Charniak, E., Riesbeck, C., and McDermott, D., 1979. Artificial Intelligence Programming, Baltimore: L. E. Erlbaum.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....in terms of a highly recursive formula conversion. For solving implicative rewrites, this conversion uses a call to itself rather than IMP SEARCH TAC. It passes another recursive call as the fconv fun argument of LOCAL SUB FCONV. For fast pattern matching, it stores rewrites in discrimination nets [ charniak80] instead of lists. A separate improvement is to expand out disjunctions during rewriting, causing automatic case splits. The conversion EXPAND DISJ FCONV derives (A B) C = A= C) B= C) A B) C = A C) B C) C (A B) C A) C B) This is applied at the same ....

E. Charniak, C. Riesbeck, and D. McDermott, Artificial Intelligence Programming (Lawrence Erlbaum Associates, 1980).

....be unified. For any two expressions, the unification algorithm either produces their most general unifier, or reports that no unifier exists. Unification plays a central role in theorem proving, polymorphic type checking [19] the language Prolog [8] and other areas of artificial intelligence [7]. Note:Ift is an expression such that t = F [t] then putting t for both x and y unifies G[x; x] and G[y; F [y] This t can be written as the infinite expression F [F [F [ which can easily be formalized in LCF. However, allowing infinite expressions would be a drastic departure from MW s ....

....let #x. # P , and let #x. ## # # P . Implicit universal quantifiers (over free variables in formulas) cover defined values only. This hides irrelevant clutter about #. 7. 1 Composition of substitution Many implementations of unification build up a substitution one variable at a time [7]. MW verify a simpler algorithm involving the operation of composing two substitutions. My symbol for composition is the infix operator THENS (by analogy with the tactical THEN) If r and s are substitutions, then MW (6) define the composition r THENS s to be the substitution satisfying, for all ....

E. Charniak, C. Riesbeck and D. McDermott, Artificial Intelligence Programming (Lawrence Erlbaum Associates, 1980).

....SM it is irrelevant whether the accessor considers this as a name or an address. Type Type Type Type Type Type Type Type 247 Type Type Type Type Type Type Type Type Type Type Type Type Type Type Type Type Integer String Array List string variable length B[1] B[2] B[3] . B[n] name Figure 2 Descriptor Pointer Examples The pervasive effect of the fixed size descriptors is to bring the efficiency of regular structure to the now inefficient world of irregular and dynamic symbolic data. Typical fixed size data structures, such as an array of ....

....the single program multiple data method (SPMD) 4.1. Extended Traditional Memory Manipulation Traditional symbolic data types which could require support from SM include sets, lists, stacks, arrays, structures, environments, text (strings) graphs, frames, semantic nets, and discriminant nets [3]. Common operations on objects, such as insert, car cdr, etc. could be implemented directly as primitives of the SM. Operations dealing with the structure of an object (such as size of, number of dimensions, etc. can be done strictly through descriptor processing (DP) The compare operation is ....

[Article contains additional citation context not shown here]

Charniak, E., C.K. Riesbeck, and D.V. McDermott, Artificial Intelligence Programming, Lawrence Erlbaum Associates, New Jersey (1980).

....A name could point into such a list at any level of the descriptor hierarchy. 8 Type Type Type Type Type Type Type Type 247 Type Type Type Type Type Type Type Type Type Type Type Type Type Type Type Type Integer String Array List string variable length B[1] B[2] B[3] . B[n] name Figure 3 Descriptor Pointer Examples A different problem arises when the structure of the data object is also variable and dynamic. An example of this is a list where the number of elements and the depth of the elements change in time. The advantage of the unchanging ....

....[16] 11 Data objects having fixed size are stored in hunks. Hunks are larger fixed sized segments of the value memory. A hunk corresponds to more normal page sizes and data objects reside within word boundaries. An example is an array of integers as shown in Figure 6. Descriptor a[1] a[2] a[3] a[i] link Figure 6 Hunk Based Array There might be a small set of different sized hunks to choose from. The same could apply with chunks) Each data value in a hunk could take up more than one word of memory. Conversely, there could be many values per word (i.e. a bit array) Optimization ....

