Oxford Advanced Learner's Dictionary

Dictionaries

This chapter defines a module for encoding lexical resources of all kinds, in particular human-oriented monolingual and multilingual dictionaries, glossaries, and similar documents. The elements described here may also be useful in the encoding of computational lexica and similar resources intended for use by language-processing software; they may also be used to provide a rich encoding for wordlists, lexica, glossaries, etc. included within other documents. Dictionaries are most familiar in their printed form; however, increasing numbers of dictionaries exist also in electronic forms which are independent of any particular printed form, but from which various displays can be produced.

Both typographically and structurally, print dictionaries are extremely complex. Such lexical resources are moreover of interest to many communities with different and sometimes conflicting goals. As a result, many general problems of text encoding are particularly pronounced here, and more compromises and alternatives within the encoding scheme may be required in the future.We refer the reader to previous and current discussions of a common format for encoding lexical resources. For example, ; ;; ; ; ; ; and ; . Two problems are particularly prominent.

First, because the structure of dictionary entries varies widely both among and within dictionaries, the simplest way for an encoding scheme to accommodate the entire range of structures actually encountered is to allow virtually any element to appear virtually anywhere in a dictionary entry. It is clear, however, that strong and consistent structural principles do govern the vast majority of conventional dictionaries, as well as many or most entries even in more exotic dictionaries; encoding guidelines should include these structural principles. We therefore define two distinct elements for dictionary entries, one (entry) which captures the regularities of many conventional dictionary entries, and a second (entryFree) which uses the same elements, but allows them to combine much more freely. It is however recommended that entry be used in preference to entryFree wherever possible. These elements and their contents are described in sections , , and .

Second, since so much of the information in printed dictionaries is implicit or highly compressed, their encoding requires clear thought about whether it is to capture the precise typographic form of the source text or the underlying structure of the information it presents. Since both of these views of the dictionary may be of interest, it proves necessary to develop methods of recording both, and of recording the interrelationship between them as well. Users interested mainly in the printed format of the dictionary will require an encoding to be faithful to an original printed version. However, other users will be interested primarily in capturing the lexical information in a dictionary in a form suitable for further processing, which may demand the expansion or rearrangement of the information contained in the printed form. Further, some users wish to encode both of these views of the data, and retain the links between related elements of the two encodings. Problems of recording these two different views of dictionary data are discussed in section , together with mechanisms for retaining both views when this is desired.

To deal with this complexity, and in particular to account for the wide variety of linguistic contexts within which a dictionary may be designed, it can be necessary to customize or change the schema by providing more restriction or possibly alternate content models for the elements defined in this chapter. Section illustrates this with the provision of a closed set of values for grammatical descriptors.

This chapter contains a large number of examples taken from existing print dictionaries; in each case, the original source is identified. In presenting such examples, we have tried to retain the original typographic appearance of the example as well as presenting a suggested encoding for it. Where this has not been possible (for example in the display of pronunciation) we have adopted the transliteration found in the electronic edition of the Oxford Advanced Learner's Dictionary. Also, the middle dot in quoted entries is rendered with a full stop, while within the sample transcriptions hyphenation and syllabification points are indicated by a vertical bar |, regardless of their appearance in the source text.

Dictionary Body and Overall Structure

Overall, dictionaries have the same structure of front matter, body, and back matter familiar from other texts. In addition, this module defines entry, entryFree, and superEntry as component-level elements which can occur directly within a text division or the text body.

The following tags can therefore be used to mark the gross structure of a printed dictionary; the dictionary-specific tags are discussed further in the following section.

As members of the classes att.entryLike and att.sortable, entry and entryFree share the following attributes:

The front and back matter of a dictionary may well contain specialized material such as lists of common and proper nouns, grammatical tables, gazetteers, a guide to the use of the dictionary, etc. These should be tagged using elements defined elsewhere in these Guidelines, chiefly in the core module (chapter ) together with the specialized dictionary elements defined in this chapter.

The body element consists of a set of entries, optionally grouped into one or several div elements. These text divisions might, for example, correspond to sections for different letters of the alphabet, or to sections for different languages in a bilingual dictionary, as in the following example:

English-French

French-English

In a print dictionary, the entries are typically typographically distinct entities, each headed by some morphological form of the lexical item described (the headword), and sorted in alphabetical order or (especially for non-alphabetic scripts) in some other conventional sequence. Dictionary entries should be encoded as distinct successive items, each marked as an entry or entryFree element. The type attribute may be used to distinguish different types of entries, for example main entries, related entries, run-on entries, or entries for cross-references, etc.

Some dictionaries provide distinct entries for homographs, on the basis of etymology, part-of-speech, or both, and typically provide a numeric superscript on the headword identifying the homograph number. In these cases each homograph should be encoded as a separate entry; the superEntry element may optionally be used to group such successive homograph entries. In addition to a series of entry elements, the superEntry may contain a preliminary form group (see section ) when information about hyphenation, pronunciation, etc., is given only once for two or more homograph entries. If the homograph number is to be recorded, the global attribute n may be used for this purpose. In some dictionaries, homographs are treated in distinct parts of the same entry; in these cases, they may be separated by use of the hom element, for which see section .

A sort key, given in the sortKey attribute, is often required for superentries and entries, especially in cases where the order of entries does not follow the local character-set collating sequence (as, for example, when an entry for 3D appears at the place where three-D would appear).

A dictionary with no internal divisions might thus have a structure like the following; a superEntry is shown grouping two homograph entries.

