Project Goal: Create a software agent that assists in the task of selecting and placing objects in a room based on a natural language description of a character. The Object Placer: given a list of object ids? lookup object in database (xml file) database entry contains: category id (concept) concept net tuples concept [OftenFoundIn] concept [OftenFoundAbove] [OftenFoundBelow] [OftenFoundInFront] [OftenFoundBehind] [OftenFoundBeside] [OftenFoundBesideLeft] [OftenFoundBesideRight] [OftenFoundNearCenter] [OftenFoundInCorner] [OftenFoundAround] relationship maps relative positioning or [LocationOf] -> parse prepositional phrase and translate into proper function to place object for each concept try to position need a prefabbed room with knowledge about room dimensions and layout (locations of walls) room, walls, floors, doors, stairs -- preplaced weapons, clothes, armor, ingrediants, chest, shelf, books still need to get various object dimensions when placing objects need to check if it can be placed there. map each object to a concept that common sense knows about get a list of possible location concepts possibly parse location into preposition and location reverse map (backward link [isA]) onto list of objects place relative to an object in that list according to preposition is there a way to ask for [locationOf] from OCMSNet? OCMSNet decomposes compound words? a few ways to guess? shortsword [isA] sword sword [isA] weapon chainmail [isA] armor weapon [locationOf] on a weapon rack book [locationOf] on a bookshelf armor [locationOf] in a chest bed [locationOf] against a wall bed [locationOf] corner of room table [locationOf] center of room chair [locationOf] around table