[Article contains additional citation context not shown here]

Charniak, E., C.K. Riesbeck, and D.V. McDermott, Artificial Intelligence Programming, Lawrence Erlbaum Associates, New Jersey (1980).

....members of the group thought it might be confusing. Probably inspired by Scheme, which in those days was using just X to indicate splicing [14] we finally decided on , X [15] 9 Meanwhile McDermott altered the Conniver notation slightly by changing X to X. In this form it appeared in [1] in 1980. As far as I know, the problems of nested splicing didn t get worked out until 1982. In January of that year Guy Steele circulated an example of quasiquotations nested three levels deep. He remarked that , X was fairly obvious , but that it took him a few tries to get his use of ....

E. Charniak, C. K. Riesbeck, and D. V. Mcdermott. Artificial Intelligence Programming. Lawrence Erlbaum Assoc., Hillsdale, NJ, first edition, 1980.

.... For example, if (PTRANS actor Person to Location) is unified with (PTRANS actor John1 to Store1) the resulting variable binding list is: T (PERSON # JOHN1: PERSON) LOCATION # STORE1: STORE) The GATE unification algorithm is based on previous unifiers (Schank Riesbeck, 1981; Charniak, Riesbeck, McDermott, 1980) with appropriate extensions for typed variables and obs, multiple slot values per slot name, special obs, and cyclic data structures. The following functions deal with unification: ob unify ob1 ob2 bd) bd or NIL Given two obs and a binding list, attempt to unify the two obs. If unification ....

Charniak, E., Riesbeck, C. K., & McDermott, D. V. (1980). Artificial intelligence programming.

....#P. Proof. D 2 Exts(S) for every S (as we observe later in Theorem 20.1) so finding an extension is trivial. The other five problems may all be computed by guessing sets E D and accepting iff the desired condition is true, all deterministic polynomial computations from the above theorems. 13 [Charniak, Riesbeck, and McDermott 1980] 18 (18.63) Conjecture. Find E 2 QExts(S) is NP complete. 18.64) Question. Is Find E 2 FGExts(S) NP complete (18.65) Theorem. Is E 2 Exts(S) Is E 2 QExts(S) and Is QExts(S) are in co NP. Proof. We see E = 2 Exts(S) is in NP by first checking E 2 Exts(S) deterministically as ....

....2 FGExts( S) depending on whether the agent is locally grounded or finitely grounded. Each of these steps is in P, so their combinations are also. 17 See [Quine 1953] Rescher 1964] and [G ardenfors 1980] 18 See [Doyle 1979] Stallman and Sussman 1977] London 1978] McAllester 1980] [Charniak, Riesbeck, and McDermott 1980], Thompson 1979] Martins 1983] Goodwin 1982] and [McDermott 1982b] 40 (33.5) Theorem. If (6 S ; Delta) is a strict finitely grounded simple reasons agent, then Is E 2 Delta(S) is in P. Proof. Since FGExts = FGExts, the previous theorem applies. 33.6) Theorem. If (6 S ; Delta) is a ....

Charniak, E., Riesbeck, C. K., and McDermott, D. V., 1980. Artificial Intelligence Programming, Hillsdale, New Jersey: Lawrence Erlbaum Associates.

.... Gamma Gamma Gamma Figure 5: A Solvable Odd Loop 7 stable well founded state if such a state exists, and otherwise will recognize and report failure. This represents an improvement over the original TMS of Doyle [Doyle 1979] as well as other published procedures for truth maintenance [Charniak, et al. 1980, Goodwin 1986] These systems all fail (perhaps even fail to terminate) in the presence of certain circular dependencies that have been characterized as odd loops. An odd loop is a cycle of arcs with an odd number of minus signs, as in Figs. 3, 4, and 5. A dependency network containing such a ....

