Detailed Information on Publication Record

The paper summarizes my experience with the acquisition of audio recordings of unofficial communication situations and with their transcription for the purposes of the Czech National Corpus: for the creation of corpora of spoken Czech. We made audio recordings of more than 220 dialogue situations together with the students and we processed their one-level conversion into written form. We had to solve a number of practical problems on that occasion. These difficulties led us to important theoretical questions. We have resigned ourselves to noting the beginning and end of sentences and clauses as they are usual in the written form of speech, for example. It was necessary to deal with the notation of communicatively relevant hesitant sounds (agreeable, indefinite, negative, etc. hmm, emm, eee...) and with phonetic-phonological deformations of words, which are the result of transposition, especially formations directed between particles and interjections (člověče, čoveče, čoeče, čéče). In all of this, the user processability of the transcribed language material was to be ensured, i.e. automatic machine recognition and provision of the material with meta-linguistic data, as well as the subsequent traceability of possible expression forms. Another area of knowledge concerns the perceptual and cognitive limits of audio recording processors. We find that editors write what they think they hear, what can or should be heard. They are more or less subject to the tendency to emphasize the irregularities of spoken language against regular phenomena. Or, on the contrary, they overlook innovations and promote established forms of written language. The same was essentially true of sentence boundaries as long as they were recorded. During the conversion of spoken language into written form, parole is systematized and idealized. But what we uncover, however, is the relativity of the units of spoken language. In my experience, the editors of the probes pay the necessary attention to the recording of the spoken speech. Vague boundaries of speech units, complicated limits of perception of reproduced speech are reflected in the fluctuations of the graphic forms he chooses. Each user of corpora of spoken communication will have to reckon with these limitations. Written corpora can be a good corrective to the spectrum of chosen expression variants. The transcript can at least be a guide to a thorough examination of the audio recording. The tendency to limit the variability of the forms in the first level of transcription is enforced in the new methodology. On the contrary, the variability of graphic forms is not a problem either. It indicates the intuitive perception of the editor about the variant function of the language unit. If students become more receptive to language as a result of this work as editors of probes, the originally unintended meaning of the existence of corpora will also be fulfilled. The result of our observations is the corpus creation methodology, which has been continuously improved since then. The contribution provides an overview of the basic questions and tries to provide theoretical answers as well as a solution methodology.