The decls attribute may be used in this way with the elements text, div, u, and writing only. It should only be used to over-ride declarations for the elements encoding.decl, script.decl and recording.decl components of the TEI header. In general, TEI recommended practice is to avoid heterogeneity in the contents of divisions: for this reason the decls attribute should only exceptionally be needed. Temporal Information

In addition to the global attributes n, id, and lang, utterances, vocals, pauses, kinesics, events and writing elements may all take a common set of attributes providing information about their position in time. For this reason, these elements are regarded as forming a class, referred to here as timed. The following attributes are common to all elements in this class: start indicates the location within a temporal alignment at which this element begins. end indicates the location within a temporal alignment at which this element ends. dur indicates the length of this element in time, using either specific units or the units specified on the associated temporal alignment. Note that if start and end point to loc elements whose temporal distance from each other is specified in the alignment map, then dur is ignored.

The tag ptr (see ) tag may be used as an alternative means of aligning the start and end of timed elements, and is required where the temporal alignment is with points within an element.

The parameter entity corresponding with this class has the following declaration: ]]> Utterances

Each distinct utterance in a spoken text is represented by a u element, described as follows: u a stretch of speech usually preceded and followed by silence or by a change of speaker. Attributes include: who supplies an identifier for the speaker or group of speakers. Its value is the identifier of a participant or participant.grp element in the TEI header. trans indicates the nature of the transition between this utterance and the previous one. Values: smooth no noticeable pause or overlap. latching overlapping or latching transition. pause noticeable pause. Use of the who attribute to associate the utterance with a particular speaker is recommended, but not required. Its use implies as a further requirement that all speakers be identified by a participant or participant.grp element in the TEI header, (see section ). The trans attribute is provided as a means of characterizing the transition from one utterance to the next at a simpler level of detail than that provided by the general alignment mechanism discussed in section below. For example Have you heard the the election results? yes it's a disaster ]]>

An utterance may contain only running text, or text within which other basic structural elements are nested. Where such nesting occurs, the who attribute is considered to be inherited for the elements pause, vocal, shift and kinesic; that is, a pause (etc.) within an utterance is regarded as being a pause by that speaker only, while a pause between utterances applies to all speakers.

Occasionally, an utterance may contain other utterances, for example where there is a change in the script associated with it. This may occur when a speaker changes script in mid-utterance. For example: Listen to this The government is confident, he said, that the current economic problems will be completely overcome by June what nonsense ]]> Here speaker A interrupts his own utterance with another nested one, which is read from a newspaper. The decls attribute on the nested utterance is used to indicate that its script is S1, rather than the default. Alternatively, the embedded utterance might be regarded as a new (non-nested) one, or represented as an event without transcribing the read material: Listen to this what nonsense ]]>

The formal definition of these elements is as follows: ]]> Pause

The pause empty element tag is used to indicate a silent pause, either between or within utterances. A pause contained by an utterance applies to the utterer of that utterance. A pause between utterances applies to all utterers. The type attribute may be used to categorize the pause, for example as short, medium or long; alternatively the attribute dur may be used to indicate its length more exactly. If detailed synchronization of pausing with other vocal phenomena is required, the alignment mechanism discussed at section should be used. Note that the trans attribute mentioned in the previous section may also be used to characterize the degree of pausing between utterances.

The formal definition of this element is as follows: ]]> Vocal, Kinesic, Event

These three empty elements are used to indicate the presence of non-transcribed semi-lexical or non-lexical phenomena either between or within utterances. vocal is used for vocalizations such as filled pauses, kinesic for gestures, and event for other types of event.

The who attribute should be used to specify the person or group responsible for a vocal, kinesic or event which is contained within an utterance, if this differs from that of the enclosing utterance. The attribute must be supplied for a vocal, kinesic or event which is not contained within an utterance.

The iterated attribute may be used to indicate that the vocal, kinesic or event is repeated, for example laughter as opposed to laugh. If detailed synchronization of these phenomena is required, the alignment mechanism discussed at section should be used.