Charniak, E., C. K. Riesbeck, and D. V. McDermott, Artificial Intelligence Programming, Lawrence Erlbaum Associates, Hillsdale, N.J., 1980. 40

....of specification and refinement. The pattern compilation which is described in this paper is a novel technique, see Peyton Jones [Peyton Jones, 1987] for a more conventional approach to the compilation of patterns which match deterministically. Nilsson [Nilsson, 1980] and Charniak et al. [Charniak, 1987] describe various search related applications for computer programs, in particular Charniak et al. Charniak, 1987] describes the connection between search problems and nondeterministic computational processes. This paper describes a simple method of adding non determinism to a functional program ....

....see Peyton Jones [Peyton Jones, 1987] for a more conventional approach to the compilation of patterns which match deterministically. Nilsson [Nilsson, 1980] and Charniak et al. Charniak, 1987] describe various search related applications for computer programs, in particular Charniak et al. [Charniak, 1987] describes the connection between search problems and nondeterministic computational processes. This paper describes a simple method of adding non determinism to a functional program using monad comprehensions, see Sondegaard [Sondegaard, 1992] for a discussion of the issues involved in ....

Charniak, E. et al. 1987 Artificial Intelligence Programming. Lawrence Erlbaum Associates.

....and Darlington (1977) are more in the spirit of the transformations described here. We have viewed the specification of rule based transition systems as computational processes from the start whereas Morgan (1990) gives a different perspective of specification and refinement. Nilsson (1980) and Charniak et al. 1987) describe various search related applications for computer programs, in particular Charniak et al. 1987) describes the connection between search problems and non deterministic computational processes. Non deterministic (functional) programs are discussed in Sondegaard (1992) and Clark (1994) ....

....specification of rule based transition systems as computational processes from the start whereas Morgan (1990) gives a different perspective of specification and refinement. Nilsson (1980) and Charniak et al. 1987) describe various search related applications for computer programs, in particular Charniak et al. 1987) describes the connection between search problems and non deterministic computational processes. Non deterministic (functional) programs are discussed in Sondegaard (1992) and Clark (1994) gives an operational description of a system with builtin non determinism. In general, systems that provide ....

Charniak, E. et al. (1987) Artificial Intelligence Programming. Lawrence Erlbaum Associates.

....facets are used to ease the complex and extremely task intensive process of creating a semantic network. Frames in a hierarchical network contain information about the software objects, their grouping, and the relationships between them. The frames system was originally proposed by Charniak et at [3], and the one used in AIRS is a modification of it. The model for similarity computation and classification in AIRS is based on three types of software objects. Features are used to describe a software component. These features are similar to the facets described earlier. A component is a set of ....

E. Charniak, C. Riesbeck, D. McDermott, and J. Meehan. Artificial Intelligence Programming. Lawrence Erlbaum, 1987.

....remain uncertain. The roots of discrimination tree indexing appear to be directly in formula manipulation systems, and the roots of path indexing appear to be in database technology. Discrimination tree indexing (also called discrimination net indexing) and some of its variations have appeared in [13, 3, 8, 14, 15, 20, 2, 5, 6]. It is used to find demodulators in [13] it is presented from a Lisp point of view in [3] it is used with very large sets of terms in [15] it is compared to path indexing in [20] it is used in the context of very high performance Knuth Bendix completion in [5, 6] and it forms the basis for a ....

....the roots of path indexing appear to be in database technology. Discrimination tree indexing (also called discrimination net indexing) and some of its variations have appeared in [13, 3, 8, 14, 15, 20, 2, 5, 6] It is used to find demodulators in [13] it is presented from a Lisp point of view in [3], it is used with very large sets of terms in [15] it is compared to path indexing in [20] it is used in the context of very high performance Knuth Bendix completion in [5, 6] and it forms the basis for a high performance deduction toolkit in [2] The predecessors of the path indexing method ....

E. Charniak, C. Reisbeck, and D. McDermott. Artificial Intelligence Programming. Lawrence Earlbaum Assoc., Hillside, NJ, 1980.

....defined collection of facets orthogonal to objects that 6 Debugging methodologies and tools for applications involving rapidly changing data is itself an interesting research topic, but outside the scope of this paper. 7 A facet is superficially similar to the frame of AI languages [Charniak 80] but analogies based on frames would be misleading. 10 must execute at the same physical location. That is, a process represents a single virtual address space in which its facets reside, and allocation of computation and communication resources to facets is handled by their process. Multiple ....

