Story grammars - Narrative Intelligence - page 200

(Colby 1973). Colby was the first to use formal grammars to describe linguistic
phenomenon beyond single sentences.
A variety of questions motivated researchers to employ formal grammars
as a way of describing stories. Some were interested in the cognitive mecha-
nisms used by people to summarize and recall stories and proposed that story
grammars were an integral part of human language ability. Others were at-
tempting to discern the common structure of stories and turned to formal
grammars as a knowledge representation device. Our own interest is closer
to the latter; and we do not make any cognitive claims concerning what hu-
man beings do when creating or reading a story. We use a formal grammar to
describe narratives; to this end, we have developed a set of structural compo-
nents along with rules for their composition. Our model is (1) general enough
to apply to folklore compilations of the sort described above, and (2) suf-
ficiently detailed to rule out constructions of non-stories. Below, we briefly
review previous work in story modeling.

Previous work in Declarative Story Modeling

Rumelhart develops a model for the organization that takes place in connected
discourse but is absent in strings of sentences (Rumelhart 1975). In general,
it is almost always necessary to infer (unstated) causal relationships to under-
stand groups of sentences. These causal relationships relate sentences to each
other. Rumelhart presents a grammar describing the inter-sentence bindings
that arise in simple stories. The grammar is context-free and consists of syn-
tactic rewrite rules each of which has a corresponding semantic interpretation
rule. The primitives are meta-sentence components such as setting, episode,
and event. Below are two rules from David Rumelhart's story grammar.

1. Attempt → Plan + Application

⇒ MOTIVATE ( Plan , Application )

2. Application → (Preaction)* + Action + Consequence

⇒ ALLOW ( AND ( Preaction , Preaction , ...),

{CAUSE | INITIATE | ALLOW} (Action, Consequence ))

In Rumelhart's grammar (and those derived from his grammar), the relation-
ship a component has to other components is expressed in “semantic” anno-
tations accompanying the “syntactic” rules. Scare quotes distinguish the story
grammar use of the terms “syntactic” and “semantic” from conventional use.
In story grammars, the terms are intended to mean something like “structural”