A conversational system which understands short stories
Gérard Sabah · 1978
We present a conversational system which can learn short stories in a natural language and perform some reasoning about them. In our model, a sentence is represented by a purely semantic case structure (deep structure) with embedding properties. The vocabulary is a tree of word representative elements, each of which having its sense precised by some semantic features (pointers to other words). To understand a sentence within a given story, we reduce the information it carries, This reduction gives way to the application of a few rules for character behaviour, whose aim is to connect the sentence to another onp by establishing some causal relationship between desires and actions. The system can then answer why questions and questions about some facts not explicitly described in the story. The natural language used is French, the program is written in SIMULA