tailieunhanh - Báo cáo khoa học: "DESIGN DIMENSIONS FOR NON-NORMATIVE ONDERSTARDING SYSTEMS"

This position paper is not based upon direct experience with the design and implementation of a "non-normative" natural language system, but rather draws upon o u r work on c a s c a d e [11] architectures for understanding systems in which syntactic, semantic and discourse processes cooperate to determine the "best" interpretation of an utterance in a given discourse context. The RUS and PSI-KLONE systems [I, 2, 3, 4, 5], which embody some of these principles, provide a strong framework for the development of non-normatlve systems as illustrated by the work of Sondhelmer and Welschedel [8, 9,. | DESIGN DIMENSIONS FOR NON-NORMATIVE UNDERSTANDING SYSTEMS Robert J. Bobrow Madeleine Bates Bolt Beranek and Newman Inc. 10 Moulton Street Cambridge Massachusetts 02238 1. Introduction This position paper Is not based upon direct experience with the design and implementation of a non-normative natural language system but rather draws upon our work on cascade 11 architectures for understanding systems in which syntactic semantic and discourse processes cooperate to determine the best interpretation of an utterance in a given discourse context. The RUS and PSI-KLONE systems 1 2 3. 4. 5 which embody some of these principles provide a strong framework for the development of non-normatlve systems as illustrated by the work of Sondheimer and Weisohedel 8 9 10land others. Here we pose a number of questions In order to clarify the theoretical and practical issues involved In building non-normatlve natural language systems. We give brief indications of the range of plausible answers In order to characterize the space of decisions that must be made in designing such a system. The first questions cover what is Intended by the Ill-defined term non-normative system beyond the Important but vague desire for a friendly and flexible computer system. The remaining questions cover several of the architectural issues involved in building such a system including the categories of knowledge to be represented In the system the static modularization of these knowledge sources and the dynamic information and control flow among these modules. The way the system Is to deal with ill-formed Input depends in a strong way on how much the system is expected to do with well-formed Input. Ad hoc data base retrieval systems a currently hot topic pose different constraints than systems that are expected to enter Into a substantial dialogue with the user. When the behavior of the system Is severely limited even given perfect Input the space of plausible Inputs is also limited and the search for a .

• Báo cáo khoa học
• DESIGN DIMENSIONS
• NON NORMATIVE ONDERSTARDING SYSTEMS
• Robert J
• Bobrow
• scientific reports
• model language
• process natural language
• Lecture Organization theory and design
• Bài giảng Lý thuyết tổ chức và thiết kế
• Organization theory and design
• Dimensions of organization design
• Structural dimensions
• Contextual dimensions
• Lecture Product Design and Development
• Product design
• Product development
• Product development capability
• Changing dimensions of competition
• Data Warehouses
• Integrity Constraints
• Physical Design
• Logical Design
• Dimensions
• Materialized Views
• Database systems
• Distributed database systems
• Principles of distributed database systems
• Cơ sở dữ liệu phân tán
• Design problem
• Dimensions of the problem
• Age groups
• Hand dimensions
• Gender friendly
• Gender friendly equipments
• Design of gender friendly equipments
• Female agricultural workers
• taxation system
• Tax benefits
• Types of entities
• Importance of treaties
• Investment through
• equity investments
• Operations management
• Competitive advantage
• Operations strategy
• Operations competitive dimensions
• Strategy design process
• Productivity measures
• Reasonable link length
• Kinematic problem
• Manufacturing tolerance
• Open chain kinematic robot
• Robot hardware
• Preform design
• Bevel Gear
• Forging process
• Non dominated sorting genetic algorithm II
• Multi criterion