validating the document.

The SGML declaration won't be described here. Further below, I'll discuss the TEI header, and the specialized tags for front matter and back matter of the main text. In work with electronic text, however, the vast majority of one's time is spent within the body of the text itself, and so I begin with a description of tags for basic text encoding: paragraphs and other paragraph-like things, character- or phrase-level elements which occur within paragraphs, and so on.

The div and head used in the example just given elements are described below.

If the page breaks of the source are of interest, as they generally are for material transcribed from existing printed editions, record them using the pb element. This element is empty: that is, it has neither content nor an end-tag. It does not mark a passage or portion of the text, just a location within the text. The attribute n, defined for all TEI elements, should be used to indicate the page number; if page numbers from more than one edition are transcribed, the attribute ed should be used to distinguish the two paginations:

I ask the wholehearted cooperation of all our citizens to the end that this return of individual freedom shall not be accompanied by the repugnant conditions that obtained prior to the adoption of the 18th Amendment and those that have existed since its adoption....

Individual verse lines should be tagged with l (that's an L), stanzas or other verse structures above the level of the line should be tagged lg (line group); the latter's type attribute may optionally be used to identify the formal structure in question, for retrieval or other purposes:

Awake! for Morning in the Bowl of Night Has flung the Stone that puts the Stars to Flight: And Lo! the Hunter of the East has caught The Sultan's Turret in a Noose of Light. ]]>

When the indentation of the lines is significant, it can be recorded using the global rend attribute, with some suitable value:

And Lo! the Hunter of the East has caught The Sultan's Turret in a Noose of Light.]]>

Of course, if the verse is quoted from another text, the l elements should be enclosed in a q element.

Drama should be encoded with the elements sp (speech) and stage (stage direction). Stage directions can occur either within speeches or between them. As may be seen in the example below, the speaker may be indicated with the who attribute on the sp element:

Speak, hands, for me! They stab Caesar. Et tu, Brute? -- then fall, Caesar! Dies.]]>

When the precise form of the speaker atribution in the source is important, the speaker may be identified by a separate speaker element at the beginning of the sp element.

Cas. Speak, hands, for me! They stab Caesar. Caes. Et tu, Brute? -- then fall, Caesar! Dies.]]>

These tags may also be used for material not written as drama, but presented using dramatic conventions (e.g. transcriptions of speeches, or of press conferences):

[Applause.] and that Governments of the people, by the people, and for the people, shall not perish from the earth. [Long-continued applause.] ]]>

As with verse, if the drama is quoted from another text, it should be enclosed in a q element.

Bibliographic references should normally be enclosed in bibl elements; within such elements, or outside them, title may be used to mark titles of articles, books, journals, etc. Its level attribute takes the values A, M, J, S, or U to show whether the title is an analytic (article) title, a monogrphic (book) title, the title of a journal, that of a series, or that of unpublished material such as a thesis. For example a reference to: in , vol. II (New York: Random House, 1938), pp. 510-514 would be encoded thus:

, in , vol. II (New York: Random House, 1938), pp. 11-16. ]]>

If material has been omitted from an electronic text (e.g. because it is illegible or not of interest to the expected users, the omission should normally be indicated using a gap element at the point of omission. The attributes desc, reason, and extent may optionally be used to describe what was omitted, to explain why, and to give an approximate size for it. For example:

Suppose I see two individuals approaching whose rank I wish to ascertain. They are, we will suppose, a Merchant and a Physician, or in other words, an Equilateral Triangle and a Pentagon: how am I to distinguish them?

It will be obvious ...

Notes in the text, whether footnotes, endnotes, or inline block notes, should be tagged with the note element. The location may be given, if desired, in the place attribute. Authorial notes may be distinguished from editorial notes by means of the resp attribute, which indicates who is responsible for the note. For example:

IN WITNESS WHEREOF, I have hereunto set my hand and caused the seal of the United States to be affixed.

Footnotes and endnotes should normally be transcribed at their point of attachment. Their number may optionally be given in the n attribute:

Philadelphia has our poor attempts and Chicago has our poor power. to add or detract.]]>

Lists should be tagged using the list and item elements; a heading or title for the list should be tagged as a head. Lists may be distinguished as ordered (numbered), unordered (bulleted), etc., by means of the type attribute. For example:

the close of the first fiscal year ending June 30 of any year after the year 1933, in which ..., or the repeal of the eighteenth amendment to the Constitution, whichever is the earlier.]]>

The full TEI scheme also defines a label element for use as an alternative to using the n attribute to give item numbers or labels.

Notes in the preceding sections have mentioned some of the elements defined in the full TEI scheme's core tag set but omitted from this bare-bones version. In addition to those already mentioned, tags omitted here include those for proper nouns and other references to people and places, addresses, numbers, units of measure and measured quantities, dates, and times of day.

The full scheme also defines optional tag sets for hypertext linking, analysis or interpretation (including both literary and linguistic analysis) of the text, manuscript transcription, text-critical apparatus, tables, figures, and other specialized interests.

In cases where text divisions have no headings, or have only headings consisting of their type value and a number, no heading need be given, as shown above. If desired, however, the heading may be given explicitly:

Section 1.

The eighteenth article of amendment to the Constitution of the United States is hereby repealed.

The transportation ...

This article shall be inoperative unless ...

The TEI encoding scheme defines specialized tags for transcribing title pages, in order to ensure that processing software can easily locate and identify the author, title, and date of the document as given on its title page. The title page itself, and its major component parts, are illustrated in this example:

The Public Papers and Addresses of Franklin D. Roosevelt With a special introduction and explanatory notes by President Roosevelt Volume Two The Year of Crisis 1933 Random House New York 1938 ]]>

The titlePart element is used both for the different parts of the document title (as shown) and also for miscellaneous parts of the title page which are neither document title, nor document author, nor imprint information.

The TEI header allows later users of the etexts you create to find out what the text is, who created the etext (i.e. you), and what source edition(s) you transcribed the etext from. In its full expansion, it also allows a full accounting of your transcription practice (did you correct typos silently? did you expand abbreviations? normalize spelling? etc.) and can also include a detailed characterization of the text itself (demographics of its author and audience, subject matter, genre, etc.) and a full change log, which is important for document management in large projects.

For bare-bones work, however, it's simplest to copy the following TEI header by rote, and replace the text in square brackets with appropriate information about the text being encoded. If the etext is not transcribed from a pre-existing source, but instead is being created in electronic form, the bibl tags within the sourceDesc element should be changed to p.

[Indicate who is publishing this electronic text (i.e. you).]

For example, the TEI header of the document you are reading looks like this:

Published electronically by the Text Encoding Initiative, Chicago and Oxford, in 1994.

A TEI-encoded electronic text is always encoded as a tei.2 element, which in turn contains a teiHeader element followed by a text element. The overall structure is thus:

]]>

The start-tag of the tei.2 element is preceded by an explicit reference to the external file containing the document type definition to be applied to the text by the SGML parser. The stripped-down DTD described here may be invoked with the following document-type declaration:

]]>

In some systems, the association of a document with a given document type is handled internally, and no such explicit declaration is visible until the document is exported from the system. In such systems, the user will be asked to select a rules or logic file when the document is first created or imported into the editor.