Programming grounded language games step by step

Suppose the goal is to use grounded language to control a model railroad. The first step is to invent a dictionary with useful words:

- locomotive_1: red, locomotive_2: blue
- switch_1: left_side, switch_2: right_side, track,
- slowdown, speedup, switch, follow, wait

These words are stored in a python dictionary as a list. The first iteration of the parser takes a command from the command line e.g. "switch_1" and searches in the dictionary if the found is available. Then the parser returns "ok".

In step 2 the vocabulary gets connected with the visual appearance of the game formalized in a pointing game. The user enters a command and the parser should highlight the object on the screen. For example the user enters "locomotive_2" and the parser draws a rectangle around the object.

In step 3 which is more advanced an instruction following game gets established. The parser has to execute actions. The user might enter a command like "speedup locomotive_1" and the parser ensures that the desired action gets executed. Programming such a behavior is the most advanced part of a language game.

In general grounded language starts always with a vocabulary which is a word list. The list contains of nouns, verbs and objects and is related to a domain. Symbol grounding in the strict sense means to play language games with the word list which are the pointing game and the instruction following game.