首页> 外文会议>International Joint Conference on Artificial Intelligence >CONCEPTUAL OVERLAYS: A MECHANISM FOR THE INTERPRETATION OF SENTENCE MEANING IN CONTEXT
【24h】

CONCEPTUAL OVERLAYS: A MECHANISM FOR THE INTERPRETATION OF SENTENCE MEANING IN CONTEXT

机译:概念覆盖:在语境中解释句子的一种机制

获取原文
获取外文期刊封面目录资料

摘要

This paper describes a context mechanism f o r a natural language understanding system. Since no sentence is ever perceived outside some context, it is reasonable to inquire into the nature of context as it affects the inter predation of sentence meaning at a deep conceptual level. A theory, called conceptual overlays, is described. This theory (1) defines C (T_1,..., T_i), the context established by the meaningful sequence of thoughts T_1 T_j: (2) defines I(T_(j+1), C (T_1,... T_j)), the high-level interpretation of T_(i+i) in the context established by T_i,..., T_i; and (3) specifies an effective algorithm and data structure for computing I (t, K) for arbitrary thought T in context K. In particular, a prototype LISP system, EX-SPECTRE-1, which solves simple cases of I (T_2.,C(T_1)) is described. The system is based on an expectancy/fulfillment paradigm. Expectancies are spontaneously activated by a pattern directed invocation technique. Each expectancy implicitly references large chunks of common - sense algorithms. A collection of such implicitly activated algorithms constitutes context, and the interpretive process is one of identifying future input as steps in these algorithms. Context switching and uses of I (T, K) in a language comprehension system are discussed.
机译:本文介绍了一种基于自然语言理解系统的上下文机制。由于在某些背景之外没有被认为没有句子,因此可以合理地询问背景的性质,因为它影响了深入概念层面的句子意义的缺席。描述了一种称为概念叠加层的理论。这个理论(1)定义C(t_1,...,t_i),由有意义的想法序列建立的上下文t_1 t_j:(2)定义I(t_(j + 1),c(t_1,... t_j )),T_I,......,T_I建立的上下文中T_(i + i)的高级解释; (3)规定了用于在上下文K中计算I(t,k)的有效算法和数据结构,特别是k的任意思想t.特别地,解决了I的原型LISP系统,其解决了I(T_2的简单情况)。描述C(T_1))描述。该系统基于期望/履行范例。预期通过模式定向调用技术自发激活。每个期望都隐含地引用了大块的常见义算法。这种隐式激活的算法的集合构成了上下文,并且解释过程是将未来输入标识到这些算法中的步骤之一。讨论了语言理解系统中I(T,K)的上下文切换和使用。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号