Definition of terms

SAMPLE

A digital representation of the analogue waveform at any one instant in time

SAMPLING RATE

The (normally constant) rate at which SAMPLEs are collected and transferred to a storage medium.

SAMPLING INTERVAL

The time between successive SAMPLES - fixed for any one SAMPLING RATE while the sampling process is not stopped.

DISCONTINUITY

The point in a stream or stored set of SAMPLES where the sampling process has been stopped for longer than one SAMPLING INTERVAL.
(Explanation - it is equivalent to the splice edit point on the tape of an analogue tape recorder.)

SEGMENT

A set of sequential SAMPLES held on the storage medium which does not contain a DISCONTINUITY. A SEGMENT will be bounded by two DISCONTINUITIES
(Explanation - this means that during the period over which a SEGMENT has been collected the process of sampling and storing has not been stopped.)

SPEECH FILE

A set of sequential SAMPLEs held on the storage medium. The SPEECH FILE will contain one or more SEGMENTS.
(Explanation - the sampling process may have been stopped, then started again during the period over which a SPEECH FILE has been collected.)

TAKE

The time over which a subject is invited to complete the production of a specific set of utterances. The TAKE will be recorded in one or more SEGMENTS. A SPEECH FILE will often (but not always) contain one TAKE.
(Explanation - the start of a single TAKE will often be identified by the issuing of a set of instructions to the subject and/or the inputting into a prompting system of the utterance set ID. It will normally be terminated by the prompting system finishing an utterance list and further operator-subject dialogue. There will normally be no operator-subject dialogue within a TAKE.

SESSION

The period of time from when a specific subject arrives at the recording location to when he leaves. During this time one or more TAKEs will be recorded.

SPEECH CORPUS

A set of SPEECH FILES.

SPEECH DATABASE

One or more SPEECH CORPORA complete with the information to allow the CORPORA to be usefully accessed. This information will usually be in the form of time-aligned annotation, speaker descriptions, recording condition files, etc.