Annotation templates

A template is a pre-built XML labeling configuration that covers a common annotation scenario. Pick one when you create a project; it gives you a working labeling interface in seconds.

Templates are the fastest path from a fresh project to your first labeled task. Customise from there.

Where to find them

When you create a project:

1. + Create Project → fill in name + workspace
2. Click Labeling Setup
3. The left panel lists the template groups in two sections: Annotate data (Computer Vision, Natural Language Processing, Audio/Speech, Sign Language Annotation, Conversational AI, Chat, Ranking & Scoring, Structured Data, Time Series, Field Data, Videos, Generative AI, Community) for labeling data you import, and Collect data (Voice Collection, Sign Language Collection, Photo Collection) for gathering new recordings and photos. On a phone the same sections appear in the category dropdown.
4. Click a template card → preview opens on the right
5. Click Use this template

You can also browse the gallery without creating a project at Project → Settings → Labeling Interface → Templates.

Built-in templates

Templates ship in groups (the left-panel headings in the picker). A representative sample of each — the full list is in the picker:

Natural Language Processing

Text Classification · Named Entity Recognition · Relation Extraction · Question Answering · Text Summarization · Machine Translation · Aspect-Based Sentiment (ABSA) · Emotion Classification (multi-label) · Stance Detection · Content Moderation · Taxonomy · Intent + Slot Filling (NLU) · Code-Switching (Language ID per span) · Named Entity Linking (KB-backed) · NER — Amharic (Ge'ez script) · NER — Arabic (RTL) · NER — Hausa / Fulfulde / Kanuri Ajami (RTL, Arabic script) · NER — Tamazight (Tifinagh script)

Computer Vision

Image Classification · Object Detection with Bounding Boxes · Semantic Segmentation with Polygons · Semantic Segmentation with Masks · Keypoint Labeling · Optical Character Recognition · OCR Labeling for PDFs · Multi-page document annotation · Image Captioning · Visual Question Answering · Inventory Tracking · Medical Image Classification with Bounding Boxes · Visual Genome

Audio/Speech Processing

Automatic Speech Recognition · Automatic Speech Recognition using Segments · Speech Transcription · Speaker Segmentation · Sound Event Detection · Signal Quality Detection · Intent Classification · Conversational Analysis

Voice Collection (AfriAnnotate-specific, <AudioRecord> + Audio QC)

Read aloud (voice corpus) · Read aloud + self-rating · Read-aloud with speech-text alignment · Listen and repeat · Pronunciation drill (single word / phrase) · Prompted question (open-ended speech) · Picture description (spontaneous speech) · Spontaneous speech (long-form monologue) · Interview response (audio question + audio answer) · Code-switch elicitation · Translation to speech

Sign Language Collection / Sign Language Annotation

Sign this prompt (video corpus) · Sign a picture (video corpus) · Prompted question (spontaneous signing) — and for annotation: Gloss and translate (sign video) · Multi-tier annotation (sign video)

Photo Collection (AfriAnnotate-specific, <ImageRecord>)

Crop health photos · Water point survey · Signs and text in local languages · Document pages

Conversational AI

Intent Classification and Slot Filling · Coreference Resolution & Entity Linking · Response Generation · Response Selection

Chat

Chatbot Evaluation · Evaluate Production Conversations for RLHF · Red-Teaming in Chat · Fine-Tune an Agent with an LLM · Fine-Tune an Agent without an LLM

Generative AI

Human Preference collection for RLHF · LLM Response Grading · LLM Ranker · Visual Ranker · Chatbot Model Assessment · Supervised Language Model Fine-tuning

Ranking & Scoring

Pairwise classification · Pairwise regression · Search Page Ranking · Document Retrieval · Content-based Image Retrieval · ASR Hypotheses Selection · Text-to-Image Generation

Structured Data Parsing

Tabular Data · HTML Entity Recognition · PDF Classification · Freeform Metadata

Time Series Analysis

Activity Recognition · Change Point Detection · Outliers & Anomaly Detection · Signal Quality · Time Series Forecasting

Field Data

Object detection in field imagery · Land cover segmentation (satellite / drone) · Field record triage & classification · Sensor time-series event labeling

Videos

Video Classification · Video Frame Classification · Video Object Tracking · Video Timeline Segmentation

Community Contributions

HTML NER Tagging · NER Tagging for Invoices (BIO Format) · OCR Invoices Pre-NER BIO Format · Two-Level Sentiment Analysis of X / Twitter posts · Breast Cancer Mammogram Classification

Picking a template

A few common decisions:

You have…	Pick…
A list of sentences + a category set	Text Classification
A corpus + person/place/organisation labels	Named Entity Recognition
Images + you want to identify objects	Object Detection with Bounding Boxes
Images + pixel-perfect masks	Semantic Segmentation with Masks
Audio + you need transcriptions	Automatic Speech Recognition
Read-aloud corpus collection (TTS / ASR training)	Read aloud (voice corpus)
Photos taken in the field (crops, water points, signs, documents)	a Photo Collection template
Comparing two model outputs (RLHF)	Human Preference collection for RLHF or Pairwise classification
Time-bucketed data labels	Activity Recognition or Change Point Detection

Customising a template

Every template is a starter — you almost always want to tweak. Two ways:

Visual editor

After picking a template:

1. Project → Settings → Labeling Interface → Visual
2. The form lets you rename labels, change colours, toggle settings (e.g. "Select text by words" vs "by character")
3. Save

The visual editor is best for label + colour changes. For deeper edits, use the code editor.

Code editor

Same page → click Code. You see the underlying XML:

<View>
  <Text name="text" value="$text"/>
  <Choices name="sentiment" toName="text" choice="single">
    <Choice value="Positive"/>
    <Choice value="Neutral"/>
    <Choice value="Negative"/>
  </Choices>
</View>

Edit the XML directly. The right pane previews changes in real-time. See Tag reference for the full XML grammar.

Save as a new template

If you've customised a config that other projects might benefit from:

1. Project → Settings → Labeling Interface → ⋮ menu → Save as template
2. Name + describe it
3. Save

In the save dialog pick who can use it: your organisation (default), specific organisations you also belong to, or — for platform staff only — platform-wide (every org). Saved templates copy the labeling config plus instructions onto new projects created from them.

Template gallery

Browse the full library (built-in + your org's custom templates) at Project → Settings → Labeling Interface → Templates — same gallery the create-project wizard uses.

The built-in templates use the same XML formats as upstream Label Studio, so any upstream template works unmodified — paste its config into the Code view.

What's next

• Labeling config overview → — XML grammar + visual editor walkthrough
• Tag reference → — every tag, its attributes, and worked examples
• Project setup → — full project create flow (the template picker sits inside this)