The desc attribute may be used to supply a conventional representation for the phenomenon, for example: lexical burp, click, cough, exhale, giggle, gulp, inhale, laugh, sneeze, sniff, snort, sob, swallow, throat, yawn semi-lexical ah, aha, aw, eh, ehm, er, erm, hmm, huh, mm, mmhm, oh, ooh, oops, phew, tsk, uh, uh-huh, uh-uh, um, urgh, yup Researchers may prefer to regard some semi-lexical phenomena as words within the bounds of the u element. See further the discussion at section below. As for all basic categories, the definition should be made clear in the encoding.decl element of the header.

It is not envisaged that events or kinesics should be included in any transcription, except where their presence is felt to be of potential significance for the interpretation of the interaction.

The formal definition of these elements is as follows: ]]> Writing

Written text may also be encountered when speech is transcribed, for example in a television broadcast or cinema performance, or where one participant shows written text to another. The writing element may be used to distinguish such written elements from the spoken text in which they are embedded. writing a passage of written text revealed to participants in the course of a spoken text. Attributes include: who supplies an identifier for the participant who reveals or creates the writing, if any. Its value is the identifier of a participant or participant.grp element in the TEI header. gradual indicates whether the writing is revealed all at once or gradually. Values: y the writing is revealed gradually. n the writing is revealed all at once. u unknown or unmarked. type categorizes the kind of writing in some way, for example as a subtitle, noticeboard etc.

The formal definition of these elements is as follows: ]]> Segmentation and Alignment

As mentioned above, an utterance may contain simply prose, possibly mixed with the other elements already discussed above. It may also be further subdivided into segments of various kinds, and may contain pointers, indicating points of synchronization within it, or shifts, indicating locations at which changes are noted in various para-linguistic phenomena. These are discussed in the following sections. Segments

For some analytic purposes it may be desirable to subdivide the divisions of a spoken text into units smaller than the individual utterance or turn. Segmentation may be performed for a number of different purposes and in terms of a variety of speech phenomena. Common examples include units defined both prosodically (by intonation, pausing etc.) and syntactically (clauses, phrases etc.) The term macrosyntagm has been used to define units peculiar to speech transcripts by a number of researchers. The term was apparently first proposed by Bengt Loman and Nils Jørgensen, in (Lund: Studentlitteratur, 1971) where it is defined as follows: A text can be analysed as a sequence of segments which are internally connected by a network of syntactic relations and externally delimited by the absence of such relations with respect to neighbouring segments. Such a segment is a syntactic unit called a macrosyntagm (trans. S. Johansson). These Guidelines propose that such analyses be performed in terms of neutrally-named segments, represented by an s element. This element may take a type attribute to specify the kind of segmentation applicable to a particular segment, if more than one is possible in a text. A full definition of the segmentation scheme or schemes used should be provided in the segmentation element of the editorial.decl element in the TEI header (see ).

The s element has the following formal definition: ]]>

It is often the case that the desired segmentation does not respect utterance boundaries, for example, syntactic units may cross utterance boundaries. This may be handled in a number of different ways, most of which are discussed in more general terms elsewhere in these Guidelines. a concurrent DTD may be defined (see ) milestone tags may be used (see ); the special-purpose shift tag discussed in section below is an extension of this method where the conflict is between utterance and segment boundaries only, it may be regarded as an instance of overlap (see section and ) where several discontinuous segments are to be grouped together to form a syntactic unit (e.g. a phrasal verb with interposed complement), the general purpose alignment mechanism should be used (see sections and Shifts

Paralinguistic features which characterize stretches of speech not co-extensive with utterances or any of the other units discussed so far may be encoded to a limited extent by marking simply their boundaries. The shift empty element is provided for this purpose. It may appear within an utterance or a segment to mark a significant change in the particular feature defined by its attributes, which is then understood to apply to all subsequent utterances for the same speaker, unless changed by a new shift for the same feature in the same speaker. Intervening utterances by other speakers do not normally carry the same feature.

