Annotation Protocol

In short

The protocol is composed of states.

State syntax:

<state-name>:
    type: <task-type>
    transitions:
        - <trigger1>:
              target: <target1-state-name>
        - <trigger2>:
              target: <target2-state-name>
    <option-key>: <option>

Components:

• state-name
  The identifier for the state. When running flask init-db or flask reset-annotations, for all state names except 'start' a column is created to save the annotations in.
• task-type
  The name of the task type. Refer to Task-Types for a list and documentation of all possible task types.
• transitions When a state has options, we need to define the target-state for each one. For example, if the annotator chooses the trigger1 option HUMAN will transition into the state target1-state-name. If the annotator, however, chooses the trigger2 option, the target-state is target2-state-name.
  • trigger
    The name of the transition: E.g. "YES" or "NO" for boolean task type. "*" is the wildcard transition name and will always transition to its target.
  • target-state-name
    The name of the state that should be transitioned into when triggered.
• options
  Options for the task type. Refer to Task-Types for the possible options for each task.

Requirements:

• A start state that loads a datatype from the server
• A transition with target 'end' to save annotations
• No duplicate states
• No state named 'end'
• For further requirements of a specific task-type refer to its documentation.