Welcome to annotlib – An Annotator Library

• Quick Start
• Overview
  • Introduction – Why annotlib?
  • Pool-based Active Learning Cycle with Multiple Annotators
  • Modelling Annotators – scikit-learn compatible
  • Annotator Types and Simulation
• Base and Standard Annotators
  • Defintion of Standard Annotators
  • Number of Annotators
  • Class Labels and Certainty Scores
  • Query Statistics and Queried Samples
  • Labelling Performance
  • Confidence Noise
• Clusters as Knowledge Areas of Annotators
  • 1. Class Labels as Clustering
  • 2. Clustering Algorithms to Find Clustering
  • 3. Feature Space as a Single Cluster
• Classifiers as Annotators
  • Varying the Size of the Training Set
  • Varying the Number of the Features Used for Training
  • Varying the Classifier Type and its Parametrisation
• Estimate Sample Difficulty to Simulate Annotators
  • Estimating the Difficulty of Samples
  • Simulating and Analyising Annotators
• Annotators with Dynamic Labelling Performance
  • Dynamic Annotators without Adversarial Labelling Performances
  • Dynamic Annotators with Adversarial Labelling Performances
• Active Learning with Multiple Types of Annotators
  • Experimental Setup
  • Results
• API Reference
  • module: base
  • module: standard
  • module: classifier_based
  • module: cluster_based
  • module: difficulty_based
  • module: dynamic
  • module: multi_types
  • module: utils
• Bibliography