I am honored to have been invited by the editors of this Festschrift to write an article in honor of Vladimir Lifschitz on the occasion of his 65th birthday. In this article I describe some of the major contributions that Vladimir has made in his exceptional career. I provide background material about Vladimir in the Former Soviet Union (FSU), where he was born; discuss his beginning years in the United States (US); how I became aware of his research; his work on stable models with Michael Gelfond, and some of my work in related areas; and describe some of the many contributions Vladimir has made in formalizing ad hoc approaches in artificial intelligence to a formal basis of logic-based analysis.

The Maryland Refutation Proof Procedure Systems (MRPPS) is an interactil perimental system intended for studying deductive search methods. Although the work is oriented towards question answering, MRPPS provides a general problem solving capability. There are three major components within MRPPS. These are: (1) an inference system, (2) a search strategy, and S (3) a base clause selection strategy. The inference system is based on the resolution principle and performs the logical deductions specified. The user may select from a wide variety of refinements of resolution. Current refinements are: set of support, linear, P1, SL, input, and combinations of the above. Paramodulation and deletion by tautologies and subsumption are also provided with the system. The search strategy directs the deductions to be made by selecting from clauses already generated those that have. the best merit. The merit of a clause is given by f(n) = wog(n) + wihi(n) + w2h2(n) + + wkhk(n). If the user can specify tie-breaking rules for equal values of clause merit, an upwards diagonal search results in the sense of Kowalski. The upwards diagonal search included in MRPPS generalizes the Kowalski upwards diagonal search to an n-dimensional search. The base clause selection strategy determines which facts and general axioms to select from the data base. Such a clause may be selected regardless of whether or not it has the best merit. Heuristic techniques are applied within each of the three major components. This technical report describes the current implementation of MRPPS. It describes each of the components and how they are integrated into what has been termed the Q* algorithm. HIPPS is written in FORTRAN V for the UNIVAC 1108 (a version of FORTRAN IV) and runs under EXEC 8 at the University of Maryland. The current imp'mentation is core bound and requires approximately 60K words of memory to run, of which 35K is for the data base and for working storage.

Michael Gelfond is one of the world leading scientists in the field of logic programming and nonmonotonic reasoning. This essay covers several aspects of Michael’s personal life, starting from his birth in the USSR, through his experiences in the USSR up to the time he emigrated to the United States (U.S.). This is followed by his first experiences in the U.S.: how he became involved in logic programming and nonmonotonic reasoning and some of his major scientific achievements. Michael is a warm, generous person, and I discuss his impact on some colleagues and students. In the concluding section, I observe that starting his career with impediments in the FSU, he overcame them to become one of the top computer scientists in logic programming and nonmonotonic reasoning.

Knowledge bases (KBs) must be able to capture a wide range of situations. One must be able to represent and answer questions regarding indefinite information where it is not clear that there is a unique answer to a question. One must also represent and answer questions about negative information. We discuss a powerful way to represent such information, namely through reasoning about knowledge bases using logic.

