Crowdee
LT Pipelines

Entity Detection

Extract named entities — persons, organizations, locations, dates, and events — from text or audio files.

Two pipelines cover named entity recognition (NER) — one for text files and one for audio. The audio variant first transcribes the file with Whisper, then runs NER on the transcript. Both return the same output schema.

lt-entity-detection-text

Extracts named entities directly from text and document files using an LLM.

150 credits per file · ~2 minutes

Accepted Files

TypeMIME types
Texttext/*
Documentsapplication/pdf, Word (.docx)

Up to 10 files per run.

Context Parameters

ParameterRequiredDescription
entity_typesNoComma-separated list of entity types to focus on (e.g. "PERSON,ORG,LOC"). If omitted, all supported types are detected.

Supported entity types: PERSON, ORG, LOC, DATE, EVENT, MISC

API Example

POST /v2/lt-pipelines/project/{projectId}/run
{
  "pipelineSlug": "lt-entity-detection-text",
  "fileIds": ["file_abc123"],
  "contextData": {
    "entity_types": "PERSON,ORG"
  }
}

lt-entity-detection-audio

Transcribes the audio file with Whisper, then extracts named entities from the transcript in a second stage. This pipeline runs two stages sequentially.

350 credits per file · ~5 minutes

Accepted Files

TypeMIME types
Audioaudio/* (MP3, WAV, M4A, OGG, FLAC, …)

One file per run. Maximum file size: 25 MB.

Context Parameters

None.

API Example

POST /v2/lt-pipelines/project/{projectId}/run
{
  "pipelineSlug": "lt-entity-detection-audio",
  "fileIds": ["file_abc123"],
  "contextData": {}
}

Output

Both pipelines return the same output schema:

{
  "pipelineSlug": "lt-entity-detection-text",
  "output": {
    "entities": [
      {
        "text": "Angela Merkel",
        "type": "PERSON",
        "start": 12,
        "end": 25,
        "confidence": 0.97
      },
      {
        "text": "Berlin",
        "type": "LOC",
        "start": 42,
        "end": 48,
        "confidence": 0.94
      }
    ],
    "entityCount": 2
  },
  "rawText": null
}

Output Fields

FieldTypeDescription
entitiesarrayAll detected entities
entities[].textstringThe entity text as it appears in the source
entities[].typestringEntity type: PERSON, ORG, LOC, DATE, EVENT, or MISC
entities[].startnumberCharacter offset of the entity start in the source text
entities[].endnumberCharacter offset of the entity end
entities[].confidencenumberDetection confidence score 0–1
entityCountnumberTotal number of detected entities

How is this guide?

© 2026 Crowdee GmbH. All rights reserved.

On this page