The Structure of Dictionary Entries

A simple dictionary entry may contain information about the form of the word treated, its grammatical characterization, its definition, synonyms, or translation equivalents, its etymology, cross-references to other entries, usage information, and examples. These we refer to as the constituent parts or constituents of the entry; some dictionary constituents possess no internal structure, while others are most naturally viewed as groups of smaller elements, which may be marked in their own right. In some styles of markup, tags will be applied only to the low-level items, leaving the constituent groups which contain them untagged. We distinguish the class of top-level constituents of dictionary entries, which can occur directly within the entry element, from the class of phrase-level constituents, which can normally occur only within top-level constituents. The top-level constituents of dictionary entries are described in section , and documented more fully, together with their phrase-level sub-constituents, in section .

In addition, however, dictionary entries often have a complex hierarchical structure. For example, an entry may consist of two or more sub-parts, each corresponding to information for a different part-of-speech homograph of the headword. The entry (or part-of-speech homographs, if the entry is split this way) may also consist of senses, each of which may in turn be composed of two or more sub-senses, etc. Each sub-part, homograph entry, sense, or sub-sense we call a level; at any level in an entry, any or all of the constituent parts of dictionary entries may appear. The hierarchical levels of dictionary entries are documented in section .

Hierarchical Levels

The outermost structural level of an entry is marked with the elements entry or entryFree. The hom element marks the subdivision of entries into homographs differing in their part-of-speech. The sense element marks the subdivision of entries and part-of-speech homographs into senses; this element nests recursively in order to provide for a hierarchy of sub-senses of any depth. It is recommended to use the sense element even for an entry that has only one sense to group together all parts of the definition relating to the word sense since this leads to more consistent encoding across entries. All of these levels may each contain any of the constituent parts of an entry. A special case of hierarchical structure is represented by the re (related entry) element, which is discussed in section . Finally, the element dictScrap may be used at any point in the hierarchy to delimit parts of the dictionary entry which are structurally anomalous, as further discussed in section .

For example, an entry with two senses will have the following structure:

An entry with two homographs, the first with two senses and the second with three (one of which has two sub-senses), may have a structure like this: In some dictionaries, homographs have separate entries; in such a case, as noted in section , the two homographs may be treated as entries, optionally grouped in a superEntry:

The hierarchic structure of a dictionary entry is enforced by the structures defined in this module. The content model for entry specifies that entries do not nest, that homographs nest within entries, and that senses nest within entries, homographs, or senses, and may be nested to any depth to reflect the embedding of sub-senses. Any of the top-level constituents (def, usg, form, etc.) can appear at any level (i.e., within entries, homographs, or senses).

Groups and Constituents

As noted above, dictionary entries, and subordinate levels within dictionary entries, may comprise several constituent parts, each providing a different type of information about the word treated. The top-level constituents of dictionary entries are: information about the form of the word treated (orthography, pronunciation, hyphenation, etc.) grammatical information (part of speech, grammatical sub-categorization, etc.) definitions or translations into another language etymology examples usage information cross-references to other entries notes entries (often of reduced form) for related words, typically called related entries Any of the hierarchical levels (entry, entryFree, hom, and sense) may contain any of these top-level constituents, since information about word form, particular grammatical information, special pronunciation, usage information, etc., may apply to an entire entry, or to only one homograph, or only to a particular sense. The examples below illustrate this point.

The following elements are used to encode these top-level constituents:

In a simple entry with no internal hierarchy, all top-level constituents can appear as children of entry. com.peti.tor /k@m"petit@(r)/ n person who competes. OALD

n person who competes. For the elements which appear within the form and gramGrp elements of this and other examples, see below, section , and section .

Any top-level constituent can appear at any level when the hierarchical structure of the entry is more complex. The most obvious examples are def and cit, which appear at the sense level when several senses or translations exist: disproof (dIs"pru:f) n. 1. facts that disprove something. 2. the act of disproving. CED

n facts that disprove something. the act of disproving.

For ease of processing of such entries containing multiple senses along with those containing only a single sense, it is recommended to use sense in all entries to wrap those elements relating to a particular word sense.

In the following example, gramGrp is used to distinguish two homographs: bray /breI/ n cry of an ass; sound of a trumpet. ∙ vt [VP2A] make a cry or sound of this kind. OALD

n cry of an ass; sound of a trumpet. vt VP2A make a cry or sound of this kind.

Information of the same kind can appear at different levels within the same entry; here, grammatical information occurs both at entry and homograph level. ca.reen /k@"ri:n/ vt,vi 1 [VP6A] turn (a ship) on one side for cleaning, repairing, etc. 2 [VP6A, 2A] (cause to) tilt, lean over to one side. OALD

vt vi VP6A turn (a ship) on one side for cleaning, repairing, etc. VP6A VP2A (cause to) tilt, lean over to one side.

Alone among the constituent groups, form can appear at the superEntry level as well as at the entry, hom, and sense levels: a.ban.don 1/@"band@n/ v [T1] 1 to leave completely and for ever; desert: The sailors abandoned the burning ship. 2 …abandon 2 n [U] the state when one's feelings and actions are uncontrolled; freedom from control...LDOCE

v T1 to leave completely and for ever … n U the state when one's feelings and actions are uncontrolled; freedom from control…

Top-level Constituents of Entries

This section describes the top-level constituents of dictionary entries, together with the phrase-level constituents peculiar to each. the form element, which groups orthographic information and pronunciations, is described in section the gramGrp element, which groups elements for the grammatical characterization of the headword, is described in section the def element, which describes the meaning of the headword, is described in section the etym element and its special phrase-level elements are documented in section the cit element and its specific applications are described in section and section the usg, lbl, xr, and note elements are described in section the re element, which marks nested entries for related words, is described in section

Information on Written and Spoken Forms

Dictionary entries most often begin with information about the form of the word to which the entry applies. Typically, the orthographic form of the word, sometimes marked for syllabification or hyphenation, is the first item in an entry. Other information about the word, including variant or alternate forms, inflected forms, pronunciation, etc., is also often given.

The following elements should be used to encode this information: the form element groups one or more occurrences of any of them; it can also be recursively nested to reflect more complex sub-grouping of information about word form(s), as shown in the examples.

