Data annotation
Data annotation is the process of attaching labels, target values, or other structured descriptions or judgments to data examples.
Machine Learning System Design
The process of assigning annotations or target values to data examples for machine learning.
- How it is used
- Supplies target labels for ML datasets while aligning annotation detail with the intended product behavior.
- How deep it goes
- Works through label granularity and a labeling-workflow checklist, but defers detailed weak-supervision methods to further reading.
- Examples
- Uses malignancy labels on medical images and relevance labels for query–image pairs.
- What you can do after
- Practice/perform annotation planning by specifying labels and sources, comparing staffing options, and preparing tasks and quality checks.
Read how Machine Learning System Design explains data annotation →
LLMs in Production
Data annotation is the process of attaching labels or metadata to data instances, often with awareness of their positions or spans.
- How it is used
- Distinguishes optional labels from the data preparation still needed for self-supervised LLM training.
- How deep it goes
- Explains why annotations may be absent, then gives high-level preparation steps without detailing annotation workflows or data pipelines.
- Examples
- Uses large Parquet or text dumps with no labels or annotations as the concrete case.
- What you can do after
- Evaluate/understand whether missing annotations matter for an LLM training task without mistaking them for proof of data quality.
Deep Learning with R, Third Edition
Data annotation is the process of assigning human-provided labels or other metadata to data examples.
- How it is used
- Supplies targets for supervised learning and guides decisions about collecting, labeling, and preparing training data.
- How deep it goes
- Explains annotation purpose, labor, quality effects, sourcing options, expertise requirements, and noise tradeoffs without detailing labeling procedures.
- Examples
- Uses sentiment-labeled movie reviews, cat-versus-dog images, dog breeds, and CT scans of bone fractures.
- What you can do after
- Evaluate annotation needs and choose labelers according to task complexity, expertise, cost, scale, control, and noise.
Read how Deep Learning with R, Third Edition explains data annotation →
Deep Learning with Python, Third Edition
The process of having people manually label or correct dataset examples.
- How it is used
- Connects supervised-learning inputs to target labels and treats annotation as a core dataset-collection task.
- How deep it goes
- Extends from manual labels and provider tradeoffs to expert correction in a semiautomated segmentation pipeline.
- Examples
- Uses cat-versus-dog labels, dog-breed classification, fracture scans, and SA-1B object masks as annotation examples.
- What you can do after
- Apply it by selecting suitable labelers and weighing expertise, cost, scalability, control, and label noise.
Read how Deep Learning with Python, Third Edition explains data annotation →
Effective Conversational AI
Data annotation is the process of assigning labels or judgments to data examples for training, testing, or evaluation.
- How it is used
- Measures and improves a classifier’s understanding of real user utterances for training, testing, or evaluation.
- How deep it goes
- Explains filtering, labeling, collaboration, disagreement resolution, and prediction review, but stops short of implementation details.
- Examples
- Uses customer-service utterances labeled Request_Agent or Office_Hours and pairs flight-baggage or refund questions with expected responses.
- What you can do after
- Diagnose/debug with it by producing golden labels, resolving ambiguous intents, and constructing representative test data.
Read how Effective Conversational AI explains data annotation →
Machine Learning Platform Engineering
An annotation is metadata attached to an input example that identifies or describes a target property for machine learning.
- How it is used
- Uses annotations to prepare labeled examples for training and evaluating machine-learning models.
- How deep it goes
- Explains modality-specific labeling, dataset splitting, and data-versioning implications without detailing annotator quality controls or tool internals.
- Examples
- Demonstrates OCR bounding boxes, handwritten ID-number transcription, fraud labels, and annotated identity cards in CVAT.
- What you can do after
- Apply annotation through CVAT and organize labeled examples for model training, validation, and testing.
Read how Machine Learning Platform Engineering explains data annotation →
Outlier Detection in Python
Labeling is the process of assigning annotations or target values to data records.
- How it is used
- Uses human judgments to classify flagged records, guide investigation, and improve future outlier detection.
- How deep it goes
- Explains binary, categorical, and scored labels plus downstream uses, but stops short of a detailed annotation workflow.
- Examples
- Uses financial transactions, unusual staff, sales combinations, and rare times of day as annotation examples.
- What you can do after
- Apply it to categorize flagged records, prioritize investigation, retune detectors, and train downstream rules or models.
Read how Outlier Detection in Python explains data annotation →
Designing Deep Learning Systems: A guide for software engineers
The process of assigning annotations or target labels to data examples for supervised machine learning.
- How it is used
- Supports data exploration by turning collected examples into datasets for later model research and prototyping.
- How deep it goes
- Places labeling in an informal early workflow and stops short of specifying annotation standards, quality controls, or a formal collection pipeline.
- Examples
- Uses public or internal data, outsourced or in-house labeling, Python or shell scripts, manual copying, and web-based analysis tools.
- What you can do after
- Build it from scratch by collecting, labeling, and assembling datasets for algorithm evaluation and proof-of-concept models.
Read how Designing Deep Learning Systems: A guide for software engineers explains data annotation →
The Art of AI Product Development
Data annotation is the process of adding labels or other structured annotations to raw data.
- How it is used
- Supports supervised learning by converting raw data into learning signals for classification, prediction, and few-shot prompting.
- How deep it goes
- Explains label sources, quality checks, human risks, and automation, but stops short of detailed annotation-tool implementation.
- Examples
- Uses sentiment labels, animal-image labels, next-word pairs, user-segment tables, and retrieved question-answering examples.
- What you can do after
- Apply it by designing labeling guidelines, checking agreement and class balance, and selecting human or automated annotation.
Read how The Art of AI Product Development explains data annotation →
Building Reliable AI Systems
The process of adding human-provided labels, rankings, or other judgments to data.
- How it is used
- Supplies human labels or judgments when fine-tuning data needs document classes or ranked response quality.
- How deep it goes
- Explains when manual labels and ratings are needed, but stops short of a labeling protocol or reviewer-agreement method.
- Examples
- Uses legal-document issue tags and human ratings or annotations of response politeness and thoroughness as examples.
- What you can do after
- Practice/perform it by tagging domain documents or judging outputs against a quality goal, while favoring accurate, domain-reviewed examples.
Read how Building Reliable AI Systems explains data annotation →
Related concepts
- labelproduces · A label is a value or annotation associated with an item that identifies its target category, outcome, or reference answer for learning or evaluation.
- content-based recommenderused by · A content-based recommender suggests items whose features resemble those of items matching a user’s interests or prior interactions.
- training datasetproduces · A collection of examples or observations supplied to a learning procedure to fit or adapt a model or learn its patterns or parameters.
- learning signalproduces · A learning signal is information or feedback that indicates what a model should learn or how it should adjust its behavior or parameters.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.