> ## Documentation Index
> Fetch the complete documentation index at: https://quartr.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Backlog transcripts
> Transcripts from events
## Overview
The transcript dataset provides access to high-accuracy transcripts from regular earnings calls and unique events. Each transcript is associated with an event.
Transcripts are only available for events conducted in English.
## How it works
Each transcript record in the API response includes a URL pointing to a JSON file hosted on our CDN. Transcripts are released in two steps:
1. **Raw transcript** (type ID = 15) is published shortly after the event concludes, with paragraph breaks.
2. **Edited transcript** (type ID = 22) follows once speaker identification has been completed, adding speaker names, roles, and company affiliations.
## Chapters
Transcripts can be divided into structured segments using [chapters](/datasets/chapters). Use the [chapters endpoint](/api-reference/transcripts/list-transcript-chapters) to retrieve hierarchical sections with titles and timestamps for a given transcript.
## Speaker identification
The edited transcript (type ID = 22) maps each paragraph to a named speaker with their role and company. The JSON structure is the same as the raw transcript, with an added `speaker_mapping` array at the top level. Older transcripts will not be retroactively updated with speaker data.
If your application filters by `typeId`, include both: `typeIds=15,22`.
It’s not always possible to identify who is speaking. In such cases, the `name`, `role`, and `company` fields may be null. This can happen when:
* The speaker is not clearly identified in the audio.
* The speaker is not part of the event’s official roster.
* We are unable to verify the speaker’s identity or role.
During high-activity periods like earnings season, some events may be prioritized for speaker attribution based on client interest and market relevance. Speaker data may appear on certain transcripts sooner than others.
## Data structure
Transcripts are structured as a hierarchy: the top-level `transcript` object contains the full text and an array of `paragraphs`, each paragraph contains `sentences`, and each sentence contains `words`. Every level includes start/end timestamps in seconds and a zero-based `speaker` index.
```json Standard theme={null}
{
"version": "1.0",
"event_id": 123456,
"company_id": 123,
"transcript": {
"text": "This is the full transcript text",
"number_of_speakers": 3,
"paragraphs": [
{
"text": "This is the paragraph text",
"start": 0,
"end": 10,
"speaker": 0,
"sentences": [
{
"text": "This is the sentence text",
"start": 0,
"end": 5,
"words": [
{
"word": "This",
"punctuated_word": "This",
"start": 0,
"end": 5,
"confidence": 0.9
}
]
}
]
}
]
}
}
```
```json Edited theme={null}
{
"version": "1.0",
"event_id": 123456,
"company_id": 123,
"speaker_mapping": [
{
"speaker": 0,
"speaker_data": {
"name": "Operator",
"role": null,
"company": null
}
},
{
"speaker": 1,
"speaker_data": {
"name": "John Doe",
"role": "CEO",
"company": "ACME Inc."
}
},
{
"speaker": 2,
"speaker_data": {
"name": null,
"role": null,
"company": null
}
}
],
"transcript": {
"text": "This is the full transcript text",
"number_of_speakers": 3,
"paragraphs": [
{
"text": "This is the paragraph text.",
"start": 0,
"end": 10,
"speaker": 0,
"sentences": [
{
"text": "This is the sentence text",
"start": 0,
"end": 5,
"words": [
{
"word": "This",
"punctuated_word": "This",
"start": 0,
"end": 5,
"confidence": 0.9
}
]
}
]
}
]
}
}
```
## How to access this data
Query audio files using company and event filters for full control.
Subscribe to webhooks for real-time updates.
Query the transcripts view directly using SQL.