In addition to those listed above, the following elements, which encode morphological details of the form, may also occur within form elements: Of these, the gram element is most general, and all of the others are synonymous with a gram element with appropriate values (gen, number, case, etc.) for the type attribute.

The use of these elements as children of form is deprecated; instead, they should always be children of a gramGrp within form when describing that particular form of the word.

Different dictionaries use different means to mark hyphenation, syllabification, and stress, and they often use some unusual glyphs (e.g., the middle dot for hyphenation). All of these glyphs are in the Unicode character set, as discussed in . When transcribing representations of pronunciation the International Phonetic Alphabet should be used. It may be convenient (as has been done in the text of this chapter) to use a simple transliteration scheme for this; such a scheme should however be properly documented in the header.

In the simplest case, nothing is given but the orthography:

Often, however, pronunciation is given. soucoupe [sukup] … DNT

For a variety of reasons including ease of processing, it may be desired to split into separate elements information which is collapsed into a single element in the source text; orthography and hyphenation may for example be transcribed as separate elements, although given together in the source text. For a discussion of the issues involved, and of methods for retaining both the presentation form and the interpreted form, see section .

This example splits orthography and hyphenation, and adds syllabification because it differs from hyphenation: ar.ea … W7

Multiple orthographic forms may be given, e.g. to illustrate a word's inflectional pattern: brag … vb. brags, bragging, bragged … CED

vb Or the inflectional pattern may be indicated by reference to a table of paradigms, as here: horrifier [ORifje] (7) vt … [C/R]

Explanatory labels may be attached to alternate forms: MTBF abbrev. for mean time between failures. CED

When multiple orthographic forms are given, a pronunciation may be associated with all of them, as here: biryani or biriani (%bIrI"A:nI) … CED

In other cases, different pronunciations are provided for different orthographic forms; here, the form element is repeated to associate the first orthographic form explicitly with the first pronunciation, and the second orthographic form with the second pronunciation: mackle ("mak^@l) or macule ("makju:l) … CED

Recursive nesting of the form element can preserve relations among elements that are implicit in the text. For example, in the CED entry for hospitaller, it is clear that U.S. is associated only with hospitaler, but that the pronunciation applies to both forms. The following encoding preserves these relations: hospitaller or U.S. hospitaler ("hQspIt@l@) … CED

"hQspIt@l@

Grammatical Information

The gramGrp element groups grammatical information, such as part of speech, subcategorization information (e.g., syntactic patterns for verbs, count/mass distinctions for nouns), etc. It can contain any of the morphological elements defined in section for form and can appear as a child of entry, form, sense, cit, or any other element containing content about which there is grammatical information. For example, in the entry pinna ('pIn@) n., pl. -nae (-ni:) or -nas CED, the word defined can be either singular or plural; the pl. specification applies only to the inflected forms provided. Compare this with pants (paents) pl. n., where pl. applies to the headword itself.

As noted above in section , the elements for morphological information are simply shorthand for the general purpose gram element. Consider this entry for the French word médire: médire v.t. ind. (de) … PLC This entry can be tagged using specialized grammatical elements:

v t ind de Or using the gram element: v t ind de

Like form, gramGrp can be repeated, recursively nested, or used at the sense level to show relations among elements. isotope adj. et n. m. … DNT

adj n m wits (wIts) pl. n. 1. (sometimes sing.) the ability to reason and act, esp. quickly … CED pl n sometimes sing. the ability to reason and act, esp. quickly …

Sense Information

Dictionaries may describe the meanings of words in a wide variety of different ways—by means of synonyms, paraphrases, translations into other languages, formal definitions in various highly stylized forms, etc. No attempt is made here to distinguish all the different forms which sense information may take; all of them may be tagged using the def element described in section .

As a special case it is frequently desirable to distinguish the provision of translation equivalents in other languages from other forms of sense information; the use of cit type="translation" (which groups a translation equivalent with related information such as its grammatical description) for this purpose is described in section .

Definitions

Dictionary definitions are those pieces of prose in a dictionary entry that describe the meaning of some lexical item. Most often, definitions describe the headword of the entry; in some cases, they describe translated texts, examples, etc.; see cit type="translation", section , and cit type="example", section . The def element directly contains the text of the definition; unlike form and gramGrp, it does not serve solely to group a set of smaller elements. The close analysis of definition text, such as the tagging of hypernyms, typical objects, etc., is not covered by these Guidelines.

Definitions may occur directly within an entry; when multiple definitions are given, they are typically identified as belonging to distinct senses, as here: demigod (…) n. 1.a. a being who is part mortal, part god. b. a lesser deity. 2. a godlike person. CP

n a being who is part mortal, part god. a lesser deity. a godlike person.

In multilingual dictionaries, it is sometimes possible to distinguish translation equivalents from definitions proper; here a def element is distinguished from the translation information within which it appears. rémoulade [Remulad] nf remoulade, rémoulade (dressing containing mustard and herbs). CR

n f remoulade rémoulade dressing containing mustard and herbs

Translation Equivalents

Multilingual dictionaries contain information about translations of a given word in some source language for one or more target languages. Minimally, the dictionary provides the corresponding translation in the target language; other material, such as morphological information (gender, case), various kinds of usage restrictions, etc., may also be given. If translation equivalents are to be distinguished from other kinds of sense information, they may be encoded using cit type="translation". The global xml:lang attribute should be used to specify the target language.

As in monolingual dictionaries, the sense element is used in multilingual dictionaries to group information (forms, grammatical information, usage, translation(s), etc.) about a given sense of a word where necessary. Information about the individual translation equivalents within a sense is grouped using cit type="translation". This information may include the translation text (tagged q or quote), morphological information (gen, case, etc.), usage notes (usg), translation labels (lbl), and definitions (def).When bibliographic data is provided, the quote element should be used.