Eugene Charniak, Christopher K. Riesbeck and Drew V. McDermott. Artificial Intelligence Programming. Lawrence Erlbaum Assoc., Hilsdale NJ, 1980.

....to compute (or estimate, or specify) the desirability of sending a particular message to a particular recipient relative to the other possible 4 combinations. An evaluation function for generalized dispatching serves somewhat the same purpose as the evaluation function in a best first search[6]. Subclass and classless dispatching can be viewed as cases of recipient resolution in which the evaluation functions are solely based on the classes and subclasses, the number of parameters, and their data types. True recipient resolution gives the programmer the power to define dispatching ....

Eugene Charniak, Christopher K. Riesbeck, Drew V. McDermott, and James R. Meehan. Artificial Intelligence Programming, page 166. Lawrence Erlbaum Associates, Hillsdale, NJ, second edition, 1987.

....them and setting their stop timestamps. For example, given the pattern: smile Jim] and the retraction timestamp: 19930809102508 the matching database assertion: 19930809102345:inf [smile Jim] would be updated to: 19930809102345:19930809102508 [smile Jim] Unification Unification (Charniak, Riesbeck, and McDermott, 1980) is a pattern matching operation performed on two list objects. The objects may contain a special kind of object called a variable. Two objects unify if values for variables can be found such that substituting the values for those variables in the objects would produce similar structures. For ....

.... human human location location When an unbound variable successfully unifies with an object, that object becomes that variable s value in the list of bindings. Second, a bound variable unifies with an object if the value of that variable unifies with the object. Instantiation Instantiation (Charniak, Riesbeck, and McDermott, 1980) takes a list object and binding list, and returns a copy of the object in which any variables have been replaced by their values in the binding list. For example, if the object: ptrans human na location] is instantiated with the binding list: human = Jim location = store87 the resulting ....

[Article contains additional citation context not shown here]

Charniak, E., Riesbeck, C. K., and McDermott, D. V. (1980). Artificial intelligence programming.

....with lsim as the operator. 4.2 Other AI Techniques The way we apply rule version space search misses common subexpressions in positive examples. Common subexpression location with DAGs, as in compilers [1] is probably not useful since compilers look for exact matches. Discrimination networks [4] may be a means of finding common subexpressions. The common subexpressions could then be found with a search and match function. The work of Feng and Muggleston [8] may be applicable. They are concerned with finding selectors which classify positive and negative examples of higher order terms. ....

Eugene Charniak, Christopher K. Riesbeck, and Drew V. McDermott. Artificial Intelligence Programming. Lawrence Erlbaum Associates, 1980, chapters 11 and 14, pp. 121--130 and 162--176.

....The approach described here is also employed to enforce changes of the KB schema, i.e. concept and role redefinition, onto the KB. A description of this work is found in [Tho92] 7 Related Work The work described in this report adopts a monotonic data dependency network management approach (see [CRM80, Chap. 16] or [Neb90, Sec. 6.6.1] A data dependency network (DDN) basically consists of nodes, which denote believed propositions, justifications, which denote sets of propositions used in derivations, and the links between them. Justifications support nodes, and nodes participate in ....

Eugene Charniak, Christopher K. Riesbeck, and Drew V. McDermott. Artificial Intelligence Programming. Erlbaum, Hillsdale, N.J., 1980.

....comparisons (not counting the computation within the delayed computation of Lmax ) 3. 1 AN EFFICIENT WAY OF MAKING Omega LAZY The operation Omega takes two ordered lists L I and L J as arguments and returns an ordered list where each element is a compound element composed of one el 4 See [Charniak et al. 1987] for details on implementing lazy list operations. 1 2 2 4 3 6 6 3 9 5 8 4 5 10 10 15 12 25 20 15 30 24 18 12 6 12 18 30 36 6 i 1 2 3 4 5 6 L I Returned previously 1 2 3 5 6 L J j Non fringe remaining element Member of current fringe Figure 5: The fringe in Omega z . ement from L I and one ....

