03Service

Text annotation

Structured labels for NLP, search and chatbot training.

We label text in English and Filipino, from short chat messages to long documents. Guidelines define every label with examples and counter-examples, so ambiguous cases are handled the same way across the team.

Request a Quote See it labeled

Annotation types: 5 types
Delivered as: JSONL, CSV or spaCy format
Quality levels: Attempt → Review → Final Review
Starts with: A paid pilot of 500–1,000 items

01Showcase

See it labeled.

One message, three kinds of labels.

Support chat · message 3 of 12EN

Hi, can you move my delivery of two office chairsPRODUCT to FridayDATE? The last order to Quezon CityLOCATION arrived damaged.

3 entities · PRODUCT, DATE, LOCATION

02Annotation types

What we label.

Each type is defined in your guidelines before the pilot, with examples of the edge cases.

Character-level spans for people, organisations, products, places, dates and your own entity types.

Topic, category or policy labels for each document or message.

Positive, negative or neutral, or a finer scale, for the whole text or per aspect.

The user's goal in each utterance, plus the values a system needs to act on it.

Links between entities, such as who works where or which product a review mentions.

03What you receive

Labels your pipeline can load as-is.

Delivered as JSONL, CSV or spaCy format. Tell us your schema and we'll match it.

Labels in the format your pipeline already reads
A quality report per batch: accuracy on hidden test items against your target
Every task through Attempt, Review and Final Review
A 7-day window to accept each batch or raise a dispute

JSONLExample

{"text": "Maria Santos ordered two Samsung monitors for the Cebu office.",
 "entities": [[0, 12, "PERSON"], [21, 24, "QTY"], [25, 41, "PRODUCT"], [50, 54, "LOCATION"]],
 "review": "final_review:approved"}

04Use cases

Where teams use it.

01
Chatbots and assistants
02
Document processing
03
Content moderation
04
Search relevance

05Quality

Every task passes three levels.

Only tasks approved at Final Review are paid to the annotator and delivered to you.

Level 1Annotator
Attempt
Labels the item to the project guidelines and label schema.
Level 2Reviewer
Review
Checks the work, then approves it, corrects it, or sends it back with a reason.
Level 3Final reviewer
Final Review
Signs the task off. Only finally approved tasks are paid and delivered.