Event Tagging and Entity Extraction
Event Tagging is the process of marking specific temporal segments where a meaningful event, action, or sound occurs within time-series data (audio, video, or sensor data). This task is focused on determining the start time and end time of an event within a longer continuous stream.
Event Tagging enables detailed understanding of what happens when and who is involved in multimedia content.
Ideal for: Smart surveillance, sports analysis, content moderation, industrial monitoring, security systems, and any scenario requiring temporal event localization.
¶ When to Use
Use Event Tagging when your data is time-series (audio, video, or sensor data) and you need to mark specific temporal segments where a meaningful event, action, or sound occurs.
¶ Audio Event Tagging Tasks (Sound Recognition)
Ideal for identifying and localizing specific non-speech sounds within an audio stream.
| Input Type | Question Example | Purpose / Requirement |
|---|---|---|
| Audio | “Tag the start and end times of every instance of a ‘dog barking’ or ‘fire alarm’ in this recording.” | Smart Surveillance/Monitoring: Developing systems to alert based on specific acoustic events (e.g., security, home monitoring). |
| Audio | “Mark the precise segment where ‘music begins’ and ‘music ends’ in this podcast.” | Content Moderation/Editing: Automatically identifying copyright material or facilitating content segmentation. |
| Audio | “Identify and tag the location (time) of abnormal machinery sounds (e.g., ‘grinding,’ ‘squealing’).” | Industrial Monitoring: Predictive maintenance by detecting early acoustic signs of equipment failure. |
¶ Video/Temporal Event Tagging Tasks (Action & Activity Recognition)
Use this for video streams to mark the specific duration of actions or states, or to tag temporal events in time-series data.
| Input Type | Question Example | Purpose / Requirement |
|---|---|---|
| Video | “Mark the exact time range when the basketball player is ‘dribbling,’ ‘shooting,’ and ‘scoring.’” | Sports Analysis: Creating a time-coded log of actions for performance review and highlight generation. |
| Video | “Tag the segment where the subject ‘opens the door’ or ‘places a package on the porch.’” | Security & Activity Monitoring: Training AI to recognize key actions in surveillance footage. |
| Sensor Data (Time Series) | “Tag the time segment where the device’s temperature readings exceed 90°C.” | IoT/System Analysis: Localizing abnormal conditions or critical phases in system logs or sensor feeds. |
¶ Key Features
- Temporal Precision - Mark exact start and end times for events
- Multi-Event Support - Tag multiple overlapping or sequential events
- Action Recognition - Identify specific actions and activities in video
- Sound Recognition - Detect and classify audio events and anomalies
¶ When to Use Other Work Types
- Need speech-to-text? → Transcription
- Need object locations? → Object Detection
- Need simple categorization? → Classification
- Need text entity extraction? → Named Entity Recognition
¶ Supported Data Types
- Video - MP4, AVI, MOV, and other video formats
- Audio - MP3, WAV, FLAC for audio-only event tagging
¶ Getting Started
- Define your event types and temporal requirements
- Specify precision requirements for start/end times
- Prepare your video or audio data
- Submit batches via API
- Track progress and download timestamped event annotations
¶ How to Submit Batches
Indirect Demand (CSV File) - Best for large-scale projects with thousands of video or audio files.
Direct Demand (Inline Data) - Best for dynamic tasks or real-time processing.
¶ Best Practices
- Clear Event Definitions - Define precise criteria for what constitutes each event type
- Temporal Boundaries - Specify how to handle event start/end ambiguity
- Overlapping Events - Provide guidelines for handling simultaneous or overlapping events
- Context Guidelines - Document edge cases and unusual scenarios
- Quality Control - Review sample annotations to ensure temporal accuracy and consistency