This document presents guidelines for the best practice in the production of speech corpora. It may be used as an introductory reading for newcomers to the field, or as a reference and check list for the experienced scientist or engineer.
The document does not cover the basic knowledge about Digital Speech Processing nor does it go into details of highly specialized topics like the SLP applications mentioned above.
Furthermore, this cookbook does not cover validation techniques for speech corpora. Please refer to  for a detailed discussion of this topic.
Finally, although a speech corpus very often contains not only the acoustic signals but also other measurable time signals derived from the process of speaking, throughout this book we will only treat the recording of speech in its various aspects. An exhaustive description of all possible signal recordings with regard to the act of speaking would be far beyond the scope of this cookbook.
The cookbook is organized in such way that a prospective producer of a new speech corpus may follow it like a recipe. In most cases (and if the reader is already familiar with the basic aspects of speech corpora) it will be sufficient to check out the summary check lists at the end of each chapter of part II or summarized in appendix . Since this cookbook describes speech corpus production step by step, many topics are discussed in more than one chapter, e.g. the recruitment of speakers: it is essential to plan the means by which speakers are recruited and to estimate the costs in the planing phase (chapter Corpus Specifications), then to prepare the recruitment (chapter Preparation of Collection) and finally to perform the recruitment during the collection phase (chapter Collection).
For a systematic and non-chronological description of speech corpus production refer to the excellent summary in the EAGLES Handbook (), chapters 3 and 4.