Recording Lifecycle
Conversation Components
Complexity Components
Testing Cycle
Caption Quality
100

This is an umbrella term for all audio recordings in the audio sample library, including raw recordings, validated recordings, modified recordings, and test clips.

What is an audio sample?

100

This is a person who has consented to being recorded.

What is a volunteer?

100

“Punctuated” and “continuous” are two kinds of this.

What is background noise?

100

This is the length of an IP CTS test cycle.

What is 6 months?

100

These are the two thigs that will be measured for captions using the v1 metric.

What are accuracy and delay?

200

This is a complete, unmodified, unvalidated audio recording of a conversation. Test clips are derived from it after they are validated and modified.

What is a raw recording?

200

This is the human or artificial entity that converses with a volunteer in a recorded conversation.

What is a confederate??

200

“Um” and “uh” are two examples of these.

What are disfluencies?

200

This is the percentage of test calls per cycle that are high complexity.

What is 50%?

200

This is the “ground truth” against which captions are compared.

What is a reference transcript?

300

This is a raw recording that has been reviewed and is approved for modification prior to becoming a test clip.

What is a validated recording?

300

This is the purpose for which a conversation took place or the type of interaction that occurred (e.g., filling a prescription, placing a customer service call)

What is a scenario?

300

This is defined as the tempo of speech excluding pauses and hesitations, measured in words per minute.

What is articulation rate?

300

This is the minimum number of test clips used per test cycle.

What is 60?

300

This occurs when a person changes the volume or pitch of their voice to compensate for background noise.


What is the Lombard Effect?

400

This is a validated recording that has undergone modification (such as the addition of background noise) but has not yet been trimmed to form a test clip.

What is a modified recording?

400

This is the overall subject matter of a conversation.

What is domain?

400

This is a commonly used measure of vocabulary complexity.


What is Flesch-Kincaid?

400

This is initiated when a provider disagrees with how a test call has been scored.

What is the challenge process?

400

These three things are tallied when calculating Word Error Rate.

What are substitutions, deletions, and insertions?

500

This is a validated recording that is derived from a raw recording, has undergone modification, and is ready to be used in an IP CTS test call

What is a test clip?

500

This is calculated by multiplying a recording’s Speech and Channel Complexity Score (SCCS) by its intelligibility multiplier.

What is a complexity score?

500

This is the percentage of the recording that is speech, calculated by dividing the total duration of speech (excluding silences) by the total duration of the recording.

What is speaking ratio?

500

This is the number of days within which a results package must be posted after the end of a test cycle.

What is 120?

500

This is the time increment in which caption delay is measured.

What is 0.1 seconds?