The feature attribute is used to identify the particular feature which noticeably changes at this point. Its value is taken from a closed list of paralinguistic features, based on those used by the Survey of English Usage; For details see S. Boase, (London: Survey of English Usage, University College London, 1990). this list may be revised or supplemented using the methods outlined in section .

The new attribute specifies the new state of the feature following the shift. If no value is specified, it is implied that the feature concerned ceases to be remarkable at this point: the special value normal may be specified to have the same effect.

A shift itself can be aligned with other elements only by being pointed at from an alignment. If therefore alignment is expressed by pointing from the text, rather than into it, a ptr element following the shift should be used. See further . In summary, shift marks the point at which some paralinguistic feature of a series of utterances by any one speaker changes. Attributes include: feature a paralinguistic feature. Values: tempo speed of utterance. loud loudness. pitch pitch range. tension tension or stress pattern. rhythm rhythmic qualities. voice voice quality. new specifies the new state of the paralinguistic feature specified.

A list of suggested values for each of the features proposed follows: tempo a allegro (fast) aa very fast acc accelerando (getting faster) l lento (slow) ll very slow rall rallentando (getting slower) loud (for loudness): f forte (loud) ff very loud cresc crescendo (getting louder) p piano (soft) pp very soft dimin diminuendo (getting softer) pitch (for pitch range): high high pitch-range low low pitch-range wide wide pitch-range narrownarrow pitch-range asc ascending desc descending monot monotonous scand scandent, each succeeding syllable higher than the last, generally ending in a falling tone tension: sl slurred lax lax, a little slurred ten tense pr very precise st staccato, every stressed syllable being doubly stressed leg legato, every syllable receiving more or less equal stress rhythm: rh beatable rhythm arrh arrhythmic, particularly halting spr spiky rising, with markedly higher unstressed syllables spf spiky falling, with markedly lower unstressed syllables glr glissando rising, like spiky rising but the unstressed syllables, usually several, also rise in pitch relative to each other glf glissando falling, like spiky falling but with the unstressed syllables also falling in pitch relative to each other voice (for voice quality): whisp whisper breath breathy husk husky creak creaky fals falsetto reson resonant giggle unvoiced laugh or giggle laugh voiced laugh trem tremulous sob sobbing yawn yawning sigh sighing

A full definition of the sense of the values provided for each feature should be provided in the encoding description section of the text header (see section )

The formal definition of these elements is as follows: ]]> Pointers and Alignment

A major difference between spoken and written texts is the importance of the temporal dimension to the former. As a very simple example, consider the following, first as it might be represented in a playscript: Let us assume that Stig and Lou respond to Jane's question before she has finished asking it --- a fairly normal situation in spontaneous speech. Three things are therefore synchronous: the end of Jane's utterance, Stig's whole utterance and Lou's kinesic. To represent such situations, these Guidelines recommend the use of a mechanism known as an alignment map, which is linked with other parts of the text by the usual SGML id/idref pointer mechanism. One way to represent the simple example above might be as follows: have you read Vanity Fair yes ]]>

The rest of this section, which should be read in conjunction with the more general discussion of alignment at section explains how this mechanism works.

A specialized form of alignment map or timeline is used to coordinate simultaneous phenomena, whether they are utterances by different speakers, utterances and gestures, or gestures and events of other kinds. The map represents a series of points in time, which are then linked or aligned with other elements in the text in one of three ways by pointing from the text to the alignment map by pointing from the alignment map into the text by linking in both directions Examples of each method are given below and in section .

An alignment map is represented by an align element, which consists of a series of loc elements. Each loc represents a synchronization point, and may bear attributes indicating its exact temporal position relative to other points in the alignment in addition to the sequencing implied by its position within the alignment itself.

For example, the following alignment map ]]> represents four points in time, named P1, P2, P6 and P3 (as with all attributes named id in the TEI scheme, the names must be unique within the document but have no other significance). P1 is located absolutely, at 12:20:01:01 BST. P2 is 4.5 seconds later than P2 (i.e. at 12:20:46). P6 is at some unspecified time later than P2 and previous to P3 (this is implied by its position within the timeline, as no attribute values have been specified for it). The fourth point, P3, is 1.5 seconds later than P6.

