Clarity on summary statistics

[bvreede]

To get to more deliberate summary statistics, this copied from the Project draft:

Tabular summary measures

For tabular summary measures, we provide at least the following key characteristics of the corpus:

• turns: number of annotations with timing information in the corpus, which in most corpora corresponds to the number of turns at talk
• turnduration: mean duration of turns in this corpus
• density: sum of all annotation durations divided by length of source. If >1, indicates a densely annotated recording with quite some overlap. If <0.7, indicates less densely annotated recording and possibly untranscribed parts.
• people: total number of distinct participants encountered in all source records for this corpus
• corpus_length: total number of hours h:m:s of transcribed data (counting from the first transcription until the last by source)

In addition, we provide a table at source level for a number of these measures:

• turns: number of annotations with timing information
• turnduration: mean duration of turns in this source
• density: sum of all annotation durations divided by length of source. If >1, indicates a densely annotated recording with quite some overlap. If <0.7, indicates less densely annotated recording and possibly untranscribed parts.
• people: total number of distinct participants encountered in this source
• source_length: amount of time h:m:s transcribed in this source (counting from the first transcription until the last in this source)