Acquisition and Maintenance of Text-based Plans

Text-based plans are plans whose steps and goals are described in a textual format. In contrast to logicbased plans plans that are represented based on variations of a first-order logic format such plans are much easier for a domain expert to describe, and is getting much easier to find given the tremendous growth of the Internet. In this paper we assume that the user is interested in an interactive plan retrieval system that is based on a library of text-based plans. We focus on two important issues on the maintenance of this library: the updating of the text-based plans themselves and the maintenance of the indexes with which to retrieve these plans. We present techniques that address the life-time maintenance of such plan libraries with techniques drawing from information retrieval, neural network learning and case based reasoning.