One or more alignment maps may be specified within a spoken text, to suit the encoder's convenience. If more than one is supplied, the origin attribute may be used on each to specify which other align element it follows. The units attribute indicates the units used for timings given on loc elements contained by the alignment map, except where otherwise specified. To avoid the need to specify times explicitly, the interval attribute may be used to indicate that any contained loc elements are a fixed distance apart from each other, or to indicate that only their sequence is specified.

Elements within a spoken text may be aligned with the times specified by a loc element in three ways, as stated above. To point from the text into the alignment, the identifier of the required alignment point should be supplied as the value of one of the following the start attribute of a timed element the end attribute of a timed element the target attribute of a ptr element within the text

For example, This is my turn ]]> The start of this utterance is aligned with P2 and its end with P3. The transition between the words my and turn occurs at point P6.

To point from the alignment into the text, ptr elements are used within the loc elements of the alignment. The alignment represented by the preceding examples could equally well be represented as follows: ... This is my turn ]]> Here, the whole of the object with identifier U1 (the utterance) has been aligned with two different points, P2 and P3. This is interpreted to mean that the utterance spans at least those two points. Note that since we are now pointing from the alignment map into the text and not vice versa, the ptr element within the utterance has been replaced by an anchor element, the function of which is to provide a name for a location in the text so we can point at it.

The two methods can of course be combined, which eliminates the need for anchors in the text: ... This is my turn ]]>

Temporal alignment at this level of precision is generally appropriate for handling the common case of speaker overlap, and examples are given of its application for that purpose below. The mechanism outlined here may however be used for a variety of other purposes, most notably the alignment of discontinuous segments in syntactic analysis, for examples of which see section . Because any element in the TEI scheme may bear an id attribute, and because the ptr and anchor elements may appear anywhere within a TEI-conformant text, it is possible to align any or all parts of a spoken text using the same basic mechanism outlined here.

The formal definition of these elements is as follows: ]]> Recommended Transcription Practice Speaker Overlap

Speaker overlap should be handled using the alignment mechanism discussed in the previous section (). This allows for any element in the transcription to be aligned with any other, either singly (from the alignment map to the element, or vice versa) or doubly (in both directions). Alignment from the align element into the text is appropriate where the text is already fully marked up and it is not to be altered; alignment from the text to the align element is appropriate where the minimum of additional tagging is required. Double alignment considerably simplifies the task of writing software to handle the alignment, at the expense of considerable density of tagging.

As an example of the three possibilities, consider the following dialogue, represented first as it might appear in a conventional playscript: A commonly used convention might be to transcribe such a passage as follows: I used to smoke [ a lot more than this ] <2> [ you used to smoke ] <1> but I never inhaled the smoke ]]> Such conventions have the drawback that they are hard to generalize or to extend beyond the very simple case presented here. Their reliance on the accidentals of physical layout also make them difficult to transport and very difficult to process. These Guidelines recommend one of the following courses:

Single linkage, text to alignment: .... I used to smoke a lot more than this but I never inhaled the smoke You used to smoke ]]> Note that the second utterance above could equally well be encoded as follows with exactly the same effect: You used to smoke ]]>

Single linkage, alignment to text: .... I used to smoke a lot more than this but I never inhaled the smoke You used to smoke ]]>

Double linkage: .... I used to smoke a lot more than this but I never inhaled the smoke You used to smoke ]]>

Note that in each case, although Bob's utterance follows Tom's sequentially in the text, it is aligned temporally with its middle, without any need to disrupt the normal syntax of the text.

