From W3C Wiki
Jump to: navigation, search

Cyc Tech Reports (recovered from archive.org)

A few old papers on AI, KR, commonsense reasoning etc., that aren't available from the Cyc site currently. These are mostly multi-page HTML docs generated via latex2html. This means that some papers may be incomplete, if only a subset of urls have been archived.

  • ACT-CYC-134-90 Cyc: A Mid-Term Report, R. V. Guha and Douglas B. Lenat, MCC and Stanford Unversity (Wed Oct 12 16:25:57 CDT 1994)
  • CYC-050-92 Microtheories: An Ontological Engineer's Guide, Paul Blair R. V. Guha Wanda Pratt, 3/92, MCC technical report number CYC-050-92 Abstract: This document is an introduction to the use of contexts in Cyc. After basic introductory concepts, the role and uses of domain assumptions and lifting rules are discussed. A final section deals with some special uses of contexts for dealing with perspectives, indefinite references, modalities, and polysemeity. More details regarding contexts may be found in the document Contexts: A formalization and some applications, by R.V. Guha, MCC Technical Report Number ACT-CYC-423-91, November 1991 (here referred to as Contexts).
  • ACT-CYC-108-90 Cyc: Toward Programs With Common Sense, Lenat, D. B., R. V. Guha, K. Pittman, D. Pratt, & M. Shepherd. (1994) published as http://www.cyc.com/tech-reports/act-cyc-108-90/act-cyc-108-90.html
  • ACT-CYC-154-90 The CycL Representation Language - Part 4, R.V.Guha and D.Lenat, 6/90, MCC technical report number ACT-CYC-129-90. (Wed Oct 12 16:31:41 CDT 1994). Abstract: "CycL, the representation language underlying CYC, has seen numerous new features and improvements in the past few years [and months. Herein we document four additional features: (1) Using iterative deepening to Get answers to queries incrementally rather than all at once; (2) automating the generation of code for performing the bookkeeping associated with newly defined inference mechanisms; (3) frame axioms for temporal projection; and (4) a new abstraction of time, involving HoldsDuring, which can be used almost as a shorthand for many of the most common sentences which are expressable using the more general SubAbstractions formalism. Most of the paper deals with the first topic, incremental Getting, as that bears on the very core of inferencing in CycL."
  • ACT-CYC-040-90 A Relevance Based Meta Level, R.V.Guha and A.Levy, 2/90,MCC technical report number.
  • ACT-CYC-083-90 The Representation of Defaults in Cyc Ramanathan V. Guha MCC and Stanford University (Wed Oct 12 17:38:33 CDT 1994) Abstract: This paper provides an account of the representation of defaults in Cyc and their semantics in terms of first order logic with reification. Default reasoning is a complex thing, and we have found it beneficial to separate out various complex issues whose ``current best solution is likely to change now and then - such as deciding between extensions, preferring one default to another, etc. - and deal with them explicitly in the knowledge base, thus allowing us to adopt a simple (and hopefully fixed) logical mechanism to handle just basic non-monotonicity itself. We also briefly sketch how this default reasoning scheme is implemented in Cyc
  • ACT-CYC-084-90 An Epistemological Level Interface for CYC Mark Derthick MCC derthick@mcc.com ACT-CYC-084-90 Submitted to AAAI-90 Topic: Knowledge Representation. Abstract: The field of AI continues to be polarized by arguments between Neats, who advocate a top-down approach beginning with a clear formalization of the problem to be solved and develop broad theories, and the Scruffies, who advocate bottom-up heuristic programs and more often develop systems of commercial use. Following McCarthy and Hayes's ##1McCarthyHayes69 suggestion for a distinction between epistemological and heuristic levels, we have added a logic-based interface to the CYC knowledge representation system, which makes it look neat. A translator converts between epistemological level and heuristic level representations. This uncouples the KB from the underlying heuristic reasoning system. If a desired conclusion is not drawn, the blame can be placed unambiguously on the knowledge enterer if the epistemological semantics do not justify it, or otherwise on the system implementor. Over the short term, at least, one does not have to depend on the other. As a matter of fact, CYC's implementors have satisfied its knowledge enterers over the long term. This paper describes the advantages of making this distinction, the user's view of CYC, the implementor's view, and how the translator keeps them in tune.
  • ACT-CYC-234-90 The Minimum Description Length Principle Applied to Feature Learning and Analogical Mapping Mark Derthick "This paper describes an algorithm for orthogonal clustering. That is, it finds multiple partitions of a domain. The Minimum Description Length (MDL) Principle is used to define a parameter-free evaluation function over all possible sets of partitions. In contrast, conventional clustering algorithms can only find a single partition of a set of data. While they can be applied iteratively to create hierarchies, these are limited to tree structures. Orthogonal clustering, on the other hand, cannot form hierarchies deeper than one layer. Ideally one would want an algorithm which does both. However there are important problems for which orthogonal clustering is desirable. In particular, orthogonal clusters correspond to feature vectors, which are widely used throughout cognitive science. Hopefully, orthogonal clusters will also be useful for finding analogies. A side effect which deserves more exploration is the induction of domain axioms in which the features are the predicates. The primary example used to demonstrate the orthogonal clustering algorithm, called MDL/OC, is finding the features {person, nationality, sex, generation}from the database of family relations used by Geoffrey Hinton ##1HintonFamilyTrees to demonstrate feature discovery by a back-propagation network. Brief examples from the literature of clustering and analogical mapping are also given, to illustrate the generality of the technique."
  • ACT-CYC-407-91 Ideas for Applying Cyc, D.Lenat and R.V.Guha, 12/91, MCC technical report number ACT-CYC-408-91.
  • ACT-CYC-406-91 Comparing Cyc to Other AI Systems, R.V.Guha and D.Lenat, 12/91, ACT-CYC-408-91.
  • ACT-CYC-423-91 The Applications of Contexts The focus of this chapter is on the different uses of the context mechanism. After listing some of these uses, we examine them in detail by means of examples chosen from the Cyc implementation of contexts.
  • ACT-CYC-212-91 The Minimum Description Length Principle Applied to Feature Learning and Analogical Mapping abridged version of ACT-CYC-234-90, "This paper discusses unsupervised learning of orthogonal concepts on relational data. Relational predicates, while formally equivalent to the features of the concept-learning literature, are not a good basis for defining concepts. Hence the current task demands a much larger search space than traditional concept learning algorithms, the sort of space explored by connectionist algorithms. However the intended application, using the discovered concepts in the Cyc knowledge base, requires that the concepts be interpretable by a human, an ability not yet realized with connectionist algorithms. Interpretability is aided by including a characterization of simplicity in the evaluation function. For Hinton's Family Relations data, we do find cleaner, more intuitive features. Yet when the solutions are not known in advance, the difficulty of interpreting even features meeting the simplicity criteria calls into question the usefulness of any reformulation algorithm that creates radically new primitives in a knowledge-based setting. At the very least, much more sophisticated explanation tools are needed."; author unclear (perhaps Derthick?).
  • CYC-234-94 1994 Index to Cyc publications
  • http://web.archive.org/web/19970630124400/http://www.cyc.com/tech-reports/cyc-234-94/chapter2_3.html#SECTION0030000000000000000
  • CYC-234-94 Key to Abbreviations/Accronyms in CYC Publications

