4.3 Running a dialogue

The dialogue can be run in one of two modes: Either the standard mode, in which the dialogue is simply executed as specified, or the "wizard" mode, which runs it as it would for a Wizard-of-Oz experiment.

Before a dialogue is run, all the clients the developer wants to attach as devices must be started up. Refer to the documentation of each client for that.

For the dialogue to be started in the standard mode, Wizard - Run from the menu must be selected, or Ctrl-R must be pressed. The dialogue manager will attach the clients as devices, and then execute the dialogue. The current node will start to blink.

For the dialogue to be started in the wizard mode, Wizard - Wizard of
Oz must be selected from the menu, or Shift-Ctrl-R must be pressed. The dialogue manager will attach the devices. Then it will prompt for information to put into the log file for the current user. This can be skipped with the No log button if a log file is not necessary.

After this a window that allows controlling the dialogue pops up. Start starts the dialogue. The wizard will send commands to output devices normally, but whenever the dialogue enters an input node, it ignores the user's input on this device. Instead a button must be pressed to indicate which outgoing edge the dialogue manager should use.

The wizard mode continually updates information about variable values and device input/outputs in the fields on the right-hand side of the window.

The wizard mode is not just useful for real wizard-of-oz experiments. It can be used for a "dry run" of a dialogue that gives more control about what happens, instead of relying on a properly attached and perfectly working recogniser.

Pressing the Stop button, stops the dialogue before it enters an end node.