Schank, R. & Abelson, R.  (1977).  Scripts, plans, goals, and understanding: An inquiry into human knowledge structure.  Hillsdale, NJ: Lawrence Erlbaum Associates.

By: Soliman Ismail

Overview

This book is a seminal work in the field of artificial intelligence. Its significance lies in the fact that it reports one of the earliest attempts to make use of our understanding of how meaning is represented in the human mind in producing computer programs to understand and/or produce natural language. The authors argue that if they could get these programs to do so, this proves the viability of their theory of the mental representation of language.

The authors start from the premise that knowledge of specific situations is stored in the human mind as scripts. A script is a structure that describes appropriate sequences of events in a particular context. Understanding a familiar situation entitles the activation of a stored script of this situation. When encountered with a new situation, people resort to plans, the mechanisms that underlie scripts. A plan is a repository for general information that will connect events that cannot be connected by the use of an available script or by standard causal chain expansion. The process of plan understanding is to ascertain the goals of the actors in a story and ascertain which particular method is being used to realize each operating goal. Goals, intentions underlying action, can be grouped into themes. Themes contain the background information upon which we base our predictions that an individual will have a certain goal. Themes can be categorized into role themes, interpersonal themes, and life themes.

Arguing that most stories can be analyzed, represented, and understood using this model, the authors describe a number of computer programs written in an attempt to test whether these propositions are valid interpretations of how human beings structure knowledge. Running these programs, the authors admit that in trying to understand and/or produce natural language, these computer programs ran into a number of problems. For example, the computer programs could not handle quantification, anaphora, and metaphors. The authors conclude their book with a chapter with a description of how children develop knowledge structures using a case study approach.

The authors should be credited for bringing to attention that content cannot be separated from form and for emphasizing the importance of context in our processing of knowledge. However, they completely exclude form from their own model of language processing. It is not clear in their model, for example, how a human being uses knowledge of syntactic structures to differentiate between the meaning of sentence pairs like: Flying planes is dangerous and Flying planes are dangerous.

Also, the authors describe the development of knowledge structure as a linear process. They feed the computer with a set of scripts, goals, plans, and/or themes and expect the computer to process this information as a human being does. What the authors do not take into consideration is that knowledge structure development by human beings is a dynamic process in the sense that humans do have mechanisms that enable them to learn from unfamiliar situations and thus generate scripts for these situations in a way that computers do not.

Ch. 1: Introduction

This chapter introduces the book by explaining how psychologists, artificial intelligence researchers, and linguists are interested in researching the nature of knowledge and how this knowledge is used. The authors also introduce their Conceptual Dependency Theory (CD), which they argue explains the representation of meaning of sentences.

Ch. 2: Causal Chains

The authors introduce their theory of Causal Chaining, which they argue explains the representation of meaning of texts. This theory is based on making explicit the implicit connections existing in the text.

Ch. 3: Scripts

This chapter examines how specific knowledge is processed in the form of scripts. The authors elaborate on how scripts can be classified and coded for computer use.

Ch. 4: Plans

This chapter examines how general knowledge is processed in the form of plans. The authors elaborate on how plans can be classified and coded for computer use in the form of planboxes, choices available to the speaker to achieve a goal.

Ch. 5: Goals

This chapter defines goals and illustrates how goals are categorized and coded. It also explains the relationship between goals and themes.

Ch. 6: Themes

This chapter introduces themes and suggests they can be categorized into three types: Role themes, interpersonal themes, and life themes. The authors also show how themes can be coded for identification by the computer programs.

Ch. 7: Representation of Stories

In this chapter, the authors argue that most stories can be analyzed, represented, and understood using the theoretical tools they developed, i.e., means of representing scripts, plans, goals, and themes. They also discuss problems arising from the use of this model of meaning representation in getting computers to process natural language.

Ch. 8: Computer Programs

In this chapter, the authors describe a number of computer programs written based on the theoretical propositions discussed earlier in the book in an attempt to test whether these propositions are valid interpretations of how human beings structure knowledge. The programs described are:

Ch. 9: A Case Study in the Development of Knowledge Structure

The authors illustrate how children acquire and develop a structure of knowledge from a very early age by studying the development of Schank's kids. The authors conclude that information is acquired as well as stored by humans in episode form and that children are active script constructors. They also argue that "the pattern of learning seems to be that first, definitions of objects are learned as episodes. Then, scripts are learned to connect events. Finally, scripts are organized by goal structures that are used to make sense of the need for them."

Web Design by Soliman Ismail