Note how in the following example, different translation equivalents are grouped into the same or different senses, following the punctuation of the source and the usage labels: dresser … (a) (Theat) habilleur m, -euse f; (Comm: window ~) étalagiste mf. she's a stylish ~ elle s'habille avec chic; V hair. (b) (tool) (for wood) raboteuse f; (for stone) rabotin m. CR

Theat habilleur m -euse f Comm étalagiste mf she's a stylish elle s'habille avec chic V. hair tool for wood raboteuse f for stone rabotin m

In the following example, a distinction is made between the translation equivalent (OAS) and a descriptive phrase providing further information for the user of the dictionary. O.A.S. ... nf (abrév de Organisation de l'Armée secrète) OAS (illegal military organization supporting French rule of Algeria). CR OAS illegal military organization supporting French rule of Algeria

Note that cit type="translation" may also be used in monolingual dictionaries when a translation is given for a foreign word: havdalah or havdoloh Hebrew. (Hebrew hAvdA"lA; Yiddish hAv"dOl@) n. Judaism. the ceremony marking the end of the sabbath or of a festival, including the blessings over wine, candles and spices. [literally: separation] CED

Judaism the ceremony marking the end of the sabbath or of a festival, including the blessings over wine, candles and spices. literally separation

Etymological Information

The element etym marks a block of etymological information. Etymologies may contain highly structured lists of words in an order indicating their descent from each other, but often also include related words and forms outside the direct line of descent, for comparison. Not infrequently, etymologies include commentary of various sorts, and can grow into short (or long!) essays with prose-like structure. This variation in structure makes it impracticable to define tags which capture the entire intellectual structure of the etymology or record the precise interrelation of all the words mentioned. It is, however, feasible to mark some of the more obvious phrase-level elements frequently found in etymologies, using tags defined in the core module or elsewhere in this chapter. Of particular relevance for the markup of etymologies are:

As in other prose, individual word forms mentioned in an etymological description are tagged with mentioned elements. Pronunciations, usage labels, and glosses can be tagged using the pron, usg, and gloss elements defined elsewhere in these Guidelines. In addition, the lang element may be used to identify a particular language name where it appears, in addition to using the xml:lang attribute of the mentioned element.

Examples: abismo m. (del gr. a priv. y byssos, fondo). Sima, gran profundidad. …

del gr. a priv. y byssos, fondo neume \'n(y)üm\ n [F, fr. ML pneuma, neuma, fr. Gk pneuma breath — more at pneumatic]: any of various symbols used in the notation of Gregorian chant … [WNC] F fr. ML pneuma neuma fr. Gk pneuma breath more at any of various symbols used in the notation of Gregorian chant

Other Information

Examples

Dictionaries typically include examples of word use, usually accompanying definitions or translations. In some cases, the examples are quotations from another source, and are occasionally followed by a citation to the author.

The cit type="example" element contains usage examples and associated information; the example text itself should be enclosed in a q or quote element. The cit element associates a quotation with a bibliographic reference to its source.

Examples frequently abbreviate the headword, and so their transcription will frequently make use of the oRef or oVar elements described below in section .

Examples: multiplex /…/ adj tech having many parts: the multiplex eye of the fly. LDOCE the multiplex eye of the fly. Or when one wants a more comprehensive representation of examples: the multiplex eye of the fly. As the following example shows, cit can also contain elements such as pron, def, etc. some … 4. (S~ and any are used with more): Give me ~ more/s@'mO:(r)/ OALD and any are used with more Give me more s@'mO:(r) In multilingual dictionaries, examples may also be accompanied by translations: horrifier … vt to horrify. elle était horrifiée par la dépense she was horrified at the expense. CR to horrify elle était horrifiée par la dépense she was horrified at the expense. When a source is indicated, the example should be marked with a bibl element: valeur … n. f. … 2. Vx. Vaillance, bravoure (spécial., au combat). La valeur n'attend pas le nombre des années (Corneille). … DNT Vx. Vaillance, bravoure (spécial., au combat) La valeur n'attend pas le nombre des années Corneille

Usage Information and Other Labels

Most dictionaries provide restrictive labels and phrases indicating the usage of given words or particular senses. Other phrases, not necessarily related to usage, may also be attached to forms, translations, cross-references, and examples. The following elements are provided to mark up such labels: As indicated in the following section (), the lbl element may be used for any kind of significative phrase or label within the text. The usg element is a specialization of this to mark usage labels in particular. Usage labels typically indicate temporal use (archaic, obsolete, etc.) register (slang, formal, taboo, ironic, facetious, etc.) style (literal, figurative, etc.) connotative effect (e.g. derogatory, offensive) subject field (Astronomy, Philosophy, etc.) national or regional use (Australian, U.S., Midland dialect, etc.) Many dictionaries provide an explanation and/or a list of such usage labels in a preface or appendix. The type of the usage information may be indicated in the type attribute on the usg element. Some typical values are: geo geographic area time temporal, historical era (archaic, old, etc.) dom domain reg register style style (figurative, literal, etc.) plev preference level (chiefly, usually, etc.) acc acceptability lang language for foreign words, spellings pronunciations, etc. gram grammatical usage In addition to this kind of information, multilingual dictionaries often provide semantic cues to help the user determine the right sense of a word in the source language (and hence the correct translation). These include synonyms, concept subdivisions, typical subjects and objects, typical verb complements, etc. These labels may also be marked with the usg element; sample values for the type attribute in these cases include: syn synonym given to show use hyper hypernym given to show usage colloc collocation given to show usage comp typical complement obj typical object subj typical subject verb typical verb hint unclassifiable piece of information to guide sense choice

In this entry, one spelling is marked as geographically restricted: colour or U.S. color … CED

In the next example, usage labels are used to indicate domains, register, and synonyms associated with different senses: palette [palEt] nf (a) (Peinture: lit, fig) palette. (b) (Boucherie) shoulder. (c) (aube de roue) paddle; (battoir à linge) beetle; (Manutention, Constr) pallet. CR Peinture lit fig palette Boucherie shoulder aube de roue paddle battoir à linge beetle Manutention Constr pallet

When the usage label is hard to classify, it may be described as a hint: rempaillage […] nm reseating, rebottoming (with straw). CR reseating rebottoming with straw

Cross-References to Other Entries

Dictionary entries frequently refer to information in other entries, often using extremely dense notations to convey the headword of the entry to be sought, the particular part of the entry being referred to, and the nature of the information to be sought there (synonyms, antonyms, usage notes, etymology, an illustration, etc.)

