Нашли опечатку? Выделите ее мышкой и нажмите Ctrl+Enter
Название: Computer Power and Human Reason: From Judgement to Calculation
Автор: Weizenbaum J.
The computer and natural language is a sub-domain of computer science in which one of the major aims is to imitation of man, focusing on two topics: psychology and linguistics. If we wish the machine to do something, we must tell it what to do and it must be able to understand us. The easiest way to tell a computer what to do is to give it a program to run. "Humans, if they are machines at all, are vastly general-purpose machines and what, is most important, they understand communications couched in natural language." Work must be done for a machine to understand natural language. "Man's capacity to manipulate symbols, his very ability to think, is inextricably interwoven with his linguistic abilities." A machine must be able to extract semantic content from the messages impinged upon it, adopt a syntactic structure of a visual scene and adopt a certain conceptual framework. The question of what comprises a visual symbol is in question. The developer defines the elements of the machines primitive vocabulary. Robert Lindsay said, "high quality translations could be produced by machines supplied with sufficiently detailed syntactic rules, a large dictionary, and sufficient speed to examine the context of ambiguous words for a few word in each direction."
Eliza was a program consisting mainly of general methods for analyzing sentences and sentence fragments, locating so-called keywords in texts, assembling sentences from fragments and so on. Eliza created the remarkable illusion of having understood in the minds of the many people who conversed with it.
In ordinary two person communication, each has a working hypothesis, a conceptual framework, concerning who the person is and what the conversation is about. The hypothesis serves an indicator of what the other person is going to say and what he is going to mean by what he is about to say. Often, the erroneous prediction is falsified before the sentence is completed and the listener makes corrections on the fly and virtually unconsciously. Each brings into mind an image of the other person, the image consists in part of the other's identity, attributes based on evidence derived from independent life experiences of the participant. "Our recognition of another person is thus an act of induction on evidence presented to us partly by him and partly by our reconstruction of the rest of the world; it is a kind of generalization". Eliza starts with the hypothesis that the system does understand.
Rogar C. Shank, based his theory on the central idea that every natural-language utterances is a manifestation, an encoding, of an underlying conceptual structure. Understanding an utterance means encoding it. The theory proposes a formal structure for the conceptual bases for making predictions. The theory creates formal rules for converting utterances into a conceptual base. One difficulty is that every individual's belief is constantly changing mean that an individuals entire base of conceptions is changing. "When a person enters a conversation he bring his belief structure with him as a kind of agenda."
Terry Winograd, of M.I.T, was working with a group were building a computer-controlled "hand-eye" machine; the computer could see its environment and manipulate objects in its environment by means of a computer-controlled mechanical arm. Winograd design and coded the software to enable humans by natural language, too instruct the computer, how to manipulate and explain events with respect to the toy world of blocks, in a natural language. "The robot can manipulate toy blocks on a table containing simple objects like a box." The robot could be ask to manipulate the objects, doing such things as building stacks and putting things in a box. It could be questions about the configuration of blocks on the table, about events that were going during the discussion, and it could be told simple facts about the objects which could be stored and used for reasoning later. The conversation goes on within a dynamic framework - "one in which the computer is an active participant, doing things to change his toy world, and discussing them."