Artificial Intelligence and Automatic Scripting - Text 3
The poem’s structure ensures a kind of poetic coherence despite the instability of its constituent elements. Computer technology would quickly take up this sort of approach, as computers excel in manipulating combinations.
In the field of artificial intelligence, research into manipulating natural language has drawn on the work of structural linguists (in particular Noam Chomsky) to model linguistic and narrative production; this takes the name “non-contextual grammars.”[3] Like Queneau’s poems, these are syntactical structures of which some terms can be replaced recursively by others. This approach makes it possible, easily and quickly, to generate structurally great quantities of structurally coherent but not necessarily semantically coherent variants. Like Queneau’s poems, the different elements are selected independently and have a strong tendency to produce incoherent and even contradictory discourses. Recently, Kate Compton’s software program Tracery[4] has popularized the use of non-contextual grammars for creative purposes, and in particular for the automatic generation of blather, a short and autonomous format which lends itself well to the unbridled and inconsistent creativity of this kind of system.
Story and Narration
Non-contextual grammars amalgamate the production of a story and its textual narration. This means that the story told does not pre-exist its being put into words and is the product of a chance recursive selection. Research soon saw these two levels as distinct problems, however. Attempts were made to produce stories out of characters, motivations, objects, actions, events, etc., independently of their discursive representation. The advantage of such separation is that it is easier for a computer to “think about” abstract entities than it is the text. In this context, systems inspired by non-contextual grammars have recently been developed precisely for the final “putting into narrative” once the story has been conceived by other means.[5].
Automatic Planning
One of the main approaches for generating coherent stories is automatic planning. In a system, this consists in finding a sequence of actions which make it possible to achieve an objective, for example determining the optimal route for getting from point A to point B, taking into account the schedules and routes of a city’s buses. In the mid-1970s, Jim Meehan developed Tale-Spin, a system which generated brief stories inspired by Aesop’s fables.[6] In these stories, characters try to accomplish objectives by planning certain actions based on their perception of a simply modelled world.
