Douglas B. Moran

Gemini: a natural language system for spoken-language understanding

Gemini is a natural language understanding system developed for spoken language applications. This paper describes the details of the system, and includes relevant measurements of size, efficiency, and performance of each of its sub-components in detail.

Synergistic use of direct manipulation and natural language

This paper shows how the integration of natural language with direct manipulation produces a multimodal interface that overcomes limitations of these techniques when used separately. Natural language helps direct manipulation in being able to specify objects and actions by description, while direct manipulation enables users to learn which objects and actions are available in the system. Furthermore, graphical rendering and manipulation of context provides a partial solution to difficult problems of natural language anaphora.

INTERLEAVING SYNTAX AND SEMANTICS IN AN EFFICIENT BOTTOM-UP PARSER

We describe an efficient bottom-up parser that interleaves syntactic and semantic structure building. Two techniques are presented for reducing search by reducing local ambiguity: Limited left context constraints are used to reduce local syntactic ambiguity, and deferred sortal-constraint application is used to reduce local semantic ambiguity. We experimentally evaluate these techniques, and show dramatic reductions in both number of chart edges and total parsing time. The robust processing capabilities of the parser are demonstrated in its use in improving the accuracy of a speech recognizer.

QUANTIFIER SLOPING IN THE SRI CORE LANGUAGE ENGINE

An algorithm for generating the possible quantifier scopings for a sentence, in order of preference, is outlined. The scoping assigned to a quantifier is determined by its interactions with other quantifiers, modals, negation, and certain syntactic-constituent boundaries. When a potential scoping is logically equivalent to another, the less preferred one is discarded.The relative scoping preferences of the individual quantifiers are not embedded in the algorithm, but are specified by a set of rules. Many of the rules presented here have appeared in the linguistics literature and have been used in various natural language processing systems. However, the co-ordination of these rules and the resulting coverage represents a significant contribution. Because experimental data on human quantifier-scoping preferences are still fragmentary, we chose to design a system in which the set of preference rules could be easily modified and expanded.The algorithm described has been implemented in Prolog as part of a larger natural language processing system. Extensions of this algorithm are in progress.

GEMINI: A NATURAL LANGUAGE SYSTEM FOR SPOKEN-LANGUAGE UNDERSTANDING

The demands on a natural language understanding system used for spoken language differ somewhat from the demands of text processing. For processing spoken language, there is a tension between the system being as robust as necessary, and as constrained as possible. The robust system will a t tempt to find as sensible an interpretation as possible, even in the presence of performance errors by the speaker, or recognition errors by the speech recognizer. In contrast, in order to provide language constraints to a speech recognizer, a system should be able to detect that a recognized string is not a sentence of English, and disprefer that recognition hypothesis from the speech recognizer. If the coupling is to be tight, with parsing and recognition interleaved, then the parser should be able to enforce as many constraints as possible for partial utterances. The approach taken in Gemini is to tightly constrain language recognition to limit overgeneration, but to extend the language analysis to recognize certain characteristic patterns of spoken utterances (but not generally thought of as part of grammar) and to recognize specific types of performance errors by the speaker.

SRI's experience with the ATIS evaluation

SRI International participated in the June 1990 Air Travel Information System (ATIS) natural-language evaluation. This report briefly describes the system that SRI used in the evaluation, analyzes SRIs results, and makes some recommendations for changes in the database structure and data collection system to be used for future ATIS evaluations.

An architecture for a multimedia teleconferencing system

We present an object-oriented architecture for a computer-based, real-time, multimedia conferencing system. This architecture divides the system into five functional areas: a multimedia shared workspace, a user interface, conference management, communications, and an information base. The structure and operation of the first four areas are modeled with object-based concepts that address design requirements identified during the development of a proof-of-concept prototype, that preceded the architecture specification. The shared workspace, the most important component, is thoroughly discussed; the other components support its realization. The modeling of workspace entities emphasizes their aggregation into composed entities and the homogeneous handling of several data media. The user interface manages uniformly the man-machine interaction for both local and remote user actions. Conference management deals with session establishment, participation, and control of multiple media floors. The communication functions replicate user actions over workspace objects in all hosts participating in a conference, matching traffic types with transmission services in the process.

Multimodal user interfaces in the Open Agent Architecture

The design and development of the Open Agent Architecture (OAA)3 system has focused on providing access to agent-based applications through an intelligent, cooperative, distributed, and multimodal agent-based user interface. Only the primary user interface agents need run on the local computer, thereby simplifying the task of using a range of applications from a variety of platforms, especially low-powered computers. An important consideration in the design of the OAA was to facilitate the reuse of agents in new and unanticipated applications, and to support rapid prototyping. The utility of the agents and tools developed has been demonstrated by their use as infrastructure in unrelated projects.