Spoken texts Basic structure: spoken texts The spoken material transcribed for the BNC is also organized into texts, which are subdivided into divisions, made up of w and mw elements grouped into s elements in the same way as written texts. However there a number of other elements specific to spoken texts, and their hierarchic organization is naturally not the same as that of written texts. For this reason, a different element (stext) is used to represent a spoken text. In demographically sampled spoken texts, each distinct conversation recorded by a given respondent is treated as a distinct div element. All the conversations from a single respondent are then grouped together to form a single stext element. Context-governed spoken texts, however, do not use the div element: the stext element for a context governed text is composed only of u elements, not grouped into any unit smaller than the stext itself. The s elements making up a spoken text are grouped not into p or other similar elements, but instead into u elements. Each u (utterance) element marks a stretch of uninterrupted speech from a given speaker; (see section ). Interspersed within and between u elements, a variety of other elements indicate para-linguistic phenomena noticed by the transcribers (see section ). The methods and principles applied in transcription and normalisation of speech are defined in a BNC working paper TGCW21 Spoken Corpus Transcription Guide, and have also been described in subsequent publications (e.g. Crowdy 1994). The editorial tags discussed in section above are also used to represent normalisation practice when dealing with transcribed speech. Utterances The term utterance is used in the BNC to refer to a continuous stretch of speech produced by one participant in a conversation, or by a group of participants. Structurally, the corresponding element behaves in a similar way to the p element in a written text — it groups a sequence of s elements together. The who attribute is required on every u: its function is to identify the person or group of people making the utterance, using the unique code defined for that person in the appropriate section of the header. A simple example follows: Mm mm. The code PS1LW used here will be specified as the value for the xml:id attribute of some person element within the header of the text from which this example is taken. Where the speaker cannot be confidently identified, or where there is more than one aspeaker, a special code is used; see further discussion at . Paralinguistic phenomena In transcribing spoken language, it is necessary to select from the possibly very large set of distinct paralinguistic phenomena which might be of interest. In the texts transcribed for the BNC, encoders were instructed to mark the following such phenomena: voice quality

for example, whispering, laughing, etc., both as discrete events and as changes in voice quality affecting passages within an utterance.

non-verbal but vocalised sounds

for example, coughs, humming noises etc.

non-verbal and non-vocal events

for example passing lorries, animal noises, and other matters considered worthy of note.

significant pauses

silence, within or between utterances, longer than was judged normal for the speaker or speakers.

unclear passages

whole utterances or passages within them which were inaudible or incomprehensible for a variety of reasons.

speech management phenomena

for example truncation, false starts, and correction.

overlap

points at which more than one speaker was active.

Other aspects of spoken texts are not explicitly recorded in the encoding, although their headers contain considerable amounts of situational and participant information. In many cases, because no standardized set of descriptions was predefined, transcribers gave very widely differing accounts of the same phenomena. An attempt has however been made to normalize the descriptions for some of these elements in the BNC XML editions. The elements used to mark these phenomena are listed below in alphabetical order: The value of the dur attribute is normally specified only if it is greater than 5 seconds, and its accuracy is only approximate. With the exception of the trunc element, which is a special case of the editorial tags discussed in section above, all of these elements are empty, and may appear anywhere within a transcription. The following example shows an event, several pauses and a patch of unclear speech: You gotta Radio Two with that. Bloody pirate station wouldn't you? Where the whole of an utterance is unclear, that is, where no speech has actually been transcribed, the unclear element is used on its own, with an optional who attribute to indicate who is speaking, if this is identifiable. For example: .... ... Here YY's remarks, whatever they are, are too unclear to be transcribed, and so no u element is provided. The values used for the desc attribute of the event element are not constrained in the current version of the corpus, and more than a thousand different values exist in the corpus. Some common examples follow: A list of the most frequent values is given in . As noted above, a distinction is made between discrete vocal events, such as laughter, and changes in voice quality, such as words which are spoken in a laughing tone. The former are encoded using the vocal element, as in the following example: , you'll have to take that off there yeah you can The shift element is used instead where the laughter indicates a change in voice quality, as in the following example: And erm and then we went and got my fruit and veg and then we went in Top Marks and got them so we never got we went through for a video really, never got round to looking for a video did we? When no new attribute is supplied on a shift element, the meaning is that the voice quality indicated reverts to a normal or unmarked state. Hence, in this example, the passage between the tags shift new="laughing"/ and shift/ is spoken with a laughing intonation A list of values currently used for the new attribute is given below in section . Alignment of overlapping speech By default it is assumed that the events represented in a transcription are non-overlapping and that they are transcribed in temporal sequence. That is, unless otherwise specified, it is implied that the end of one utterance precedes the start of the next following it in the text, perhaps with an interposed pause element. Where this is not the case, the following element is used: The with attribute of an align element may be thought of as identifying some point in time. Where two or more align elements specify the same value for this attribute, their locations are assumed to be synchronised. The following example demonstrates how this mechanism is used to indicate that one speaker's attempt to take the floor has been unsuccessful: Poor old Luxembourg's beaten. You you've you've absolutely just gone straight over it I haven't. and forgotten the poor little country. This encoding is the CDIF equivalent of what might be presented in a conventional playscript as follows: W0001: Poor old Luxembourg's beaten. You, you've, you've absolutely just gone straight over it -- W0014: (interrupting) I haven't. W0001: (at the same time) and forgotten the poor little country.