Cross-references may be tagged in dictionaries using the ref and ptr elements defined in the core module (section ). In addition, the xr element may be used to group all the information relating to a cross-reference.

As in other types of text, the actual pointing element (e.g. ref or ptr) is used to tag the cross-reference target proper (in dictionaries, usually the headword, possibly accompanied by a homograph number, a sense number, or other further restriction specifying what portion of the target entry is being referred to). The xr element is used to group the target with any accompanying phrases or symbols used to label the cross-reference; the cross-reference label itself may be tagged as a lbl or may remain untagged. Both of the following are thus legitimate: glee … Compare madrigal (sense 1) CED

Compare hostellerie Syn. de hôtellerie (sens 1). DNT Syn. de hôtellerie (sens 1). In addition to using, or not using, lbl to mark the cross-reference label, the two examples differ in another way. The former assumes that the first sense of madrigal has the identifier madrigal.1, and that the specific form of the reference in the source volume can be reconstructed, if needed, from that information. The latter does not require the first sense of hôtellerie to have an identifier, and retains the print form of the cross-reference; by omitting the target attribute of the ref element, however, the second example does assume implicitly either that some software could usefully parse the phrase tagged as a ref and find the location referred to, or else that such processing will not be necessary.

The type attribute on the pointing element or on the xr element may be used to indicate what kind of cross-reference is being made, using any convenient typology. Since different dictionaries may label the same kind of cross-reference in different ways, it may be useful to give normalized indications in the type attribute, enabling the encoder to distinguish irregular forms of cross-reference more reliably: rose2 … vb. the past tense of rise. CED

the past tense of rise from cross-references for synonyms and the like: antagonist … syn see adverse W7 syn see adverse Strictly speaking, the reference above is not to the entry for adverse, but to the list of synonyms found within that entry. In some cases, the cross-reference is to a particular subset of the meanings of the entry in question: globe …V. armillaire (sphère) PR V. armillaire sphère

Cross-references occasionally occur in definition texts, example texts, etc., or may be free-standing within an entry. These may typically be encoded using ref or ptr, without an enclosing xr. For example: entacher … Acte entaché de nullité, contenant un vice de forme ou passé par un incapable*. DNT The asterisk signals a reference to the entry for incapable. contenant un vice de forme ou passé par un . In some cases, the form in the definition is inflected, and thus ref must be used to indicate more exactly the intended target, as here: justifier …4. IMPRIM Donner a (une ligne) une longeur convenable au moyen de blancs (2, sens 1, 3). DNT imprim Donner a (une ligne) une longeur convenable au moyen de blancs (2, sens 1, 3) ...

Notes within Entries

Dictionaries may include extensive explanatory notes about usage, grammar, context, etc. within entries. Very often, such notes appear as a separate section at the end of an entry. The standard note element should be used for such material.

For example: ain't (eInt) Not standard. contraction of am not, is not, are not, have not or has not: I ain't seen it. ….Usage. Although the interrogative form ain't I? would be a natural contraction of am I not?, it is generally avoided in spoken English and never used in formal English. CED

Not standard I ain't seen it. Although the interrogative form ain't I? would be a natural contraction of am I not?, it is generally avoided in spoken English and never used in formal English.

The formal declaration for note is given in section .

Related Entries

The re element encloses a degenerate entry which appears in the body of another entry for some purpose. Many dictionaries include related entries for direct derivatives or inflected forms of the entry word, or for compound words, phrases, collocations, and idioms containing the entry word.

Related entries can be complex, and may in fact include any of the information to be found in a regular entry. Therefore, the re element is defined to contain the same elements as an entry element, with the exception that it may not contain any nested re elements.

Examples: bevvy ("bEvI) Dialect. ~ n., pl. -vies. 1. a drink, esp. an alcoholic one: we had a few bevvies last night. 2. a night of drinking. ~ vb. - vies, -vying, -vied (intr.) 3. to drink alcohol [probably from Old French bevee, buvee, drinking] —'bevvied adj. CED

Dialect n a drink, esp. an alcoholic one: we had a few bevvies last night. vb to drink alcohol probably from Old French bevee, buvee drinking adj

Headword and Pronunciation References

Examples, definitions, etymologies, and occasionally other elements such as cross-references, orthographic forms, etc., often contain a shortened or iconic reference to the headword, rather than repeating the headword itself. The references may be to the orthographic form or to the pronunciation, to the form given or to a variant of that form. The following elements are used to encode such iconic references to a headword:

These elements all inherit the following attributes from the class att.pointing which may optionally be used to resolve any ambiguity about the headword form being referred to.

Headword references come in a variety of formats: ~ indicates a reference to the full form of the headword pref~ gives a prefix to be affixed to the headword ~suf gives a suffix to be affixed to the headword A~ gives the first letter in uppercase, indicating that the headword is capitalized pref~suf gives a prefix and a suffix to be affixed to the headword a. gives the initial of the word followed by a full stop, to indicate reference to the full form of the headword A. refers to a capitalized form of the headword

The oRef element should be used for iconic or shortened references to the orthographic form(s) of the headword itself. It is an empty element and replaces, rather than enclosing, the reference. Note that the reference to a headword is not necessarily a simple string replacement. In the example colour1, (US = color) …~ films; ~ TV; Red, blue and yellow are ~s. OALD, the tilde stands for either headword form (colour, color).

