# Attribution pilot — speakers-smoke (diarized): what the labels actually say

Read this file, not the metrics, to answer *are the labels right*. Each speaker
is shown with its talk-time share and the verbatim transcript at its longest
segments — the same text, in the same `[HH:MM:SS] line` form, the model saw.

## shondo/OWuZuOrbQ10

**[ASMR] Asking You Lots Of Questions | THE SURVEY [Keyboard Typing] [Softly Spoken]**

00:42:26 · 1 chunk(s) · outcome `attributed`

### Host — 67% of attributed time, confidence 0.90, first seen in chunk 1, present in 1/1 chunk(s)

_00:00:41 – 00:00:49_

```
[00:00:41] You may hear your results being typed out on the computer so that they can be recorded in our online portal.
```

_00:03:47 – 00:03:56_

```
[00:03:47] It is really important to us to record your sincere answers.
[00:03:53] Your answers are really important to us.
```

_00:21:16 – 00:21:24_

```
[00:21:16] Question twenty eight
[00:21:19] standardized testing an accurate measure of a child's intellect
```

### Caller — 18% of attributed time, confidence 0.50, first seen in chunk 1, present in 1/1 chunk(s)

_00:04:44 – 00:04:48_

```
[00:04:44] The option that you choose will be recorded in our online portal.
```

_00:17:28 – 00:17:33_

```
[00:17:32] live inside of it
```

_00:29:26 – 00:29:31_

```
[00:29:26] you may hear your chosen option being typed on the computer
```

### Interviewer — 15% of attributed time, confidence 0.80, first seen in chunk 1, present in 1/1 chunk(s)

_00:36:40 – 00:36:45_

```
[00:36:40] Question three one
[00:36:43] intelligence
[00:36:44] or beauty
```

_00:41:07 – 00:41:12_

```
[00:41:07] there are no more questions in Tony's survey.
```

_00:41:29 – 00:41:34_

```
[00:41:29] thank you so much for volunteering for today's survey. You have answered all of the
```

- [ ] labels are right
- [ ] a label is wrong (which: ____)
- [ ] a name appears that the transcript never states
- [ ] one person is split across two labels
- [ ] two people are merged into one label