341 deened relations. Whenever base relations change, we may utilize the algorithm to compute the tuples that have to be deleted from and inserted into the deened relations, instead of recomputing the deenitional expressions from scratch. Another application area is integrity constraint enforcement. Suppose we have a constraint (8x)P(x) expressed in rst-order logic. The fact that (8x)P(x) is true can be equally represented as fxj:P(x)g = ;. By the equivalence of relational calculus and relational algebra, we can nd a relational expression E equivalent to fxj:P(x)g (assuming of course that :P is safe). Whenever the constrained relations change, we just need to ensure that 4E is empty. As a consequence, updates that only contribute to 5E have no eeect on the validity of the constraint. Notice however that the orthogonal issue discussed above becomes extremely important here, because we want to compute 4E before transaction execution in order to avoid the expensive aborting operation. There is growing interest in more advanced database management systems, such as object-oriented databases and active databases. One important feature of these systems is the ability to respond to changes in user-speciied situations, by either notifying users or triggering user-supplied procedures. When the underlying databases are relational, the monitoring situations are best expressed as relational expressions. The eeciency requirement for their evaluation is usually high, especially for real-time systems. By building our algorithm into these monitoring facilities, signiicant performance enhancement can be expected. In summary, the incremental recomputation of active re-lational expressions is essential for many database function-alities. We presented a complete set of propagation rules and an eecient incremental recomputation algorithm that derives, by update propagation, the minimal incremental relational expressions that need recomputation. Compared with existing approaches, we separated the problem of in-cremental recomputation of arbitrary relational expressions from the issue of computing incremental base relations. Our algorithm uniformly applies to all expressions of rela-tional algebra in the transformational framework. Moreover , the algorithm does not rely on the availability of transactions, but information about transactions can be easily incorporated for further optimization. Acknowledgment The authors would like to thank Tore Risch and Peter Ladkin for valuable comments and helpful discussions.

The authors acknowledge Mr. James Albus and Mr. Sam Rosenfeld of the National Aeronautics and Space Administration for their support of this work. We express our appreciation to Mr. Daniel Fishman of the University of Maryland for participating in the early efforts and for valuable discussions. We wish to thank Dr. Cordell Green of Stanford University for reading an early draft of the work and for providing many suggestions for improvements.

This article describes the first six years of the Computer Science Department, established in 1973 at the University of Maryland. The department evolved out of the Computer Science Center, which had been instituted in February 1962. In 1980, the National Academy of Sciences judged the department as being among the leading computer science departments in the US.

This paper provides an overview of the cooperative answering strategies used in the CARMINcooperative answering system, an ongoing research effort at Maryland. Section 2 gives some usefulbackground definitions. Section 3 describes techniques for collecting cooperative logical formulae.Section 4 discusses which natural language generation techniques are useful for presenting the logicformula in natural language text. Section 5 presents a diagram of the system.

The use of and first courses on computers at Maryland began in the 1940s. The computer science center beginnings are examined in the context of founding director Werner C. Rheinboldt's noteworthy accomplishments to develop a computation facility with research and educational components, supported by John Menard, assistant director. William F. Atchison, who replaced Rheinboldt as Center director, continued these accomplishments

The field of cooperative answering goes back to work started by Joshi and Webber [12] in natural language processing in the early 1980s at the University of Pennsylvania. The work was applied to databases and information systems at the University of Pennsylvania by Kaplan [14, 15] and Mays [17]. Other early work at the University of Pennsylvania and at other universities is discussed in [25, 13, 26, 16, 18, 24]. Databases and knowledge base systems are often difficult to use because they do not attempt to cooperate with their users. A database or a knowledge base query system provides literal answers to queries posed to them. Such answers to queries may not always be the best answers. Instead, an answer with extra or alternative information may be more useful and less misleading to a user. This lecture surveys foundational work that has been done toward developing database and knowledge base systems with the ability to exhibit cooperative behavior. In the 1970s, Grice [11] proposed maxims of cooperative conversation. These maxims provide the starting point for the field of cooperative answering. To develop a general system for data and knowledge bases, it is important to specify both the sources of information needed to provide cooperative behaviors, and what constitutes cooperative behavior. Several sources of knowledge apply: the basic knowledge in a system is given by explicit data, referred to as data in a relational database, or facts (extensional data) in a deductive database; by general rules that permit new relations (or new predicates) to be developed from existing data, referred to as views in relational databases and as intensional data in deductive databases; and by integrity constraints that must be consistent with the extensional and intensional data. Integrity constraints may be obtained either through the user or the database administrator, or by a data mining capability. Whereas integrity constraints must be consistent with every instance of the database schema, another source of knowledge is state constraints, which may apply to the current state, and need not apply to a subsequent state upon update. Two additional sources of knowledge arise from information about the users. One describes the class of the user, for example, an engineer or a child, each of whom expect different kinds of answers to queries, and user constraints that must be satisfied. User constraints need not be consistent with the database, but reflect the interests, preferences and desires of the user. Alternative cooperative behaviors are explained and illustrated by examples. The cooperative behaviors discussed are: A brief description is provided of three systems that have been implemented, which exhibit cooperative behavior for relational and deductive databases. The systems and their features are: For a discussion of the state-of-the-art of cooperative answering systems, see A functional description is provided of a cooperative database system. It is currently possible to develop a cooperative capability that interfaces with any of the existing database systems. Since proposed and future versions of relational databases will include capabilities to handle recursion and semantic query optimization, the ability to include cooperative capabilities will, in the not too distant future, be incorporated into such systems.

