LabelRight Get free sample annotation

Services

Business-to-business data annotation for AI start-ups, model developers, computer vision teams and research organisations. Each service shows what we can deliver today.

Available run through our full workflow, upload to export   By arrangement scoped with you first

Text annotation

Labels for language data: from simple categories to judgements on model output.

Text classification and categorisation Available
One or more categories for each text, from your own label set.
Named entity recognition Available
People, organisations, places, products or your own entity types, marked as spans.
Sentiment analysis Available
Positive, neutral, negative, or a finer scale that you define.
Document labelling Available
Type, topic or handling labels for whole documents.
Intent classification Available
The intent behind a message or query, against your intent list.
LLM response evaluation Available
A rating and a short reason for each model response, against your rubric.
Human preference and response ranking Available
Two responses side by side; a person chooses the better one.

Image annotation

Labels for computer vision datasets.

Image classification Available
One category for each image.
Object detection and bounding boxes Available
A labelled box around each object of interest.
Image segmentation By arrangement
Polygon outlines around objects. Pixel-level masks by arrangement.
Image tagging Available
Several descriptive tags for each image.

Audio annotation

Labels and text for speech and sound recordings.

Speech transcription By arrangement
The words spoken, typed out to your style guide.
Speaker identification and diarisation By arrangement
Who speaks when, marked as segments on the recording.
Audio classification By arrangement
A category for each clip.

Dataset preparation

The work around the labels that decides if a dataset is usable.

Cleaning and organising datasets Available
Duplicates, empty records and broken rows found and reported before labelling starts.
Creating annotation guidelines Available
Written rules with examples and edge cases, agreed with you before work starts.
Data formatting and export Available
CSV, JSON, JSONL, and COCO or YOLO for image boxes.
Quality control and annotation review Available
A second look at a sample of the labels, with the error count recorded.
AI-assisted pre-labelling with human verification By arrangement
A model suggests, a person checks every item. Only with your written permission.

What we do not claim

We do not publish accuracy percentages, because a number means nothing without your data and your guidelines. For each project we record how many labels were reviewed and how many needed correction, and we give you those figures with the delivery.

Not sure which service fits?

Describe the data and what the model needs to learn. We will tell you what we can do.