Examples: colonel … army officer above a lieutenant-~. OALD army officer above a lieutenant- academy … The Royal A~ of Arts OALD The Royal of Arts

The following example demonstrates the use of the target attribute to refer to a specific form of the headword: vag- or vago- comb form … : vagus nerve < vagal > < vagotomy > W7

vagus nerve al tomy

In many cases the reference is not to the orthographic form of the headword, but rather to another form of the headword—usually to an inflected form. In these cases, the element oVar should be used; this element takes as its content the string as it appears in the text. take … < Mr Burton took us for French > NPEG Mr Burton took us for French take … < was quite ~n with him > NPEG was quite n with him

The next example shows a discontinuous reference, using the attributes next and prev, which are defined in the additional module for linking, segmentation, and alignment (see chapter ) and therefore require that that module be selected in addition to that for dictionaries. mix up… < it's easy to mix her up with her sister > NPEG it's easy to mix her up with her sister

In addition, some dictionaries make reference to the pronunciation of the headword in the pronunciation of related entries, variants, or examples. The pRef and pVar elements should be used for such references. hors d'oeuvre /,aw'duhv (Fr O:r dœvr)/ n, pl hors d'oeuvres also hors d'oeuvre /'duhv(z) (Fr ~)/ NPEG

Because headword and pronunciation references can occur virtually anywhere in an entry, the oRef, oVar, pRef, and pVar elements can appear within any other element defined for dictionary entries.

Since existing printed dictionaries use different conventions for headword references (swung dash, first letter abbreviated form, capitalization, or italicization of the word, etc.) the exact method used should be documented in the header.

Typographic and Lexical Information in Dictionary Data

Among the many possible views of dictionaries, it is useful to distinguish at least the following three, which help to clarify some issues raised with particular urgency by dictionaries, on account of the complexity of both their typography and their information structure. (a) the typographic view—the two-dimensional printed page, including information about line and page breaks and other features of layout (b) the editorial view—the one-dimensional sequence of tokens which can be seen as the input to the typesetting process; the wording and punctuation of the text and the sequencing of items are visible in this view, but specifics of the typographic realization are not (c) the lexical view—this view includes the underlying information represented in a dictionary, without concern for its exact textual form

For example, a domain indication in a dictionary entry might be broken over a line and therefore hyphenated (naut- ical); the typographic view of the dictionary preserves this information. In a purely editorial view, the particular form in which the domain name is given in the particular dictionary (as nautical, rather than naut., Naut., etc.) would be preserved, but the fact of the line break would not. Font shifts might plausibly be included in either a strictly typographic or an editorial view. In the lexical view, the only information preserved concerning domain would be some standard symbol or string representing the nautical domain (e.g. naut.) regardless of the form in which it appears in the printed dictionary.

In practice, publishers begin with the lexical view—i.e., lexical data as it might appear in a database—and generate first the editorial view, which reflects editorial choices for a particular dictionary (such as the use of the abbreviation Naut. for nautical, the fonts in which different types of information are to be rendered, etc.), and then the typographic view, which is tied to a specific printed rendering. Computational linguists and philologists often begin with the typographic view and analyse it to obtain the editorial and/or lexical views. Some users may ultimately be concerned with retaining only the lexical view, or they may wish to preserve the typographic or editorial views as a reference text, perhaps as a guard against the loss or misinterpretation of information in the translation process. Some researchers may wish to retain all three views, and study their interrelations, since research questions may well span all three views.

In general, an electronic encoding of a text will allow the recovery of at least one view of that text (the one which guided the encoding); if editorial and typographic practices are consistently applied in the production of a printed dictionary, or if exceptions to the rules are consistently recorded in the electronic encoding, then it is in principle possible to recover the editorial view from an encoding of the lexical view, and the typographic view from an encoding of the editorial view. In practice, of course, the severe compression of information in dictionaries, the variety of methods by which this compression is achieved, the complexity of formulating completely explicit rules for editorial and typographic practice, and the relative rarity of complete consistency in the application of such rules, all make the mechanical transformation of information from one view into another something of a vexed question.

This section describes some principles which may be useful in capturing one or the other of these views as consistently and completely as possible, and describes some methods of attempting to capture more than one view in a single encoding. Only the editorial and lexical views are explicitly treated here; for methods of recording the physical or typographic details of a text, see chapter . Other approaches to these problems, such as the use of repetitive encoding and links to show their correspondences, or the use of feature structures to capture the information structure, and of the ana and inst attributes to link feature structures to a transcription of the editorial view of a dictionary, are not discussed here (for feature structures, see chapter . For linkage of textual form and underlying information, see chapter ).

Editorial View

Common practice in encoding texts of all sorts relies on principles such as the following, which can be used successfully to capture the editorial view when encoding a dictionary: All characters of the source text should be retained, with the possible exception of rendition text (for which see further below). Characters appearing in the source text should typically be given as character data content in the document, rather than as the value of an attribute; again, rendition text may optionally be excepted from this rule. Apart from the characters or graphics in the source text, nothing else should appear as content in the document, although it may be given in attribute values. The material in the source text should appear in the encoding in the same order. Complications of the character sequence by footnotes, marginal notes, etc., text wrapping around illustrations, etc., may be dealt with by the usual means (for notes, see section ).Complications of sequence caused by marginal or interlinear insertions and deletions, which are frequent in manuscripts, or by unconventional page layouts, as in concrete poetry, magazines with imaginative graphic designers, and texts about the nature of typography as a medium, typically do not occur in dictionaries, and so are not discussed here.