Still missing

Not found in archive.org at these urls

  • Microtheories: An Ontological Engineer's Guide, Paul Blair R. V. Guha Wanda Pratt (March 5, 1992)
  • Context Dependence of Representations in Cyc, R.V.Guha, 8/93, MCC technical report number CYC 066-93. [Guha] R. V. Guha: “Context dependence of representations in CYC”, Technical Report CYC 066 – 93, Microelectronics and Computer Technology Corporation,

1993. http://www.cyc.com/tech-reports/act-cyc-066-93/act-cyc-066-93.html

  • Representing Information Bearing Things in Cyc, R.V.Guha and K.Pittman, 1/91 MCC technical report number ACT-CYC-022-91.
  • Micro-theories and Contexts in Cyc Part I: Basic Issues, R.V.Guha, 12/90, MCC technical report number ACT-CYC-300-90.
  • A Functional Interface for Cyc , D.Pratt and R.V.Guha, 3/90, MCC technical report number ACT-CYC-089-90-P.
  • The World According to Cyc, Part 3, R.V. Guha and D.Lenat, 12/89, MCC technical report number ACT-AI-455-89.
  • CycL: The Cyc Representation Language Part 3, R.V.Guha and D.Lenat, 12/89, MCC technical report number ACT-AI-454-89.
  • The World According to Cyc, Part 2: Agenthood, Institutions, and Agreements, R.V.Guha and D.Lenat, 12/89, MCC technical report number ACT-AI-453-89.
  • CycLing: Inferencing in Cyc, R.V.Guha and D.Lenat, 9/88, MCC technica report number ACA-AI-303-88.
  • The CycL Representation Language, D.Lenat, R.V.Guha, and D.V.Wallace, 9/88, MCC technical report number ACA-AI-302-88.
  • The World According to Cyc, D.Lenat and R.V.Guha, 9/88, MCC technical report number ACA-AI-300-88.

See also