Abstract
PARROT is a system for paraphrasing simple stories about stereotypical events like getting up in the morning or buying food in a shop. The ideas come from SAM and other Yale School work, but instead of having scripts, PARROT has Knowledge Structures (KSs) implemented in a system consisting of five networks plus an inference network for each KS. The system works in three phases, input, inference, and output. During input, text like 'John eat breakfast' is first converted by a network into four field form. A second network determines which KS action (if any) should 'recognise' the input. Such actions save the input in the weights of a third network.
Original language | English |
---|---|
Pages (from-to) | 318 |
Number of pages | 1 |
Journal | Neural Networks |
Volume | 1 |
Issue number | 1 SUPPL |
DOIs | |
Publication status | Published - 1988 |
Externally published | Yes |
Event | International Neural Network Society 1988 First Annual Meeting - Boston, MA, USA Duration: 6 Sep 1988 → 10 Sep 1988 |