We present and discuss problems in creating a lemmatised index to transcriptions of Bulgarian speech, including the prerequisites for such an index, and why we consider an index preferable to a search engine for this particular kind of text.