In a very conservative transcription of the editorial view of a text, rendition characters (e.g. the commas, parentheses, etc., used in dictionary entries to signal boundaries among parts of the entry) and rendition text (for example, conjunctions joining alternate headwords, etc.) are typically retained. Removing the tags from such a transcription will leave all and only the characters of the source text, in their original sequence.This is a slight oversimplification. Even in conservative transcriptions, it is common to omit page numbers, signatures of gatherings, running titles and the like. The simple description above also elides, for the sake of simplicity, the difficulties of assigning a meaning to the phrase original sequence when it is applied to the printed characters of a source text; the original sequence retained or recovered from a conservative transcription of the editorial view is, of course, the one established during the transcription by the encoder.

Consider, for example, the following entry: pinna ('pIn@) n., pl. -nae (-ni:) or -nas. 1. any leaflet of a pinnate compound leaf. 2. Zoology. a feather, wing, fin, or similarly shaped part. 3. another name for auricle (sense 2). [C18: via New Latin from Latin: wing, feather, fin] CED A conservative encoding of the editorial view of this entry, which retains all rendition text, might resemble the following:

n., or -nas 1. any leaflet of a pinnate compound leaf. 2. Zoology a feather, wing, fin, or similarly shaped part. 3. another name for auricle (sense 2). [C18: via New Latin from Latin: wing, feather, fin]

A somewhat simplified encoding of the editorial view of this entry might exploit the fact that rendition text is often systematically recoverable. For example, parentheses consistently appear around pronunciation in this dictionary, and thus are effectively implied by the start- and end-tags for pron.The omission of rendition text is particularly common in systems for document production; it is considered good practice there, since automatic generation of rendition text is more reliable and more consistent than attempting to maintain it manually in the electronic text. In such an encoding, removing the tags should exactly reproduce the sequence of characters in the source, minus rendition text. The original character sequence can be recovered fully by replacing tags with any rendition text they imply.

Encoding in this way, the example given above might resemble the following. The tagUsage element in the header would be used to record the following patterns of rendition text: parentheses appear around pron elements commas appear before inflected forms the word or appears before alternate forms brackets appear around the etymology full stops appear after pos, inflection information, and sense numbers senses are numbered in sequence unless otherwise specified using the global n attribute

n -nas any leaflet of a pinnate compound leaf. Zoology a feather, wing, fin, or similarly shaped part. another name for auricle (sense 2). C18: via New Latin from Latin: wing, feather, fin

When rendition text is omitted, it is recommended that the means to regenerate it be fully documented, using the tagUsage element of the TEI header.

If rendition text is used systematically in a dictionary, with only a few mistakes or exceptions, the global attribute rend may be used on any tag to flag exceptions to the normal treatment. The values of the rend attribute are not prescribed, but it can be used with values such as no-comma, no-left-paren, etc. Specific values can be documented using the rendition element in the TEI header.

In the following (imaginary) example, no left parenthesis precedes the pronunciation: biryani or biriani %bIrI"A:nI) any of a variety of Indian dishes … [from Urdu] This irregularity can be recorded thus:

any of a variety of Indian dishes … from Urdu

Lexical View

If the text to be interchanged retains only the lexical view of the text, there may be no concern for the recoverability of the editorial (not to speak of the typographic) view of the text. However, it is strongly recommended that the TEI header be used to document fully the nature of all alterations to the original data, such as normalization of domain names, expansion of inflected forms, etc.

In an encoding of the lexical view of a text, there are degrees of departure from the original data: normalizing inconsistent forms like nautical, naut., Naut., etc., to nautical is a relatively slight alteration; expansion of delay -ed -ing to delay, delayed, delaying is a more substantial departure. Still more severe is the rearranging of the order of information in entries; for example: reorganizing the order of elements in an entry to show their relationship, as in clem (klEm) or clam vb. clems, clemming, clemmed or clams, clamming, clammed CED where in a strictly lexical view one might wish to group clem and clam with their respective inflected forms. splitting an entry into two separate entries, as in celi.bacy /"selIb@sI/ n [U] state of living unmarried, esp as a religious obligation. celi.bate /"selIb@t/ n [C] unmarried person (esp a priest who has taken a vow not to marry). OALD For some purposes, this entry might usefully be split into an entry for celibacy and a separate entry for celibate.

An encoding which captures the lexical view of the example given in the previous section might look something like the following. In this encoding: abbreviated forms have been silently expanded some forms have been moved to allow related forms to be grouped together the part of speech information has been moved to allow all forms to be given together the cross-reference to auricle has been simplified

pinnas n any leaflet of a pinnate compound leaf. Zoology a feather, wing, fin, or similarly shaped part. C18: via New Latin from Latin: wing, feather, fin

Whether the given dictionary encoding focusses on the lexical view and thus approaches the status of lexical databases, or uses the typographic/editorial view approach and needs to communicate the sometimes informally stated values for the particular descriptive features, the issue of interoperability of the content and of the container objects becomes relevant, in view of the growing tendency to interlink pieces of information across Internet resources. In such cases, it becomes crucial to be able to encode the fact that whether the information on, for instance, the value of the grammatical category of Number is provided as "sg.", "sing.", "Singular", or equivalently "poj." in Polish, or "Ez." in German, etc., what is actually referred to is always the same grammatical value that can be rendered with a plethora of markers, depending on the publisher, language, or lexicographic tradition. In order to signal that this variety of surface markers in fact indicate the same underlying value, it is possible to align them with an external inventory of standardized values. The TEI provides means to align grammatical categories as well as their content with the ISOcat reference, which is a Web implementation of ISO 12620.

In the example below, a fragment of the entry for isotope cited in section is adorned by references to ISOcat definitions for "part of speech" (dcr:datcat) and "adjective" (dcr:valueDatcat). Depending on the status and extent of the dictionary, various strategies may be used to reduce the redundancy of the repeated ISOcat references.

adj

Retaining Both Views

It is sometimes desirable to retain both the lexical and the editorial view, in which case a potential conflict exists between the two. When there is a conflict between the encodings for the lexical and editorial views, the principles described in the following sections may be applied.