E. Charniak, C. K. Riesbeck, D. V. McDermott, and J. R. Meehan. Artificial Intelligence Programming. Lawrence Erlbaum Associates, Inc., Hillsdale, NJ, 1987.

....techniques cannot reduce the size of the search space to reasonable proportions. Under such circumstances, guesses have to be made to guide the search engine to 4 STEPHEN BEALE the area of the search space most likely to contain acceptable answers. Best first search (see, among many others, (Charniak et al. 1987)) is an example of how to use heuristics. Unfortunately, heuristic search cannot guarantee optimal answers. The hunting techniques applied in this research are most closely related to the field of constraint satisfaction problems (CSP) Beale, 1996) overviews this field and (Tsang, 1993) ....

Charniak, E.; Riesbeck, C.K.; McDermott, D.V.; and Meehan, J. R. 1987. Artificial Intelligence Programming. Hillsdale, NJ: Erlbaum.

....(x y) Given a term A and in order to find all the rewritings that can be applied on A, it is necessary to test if the pattern B matches A for each rewriting rule B Gamma C. Instead of doing the matching test separately for each rule, all the patterns are grouped in a discrimination net [5]. The possible rewritings are then found with only one matching test. A possibility of limiting the search space is also given to the user by defining in which theories the search for rewritings has to be performed. The composition of these two techniques, the discrimination net and the limitation ....

E. Charniak, C. K. Riesbeck, D. V. McDermott, J. R. Meehan, "Artificial intelligence programming", Lawrence Erlbaum Associates Publishers, New Jersey 1987.

.... of stratified default theories or extensions of default theories that have no odd cycles (in some precise sense) are given by Papadimitriou and Sideri (1994) and Cholewi nski (1995a, 1995b) Algorithms for handling a TMS with nogoods have been developed in the AI community by Doyle (1979) and Charniak et al. 1980). But, as Elkan (1990) points out, these algorithms are not always faithful to the semantics of the TMS and their complexities have not been analyzed. Dechter and Dechter (1994) provide algorithms for manipulating a TMS when it is represented as a constraint network. The efficiency of their ....

Charniak, E., Riesbeck, C. K., & McDermott, D. V. (1980). Artificial Intelligence Programming, chap. 16. Lawrence Erlbaum, Hillsdale, NJ.

....passed not only an expression to be expanded, but also another expansion function. This function may or may not be used to perform further expansion. This is analogous to passing an explicit continuation to a function, which is well known to be a powerful programming technique (for example, see [1, 12]) In this paper we have demonstrated that the related technique of passing expanders explicitly provides a powerful tool for defining syntactic extensions. Using EPS, it is possible to implement several special forms typically found in Lisp systems, avoiding the need to build these forms into the ....

Charniak, E., Riesbeck, C. K., and McDermott, D. V., Artificial Intelligence Programming, Lawrence Erlbaum Associates, 1980.

....for the partial sequences. The templates are described by the regular expressions shown in Fig. 15. Identification of a partial procedure and relating it to an upper level operator is carried out by matching the sequence with the templates by an ATN (Augmented Transition Network) interpreter [28]. At the end of this step, the whole input sequence is organized into a set of ACHIEVE operators. C.2 Step 2 Analyzing Dependencies: Dependency among operators above the OSL are analyzed by tracing the coincidence of focused objects. An operator OP i moves the objects included in its base ....

E. Charniak et al., Artificial Intelligence Programming, Lawrence Erlbaum, 1987.

....than the original truth maintenance systems ) have been studied by a variety of authors and have gained currency in artificial intelligence in spite of rather unwieldy descriptions in terms of complex procedures. See [Doyle 1979] Stallman and Sussman 1977] London 1978] McAllester 1980] [Charniak, Riesbeck, and McDermott 1980], Thompson 1979] G ardenfors 1980] Steele 1980] de Kleer and Doyle 1982] McDermott 1982] Goodwin 1982] and [Martins 1983] There is little hope for improving on existing RMS implementations without clearer statements of their intended behaviors and better analyses of their ....

