The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #92123

Title:Computing Action Plans from Logical Representations in a Natural Language Front-End
Date: 1992
Abstract:MASQUE is a Natural Language Front-End which uses queries expressed in a subset of English for accessing information about some domain, stored in a database. Spenceley's MASQUE is an extended version of MASQUE which covers imperative sentences in addition to questions. This thesis describes how workable plans can be constructed from logical representations produced for imperatives, by integrating a STRIPS-like planner within the query answering module of Spenceley's MASQUE. It is explored how the logical representations for imperatives can be transformed into Goals for the planner, while keeping all the essential information encoded in them and especially in the quantifiers involved in them. It also described a way of keeping uninstantiated variables into the Goals which increases the flexibility of the planner. Within the planner there is implemented a special filter mechanism for detecting invalid Goals and for instantiating properly the uninstantiated Goals. Finally two ways are explored for preventing the planner from seeking plans for already satisfied Goals.

[Search These Pages] [DAI Home Page] [Comment]