As has been seen in the previous sections, NERC Level IV includes a digitized speech wave together with the symbolic annotation. This will make it possible to use, in corpus linguistics work, the (semi)automatic procedures for aligning orthographic transcriptions with the speech signal, developed in the field of speech technology (see, for example, Andersson & Broman, 1993; Blomberg & Carlson, 1993). However, it must be clear that, at the present stage, automatic alignment systems present certain limitations when applied to spontaneous or conversational speech.
In line with the NERC recommendation, it was suggested during the Madrid workshop `Issues in Corpus Work' organized by the EAGLES Text Corpora Working Group in January 1996 that an alignment between the speech signal and word end-points would be desirable both for speech and for spoken corpora. Of course, this would require the development of adequate TEI mechanisms.
The work carried out in MARSEC (Roach & Arnfield,1995; Knowles,1995; more information is found at URL http://midwich.reading.ac.uk/research/speechlab/marsec/marsec.html) to link the transcriptions ofthe Spoken English Corpus with the acoustic waveform is a recent example of the conversion of a spoken corpus into a segmented and labelled database.