IBM Research undertook a challenge to build a computer system that could compete at the human champion level in real time on the American TV Quiz show, Jeopardy! The extent of the challenge includes fielding a real-time automatic contestant on the show, not merely a laboratory exercise. The Jeopardy! Challenge helped us address requirements that led to the design of the DeepQA architecture and the implementation of Watson. After 3 years of intense research and development by a core team of about 20 researches, Watson is performing at human expert-levels in terms of precision, confidence and speed at the Jeopardy! Quiz show. Our results strongly suggest that DeepQA is an effective and extensible architecture that may be used as a foundation for combining, deploying, evaluating and advancing a wide range of algorithmic techniques to rapidly advance the field of QA.

/pdf/building-watson-an-overview-of-the-deepqa-project-215b8hl6in.pdf

Building Watson: An Overview of the DeepQA Project

: This thesis investigates the problem of constraint-directed reasoning in the job-shop scheduling domain. The job-shop scheduling problem is defined as: selecting a sequence of operations whose execution results in the completion of an order, and assigning times (i.e., start and end times) and resources to each operation. The number of possible schedules grows exponentially with the number of orders, alternative production plans, substitutable resources, and possible times to assign resources and perform operations. The acceptability of a particular schedule depends not only on the availability of alternatives, but on other knowledge such as organizational goals, physical limitations of resources, causal restrictions amongst resources and operations, availability of resources, and preferences amongst alternatives. By viewing the scheduling problem from a constraint-directed search perspective, much of this knowledge can be viewed as constraints on the schedule generation and selection process. In this thesis, we present a system called ISIS. ISIS uses a constraint-directed search paradigm to solve the scheduling problem. ISIS provides: a knowledge representation language (SRL) for modeling organizations and their constraints; hierarchical, constraint-directed scheduling of orders, which includes: constraint-directed bounding of the solution space; context-sensitive selection of constraints, and weighted interpretation of constraints; analytic and generative constraint relaxation; and techniques for the diagnosis of poor schedules.

/pdf/constraint-directed-search-a-case-study-of-job-shop-5gox5qeuan.pdf

Constraint-Directed Search: A Case Study of Job-Shop Scheduling

A survey on question answering technology from an information retrieval perspective

Rule-based expert systems are surveyed. The most important considerations are representation and inference. Rule-based systems make strong assumptions about the representation of knowledge as conditional sentences and about the control of inference in one of three ways. The problem of reasoning with incomplete or inexact information is also discussed, as are several other issues regarding the design of expert systems.

http://i.stanford.edu/pub/cstr/reports/cs/tr/82/926/CS-TR-82-926.pdf

Principles of rule-based expert systems

A command and control system is a technological, procedural, and organizational extension of the sensing, processing, and communicating capabilities of the military commanders whose decisions it supports. However, the increase in battlefield information rate brought about by modern weapons, sensors, and tactics requires selective but extensive application of automation to assist commanders and their staffs in reaching timely and appropriate decisions. The background and context of the problem and the development of a taxonomy of force management decisions in U.S. Air Force tactical command and control is presented. The processing requirements for those decisions are identified along with the types of processing errors and other critical problems in man-computer interaction. Finally, Air Force developments in decision aids for tactical command and control are reviewed, and implications and issues for command and control systems research, development, and design are drawn.

Force Management Decision Requirements for Air Force Tactical Command and Control

A mathematical scientist experiments. Today, his test tube and his breadboard are blackboard and paper. He may, it is true, have available a computer, but its role is numerical and its results are delivered not today, not tomorrow, but the day after the final programming bug is corrected. The computer is not present during the most creative phases of the scientist's labor. The purpose of MATHLAB is to provide the scientist with computational aid of a much more intimate and liberating nature.

MATHLAB: a program for on-line machine assistance in symbolic computations

The natural language database query system incorporated in the KNOBS interactive planning system comprises a dictionary driven parser, APE-II, and script interpreter which yield a conceptual dependency conceptualization as a representation of the meaning of user input. A conceptualization pattern matching production system then determines and executes a procedure for extracting the desired information from the database. In contrast to syntax driven Q-A systems, e.g., those based on ATN parsers, APE-II is driven bottom-up by expectations associated with word meanings. The processing of a query is based on the contents of several knowledge sources including the dictionary entries (partial conceptualizations and their expectations), frames representing conceptual dependency primitives, scripts which contain stereotypical knowledge about planning tasks used to infer states enabling or resulting from actions, and two production system rule bases for the inference of implicit case fillers, and for determining the responsive database search. The goals of this approach, all of which are currently at least partially achieved, include utilizing similar representations for questions with similar meanings but widely varying surface structures, developing a powerful mechanism for the disambiguation of words with multiple meanings and the determination of pronoun referents, answering questions which require inferences to be understood, and interpreting ellipses and ungrammatical utterances.

/pdf/knowledge-based-question-answering-3jqxgfi70x.pdf

Knowledge based question answering

This paper discusses the requirements that interactive frame instantiation imposes on constraint verification. The representations and algorithms of an implemented software solution are presented.

Interactive frame instantiation

KNOBS (from Knowledge Based System), proper, refers to an experimental system under development at the MITRE Corporation to explore the applicability of artificial intelligence techniques to the implementation of an automated, extremely flexible tactical mission planning consultant. This paper will discuss that system and then another which supports the rule based simulation of aircraft identification strategies.

Carl Engelman

Papers

MATHLAB: a program for on-line machine assistance in symbolic computations

Knowledge based question answering

Interactive frame instantiation

KNOBS: an experimental knowledge based tactical air mission planning system and a rule based aircraft identification simulation facility

An integrated frame/rule architecture