Using Attribute Values to Capture Alternate Views

If the order of the data is the same in both views, then both views may be captured by encoding one dominant view in the character data content of the document, and encoding the other using attribute values on the appropriate elements. If all tags were to be removed, the remaining characters would be those of the dominant view of the text.

The attribute class att.lexicographic is used to provide attributes for use in encoding multiple views of the same dictionary entry. These attributes are available for use on all elements defined in this chapter when the base module for dictionaries is selected.

When the editorial view is dominant, the following attributes may be used to capture the lexical view:

When the lexical view is dominant, the following attributes may be used to record the editorial view:

One attribute is useful in either view:

For example, if the source text had the domain label naut., it might be encoded as follows. With the editorial view dominant: naut. The lexical view of the same label would transcribe the normalized form as content of the usg element, the typographic form as an attribute value: nautical

If the source text gives inflectional information for the verb delay as delay, -ed, -ing, it might usefully be expanded to delayed, delayed, delaying. An encoding of the editorial view might take this form:

delay -ed

-ing Note the use of the tns tag with null content, to enable the representation of implicit information even though it has no print realization.

The lexical view might be encoded thus:

A particular problem may be posed by the common practice of presenting two alternate forms of a word in a single string, by marking some parts of the word as optional in some forms. The following entry is for a word which can be spelled either thyrostimuline or thyréostimuline: thyr(é)ostimuline [tiR(e)ostimylin] … With the editorial view dominant, this entry might begin thus:

With the lexical view dominant, however, two orth and two pron elements would be encoded, in order to disentangle the two forms; the orig attribute would be used to record the typographic presentation of the information in the source.

This example might also be encoded using the opt attribute combined with the attributes next and prev defined in chapter .

Note that this transcription preserves both the lexical and editorial views in a single encoding. However, it has the disadvantage that the strings corresponding to entire words do not appear in the encoding uninterrupted, and therefore complex processing is required to retrieve them from the encoded text. The use of the opt attribute is recommended, however, when long spans of text are involved, or when the optional part contains embedded tags.

For example, the following gives two definitions in one text: picture drawn with coloured chalk made into crayons, and coloured chalk made into crayons: pas.tel /"pastl US: pa"stel/ n 1 (picture drawn with) coloured chalk made into crayons. 2… OALD

A simple encoding solution would be to leave the definition text unanalysed, but this might be felt inadequate since it does not show that there are two definitions. A possible alternative encoding would be: coloured chalk made into crayons picture drawn with coloured chalk made into crayons

This transcribes some characters of the source text twice, however, which deviates from the usual practice. The following encoding records both the editorial and lexical views: picture drawn with coloured chalk made into crayons

Recording Original Locations of Transposed Elements

The attributes described in the previous section are useful only when the order of material is the same in both the editorial and the lexical view. When the two views impose different orders on the data, the standard linking mechanisms may be used to show the original location of material transposed in an encoding of the lexical view.

If the original is only slightly modified, the anchor element may be used to mark the original location of the material, and the location attribute may be used on the lexical encoding of that material to indicate its original location(s). Like those in the preceding section, this attribute is defined for the attribute class att.lexicographic:

For example: pinna ("pIn@) n., pl. -nae (-ni:) or -nas. CED

Unstructured Entries

The content model for the entry element provides an entry structure suitable for many average dictionaries, as well as many regular entries in more exotic dictionaries. However, the structure of some dictionaries does not allow the restrictions imposed by the content model for entry. To handle these cases, the entryFree and dictScrap elements are provided to support much wider variation in entry structure. The dictScrap element offers less freedom, in that it can only contain phrase level elements, but it can itself appear at any point within a dictionary entry where any of the structural components of a dictionary entry are permitted. As such, it acts as a container for otherwise anomalous parts of an entry.

The entryFree element places no constraints at all upon the entry: any element defined in this chapter, as well as all the normal phrase-level and inter-level elements, can appear anywhere within it. With the entryFree element, the encoder is free to use any element anywhere, as well as to use or omit grouping elements such as form, gramGrp, etc.

The entryFree element allows the encoding of entries which violate the structure specified for the entry element. For example, in the following entry from a dictionary already in electronic form, it is necessary to include a pron element within a def. This is not permitted in the content model for entry, but it poses no problem in the entryFree element. <ent h="demigod"> <hwd>demi|god</hwd> <pr> <ph>"demIgQd</ph> </pr> <hps ps="n"> <hsn> <def>one who is partly divine and partly human</def> <def>(in Gk myth, etc) the son of a god and a mortal woman, eg<cf>Hercules</cf> <pr> <ph>"h3:kjUli:z</ph> </pr> </def> </hsn> </hps> </ent> (OALD)

n one who is partly divine and partly human (in Gk myth, etc) the son of a god and a mortal woman, eg Hercules "h3:kjUli:z

The entryFree element also makes it possible to transcribe a dictionary using only phrase-level (atomic) elements—that is, using no grouping elements at all. This can be desirable if the encoder wants a completely flat view, with no indication of or commitment to the association of one element with another. The following encoding uses no grouping elements, and keeps all rendition text: biryani or biriani (%bIrI"A:nI) any of a variety of Indian dishes…[from Urdu] CED biryani or biriani (%bIrI"A:nI) any of a variety of Indian dishes … [from Urdu]

Here is an alternative way of representing the same structure, this time using dictScrap: biryani or biriani (%bIrI"A:nI) any of a variety of Indian dishes … [from Urdu]

The Dictionary Module

The module defined in this chapter makes available the following components: Dictionaries Dictionaries Dictionnaires 紙本字典 Dizionari a stampaDicionários impressos辞書モジュール The selection and combination of modules to form a TEI schema is described in .