Charniak, E., Riesbeck, C. K., and McDermott, D. V., 1980. Artificial Intelligence Programming, Hillsdale, New Jersey: Lawrence Erlbaum Associates.

....c. 7.3. 2 Example: arithmetic expressions This theory specification contains a syntax section with mixfix declarations encoding the priority grammar from x7.1: ExpSyntax = Pure types exp syntax 0 : exp ( 0 9) exp, exp] exp ( 0, 1] 0) exp, exp] exp ( [3, 2] 2) exp = exp ( 3] 3) end CHAPTER 7. DEFINING LOGICS 74 If you put this into a file ExpSyntax.thy and load it via usethy ExpSyntax , you can run some tests: val readexp = Syntax.testread (synof ExpSyntax.thy) exp ; val it = fn : string unit readexp 0 0 0 0 0 0 ....

.... provided by the macro expander (see x8.5) Executing Syntax.printgram reveals the productions derived from the above mixfix declarations (lots of additional information deleted) Syntax.printgram (synof ExpSyntax.thy) exp = 0 = 0 (9) exp = exp[0] exp[1] 0) exp = exp[3] exp[2] = 2) exp = exp[3] 3) Note that because exp is not of class logic, it has been retained as a separate nonterminal. This also entails that the syntax does not provide for identifiers or paranthesized expressions. Normally you would also want to add the declaration arities ....

E. Charniak, C. K. Riesbeck, and D. V. McDermott. Artificial Intelligence Programming. Lawrence Erlbaum Associates, 1980.

....on each rule s first non [NULL ] daughter. The index itself consists of a vector (one dimensional array) the i th element of the vector giving access to all the rules whose first daughter category s object grammar category identifier is i. Each element of the vector is a discrimination tree (Charniak, Riesbeck McDermott, 1980) keyed on the feature values of the first (non null) daughter, the leaves of the trees containing rules. The index is cheap to compute, and in addition is cached, only being recomputed if the grammar is modified. So, for example, the object grammar rule shown in figure 3.4 is stored in the ....

Charniak, E., C. Riesbeck & D. McDermott (1980) Artificial intelligence programming.

....not entirely new. ON conditions in programming languages and early DBMSs [Oll78] and signals in operating systems [Bac88, Geh84, Hug79, LMKQ89] have been used early on. Later, Artificial Intelligence (AI) systems have used daemons for asynchronous execution of procedures attached to frame slots [CRM80, WB79] Furthermore, multi paradigm systems, such as LOOPS [BS83] and KEE [Int85] have incorporated active values as a new technique that generalizes asynchronous rule processing. However, most of these systems do not support typical database functionality, such as data sharing, consistency, and ....

Eugene Charniak, Christopher K. Riesbeck, and Drew V. McDermott. Artificial Intelligence Programming. Lawrence Erlbaum Associates, Publishers, Hillsdale, New Jersey, 1980.

....because of a valid justification, all its supportees must be reconsidered, because they might have been assigned a new status based on the assumption that the supporter be OUT. Details of this algorithm can be found in (Hern andez 1984) where several variants were implemented based on ideas in (Charniak et al. 1980) and (Doyle 1979) 4 Conclusion Even though the solution techniques for the general constraint satisfaction problem available in the literature represent large efficiency improvements over the obvious backtracking algorithm, they are limited by their generality . That is, being general, ....

Charniak, E., Riesbeck, C. K., and McDermott, D. V. (1980). Artificial Intelligence Programming. Erlbaum, Hillsdale, N.J.

....A Reason Maintenance System (Doyle [6] de Kleer [3] Reiter and de Kleer [12] provides facilities for building and accessing just such a record of logical structure. In particular, an RMS records a structure of justifications and nogoods, which are both kinds of data dependency (Charniak et al. [2]) If one formula is shown to be a logical consequence of a set of others then the RMS can be informed of this fact, which it records as a justification. If a set of formulae are shown to be inconsistent, then the RMS can be informed of this fact too, which it records as a nogood. Now, how can ....

Charniak E., Riesbeck C.K., McDermott D.V. and Meehan J.R., