Transcription tools, transcription conventions and the TEI guidelines for transcriptions of speech

(Thomas Schmidt, Andreas Witt)

The large majority of TEI related work has been concerned with various forms of written language. In our paper, we would like to draw attention to the TEI-based markup of spoken language (Witt 1998), which is important for many research questions in linguistics and other humanities disciplines, and which we believe to be a rewarding topic for the TEI community.

The relevant chapter 8 on ‘Transcriptions of Speech’, one might argue, is confined to a marginal position in the TEI guidelines, that is, the TEI’s recommendations for the encoding of spoken language are not as widely known and used as those parts which deal with the encoding of written language. In our opinion, the reasons for this are to be found not so much in the guidelines themselves as in their relationship to two external factors:

  1. First, transcription as a very labour intensive process depends strongly on specialised tools, the most widely used of which work with a multi-layer, time-based conception of data which is not easy to consolidate with the TEI’s view of transcriptions as “texts of a special kind” (Burnard 1995). Consequently, few of these tools cater for TEI compliance, and interoperability is achieved through other means.
  2. Second, transcription conventions, as devised by conversation and discourse analysts, are still formulated as layout-oriented instructions for the representation of transcribed utterances on paper. The principle of the separation of form from content, fundamental to the application of markup languages and the computer-assisted processing of language data in general, has not yet found its way into these fields of research. Consequently, the benefits of the TEI’s approach to “reduce a great deal of the variety in the transcription of speech [by] focussing [not on display, but] on an underlying representation” (Johansson 1995) are still poorly understood and hardly appreciated by researchers working with transcriptions.

We present a proposal for establishing a link between time-based data models (Schmidt 2005) and the TEI’s guidelines for transcriptions speech on the one hand and between layout-oriented transcription-conventions and the TEI’s content-oriented approach on the other hand. Based on the observation that all transcription systems share global structural properties, but differ with respect to local semantic distinctions, we suggest an architecture in which different transcription conventions can be used within one and the same tool to produce TEI compliant transcription data. We will demonstrate an actual implementation of this architecture and make clear what benefits can be derived from using a TEI conformant representation of transcriptions of speech. We will also make a suggestion for a revision of chapter 8 of the guidelines which should improve its utility and usability for spoken language researchers and developers of transcription tools.

References

  • Burnard, Lou (1995): The Text Encoding Initiative: an overview. In: Leech et al. (eds.): Spoken English on Computer: Transcription, Markup and Application. Harlow: Longman, 69-81.
  • Johansson, Stig (1995): The approach of the Text Encoding Initiative to the encoding of spoken discourse. In: Leech et al. Leech et al. (eds.): Spoken English on Computer: Transcription, Markup and Application. Harlow: Longman, 82-98.
  • Schmidt, Thomas (2005): Time-based data models and the Text Encoding Initiative’s Guidelines for Transcriptions of Speech. Working papers in Multilingualism 62. Hamburg.