We extend the well-founded semantics of Van Gelder et al. [12] so that it is able to reason through clauses instead of single literals. This is necessary to infer p to be true in the program {p ← a; p ← b; a ← ¬b; b ← ¬a}. We call this generalized semantics, the generalized well-founded semantics. We present fixpoint and model theoretic definitions for generalized well-founded semantics and show their equivalence.

At a workshop held in Toulouse, France in 1977, Gallaire, Minker and Nicolas stated that logic and databases was a field in its own right (see [131]). This was the first time that this designation was made. The impetus for this started approximately twenty years ago in 1976 when I visited Gallaire and Nicolas in Toulouse, France, which culminated in a workshop held in Toulouse, France in 1977. It is appropriate, then to provide an assessment as to what has been achieved in the twenty years since the field started as a distinct discipline. In this retrospective I shall review developments that have taken place in the field, assess the contributions that have been made, consider the status of implementations of deductive databases and discuss the future of work in this area.

Database systems can be difficult to use. Part of the problem is that systems do not, for the most part, help a user when a query fails or fails to evaluate as the user expects. Schema and semantics of databases are often complex, and are rarely understood in entirety by the lay user. As a consequence, queries a lay user casts may not make sense with respect to a database's semantics. A system which returns informative responses beyond a query's answer set itself can elucidate the schema and semantics of the database, which can greatly help the user to cast the queries intended. Such a database system is to be called a cooperative database system (CDBS). A number of cooperative behaviors and methods have been introduced to be incorporated into information systems to make them more informative and, hence, easier to use. We have identified a fundamental set of such cooperative techniques that we believe relational and deductive systems should be extended to include. We present an architecture for such a CDBS which supports these, and which uses existing relational technologies. We have built the Carmin system based on the proposed architecture. Carmin is intended both as a research platform for cooperative answering and as a practical, efficient CDBS. The paper presents the architecture of Carmin, and, by example, a general architectural approach for CDBSs. The architectural components of Carmin are outlined and discussed in turn. The implementation issues and cost overhead of Carmin are discussed, and the architectural choices are argued.

Semantic query optimization uses semantic knowledge in databases (represented in the form of integrity constraints) to rewrite queries and logic programs to achieve efficient query evaluation. Much work has been done to develop various techniques for optimization. Most of it, however, is applicable to top-down query evaluation strategies. Moreover, little attention has been paid to the cost of the optimization. We address the issue of semantic query optimization for bottom-up query evaluation strategies with an emphasis on overall efficiency. We focus on a single optimization technique, join elimination. We discuss factors that influence the cost of semantic optimization, and present two different abstract algorithms for optimization. The first pre-processes a query statically before it is evaluated; the second combines query evaluation with semantic optimization using heuristics to achieve the largest possible savings.

Consider the construction of an expert system by encoding the knowledge of different experts. Suppose the knowledge provided by each expert is encoded into a knowledge base. Then the process of combining the knowledge of these different experts is an important and non-trivial problem. We study this problem here when the expert systems are considered to be first order theories. We present techniques for resolving inconsistencies in such knowledge bases. We also provide algorithms for implementing these techniques.