As a further example, consider the following exchange, first as it might be represented using a musical score like notation, in which points of synchronization are represented by vertical alignment of the text All three speakers are simultaneous at the words my, Balderdash and No; speakers A and C are simultaneous at the words turn and it's. This could be encoded as follows, using pointers from the alignment map into the text: ... this is my turn balderdash no it's mine ]]> Word Form

When speech is transcribed into writing, it is customary to use ordinary orthographic notation. This necessarily implies some compromise between the sounds produced and conventional orthography. Particularly when dealing with informal, dialectal or other varieties of language, the transcriber will frequently have to decide whether a particular sound is to be treated as a distinct vocabulary item or not. For example, while kinda is probably not worth distinguishing as a vocabulary item from kind of, isn't is clearly worth distinguishing from is not; for some purposes, the regional variant isnae might also be worth distinguishing in the same way. One rule of thumb might be to allow such variation only where a generally accepted orthographic form exists, for example, in published dictionaries of the language register being encoded; this has the disadvantage that such dictionaries may not exist. Another is to maintain a controlled (but extensible) set of normalized forms for all such words; this has the advantage of enforcing some degree of consistency amongst different transcribers. Occasionally, as for example when transcribing abbreviations or acronyms, it may be felt necessary to depart from conventional spelling to distinguish between cases where the abbreviation is spelled out letter by letter (for example, B B C or V A T) and where it is pronounced as a single word (for example VAT or RADA). Similar considerations might apply to pronunciation of foreign words (for example Monsewer vs Monsieur).

In general, use of punctuation, capitalization etc in spoken transcripts should be carefully controlled. It is important to distinguish the transcriber's intuition as to what the punctuation should be from the marking of prosodic features such as pausing, intonation etc.

Whatever practice is adopted, it is essential that it be clearly and fully documented in the editorial declarations section of the header. It may also be found helpful to include normalized forms of non-conventional spellings within the text, using the regularization element as described in section . Prosody

In the absence of conventional punctuation, the marking of prosodic features becomes of paramount importance, since these structure and organize the spoken message. Pauses have already been dealt with in section ; while tone units (or intonational phrases) can be indicated by the segmentation tag discussed in section . The shift tag discussed in section may also be used to encode some prosodic features, for example where all that is required is the ability to record shifts in voice quality.

For more detailed work, involving a detailed phonological transcription including representation of stress and pitch patterns, it is probably best to maintain the prosodic description in parallel with the conventional written transcription, rather than attempt to embed detailed prosodic information within the written transcription. The two parallel streams may be aligned with each other and with other streams, for example an acoustic encoding, using the general alignment mechanisms discussed in section . For representation of phonemic information, we recommend the use of the International Phonetic Alphabet, as further discussed in section Speech Management

By speech management we mean disfluencies such as filled and unfilled pauses, interrupted or repeated words, corrections and reformulations as well as interactional devices asking for or providing feedback. Depending on the importance attached to such features, transcribers may choose to adopt conventionalized representations for them (as discussed in section above), or transcribe them using IPA or some other phonemic transcription system. To simplify analysis of the lexical features of a speech transcript, it may be felt useful to tidy away in some sense many of these disfluencies. Where this policy has been adopted, these Guidelines recommend the use of the tags for simple editorial intervention discussed in section to make explicit the extent of regularization or normalization performed by the transcriber.

For example, false starts, repetition and truncated words might all be included within a transcription, but marked as editorially deleted, in the following way: ssee you youyou know it'she's crazy ]]>

Similarly, where a transcriber is believed to have incorrectly identified a word, the regularization/non-regularization tags sic and reg may be used to indicate both the original and a regularization of it: skuzzy SCSI ]]> As discussed in section , the first of these would be appropriate where faithfulness to the transcribers' intuition is paramount, and the second where the editorial interpretation is felt more significant. In either case, the user of the text can perceive the basis of the choice being offered. Analytic Coding

The recommendations made here only concern the establishment of a basic text. Where a more sophisticated analysis is needed, more sophisticated methods of markup will also be necessary, requiring the use of concurrent markup streams for multiple segmentation of the stream of discourse, or complex alignment of several segments within it. The general purpose analytic tools discussed in section should be used for such purposes, as they should for representation of structures larger than the individual utterance. As yet, however, we have been unable to find any consensus in the field as to what kinds of units should be identified at this level, and cannot therefore make more specific recommendations. It is hoped however that the basic building blocks offered here will be serviceable in the expression of such a consensus when it is achieved.