Transcription of a recording

Learn in five minutes

• Wikipedia defines speech analytics as analyzing recorded calls to gather customer information.
• That page says it includes automatic speech recognition and is known for setting the topic against the emotional character of the speech.
• It says applications can spot keywords on live audio or as a later pass on a recording.
• It says word error rate compares a transcription with a reference, and that word error rate is not always what a search user needs.
• A transcript is text. It is not proof of the label someone adds afterward.

Keep the audio, the transcript, and the label in three boxes. A fluent transcript can still contain the wrong word.

This section HTML

Visual

Diagram SVGDiagram PNGDiagram HTML

Flow SVGFlow PNGFlow HTML

Three boxes

1. Box 1: the recording you are allowed to hear.
2. Box 2: the transcript.
3. Box 3: empty, until a later room, for a sentiment or quality label.
4. If you have no reference transcript, do not invent a word error rate.

This section HTML

Quiz

This section HTML

Ecosystem

Transcript

Text from speech. It still needs a check against audio you are allowed to hear.

Sisters: Automatic speech recognition, Sentiment analysis, Quality rubrics, Personal data.

The live voice pipeline answers the person. Analytics reads a recording later. Keep those clocks apart.

A later review can add a rubric you wrote. It cannot borrow a customer's tape.

Config

• Write the goal in one sentence.
• Name the input and the allowed output.
• Save the trace next to the result.

Troubleshooting

• The answer looks fluent and is wrong. Check whether a source passage actually contains the claim.
• The loop never stops. Set a step cap and stop when the tool result repeats.

This section HTML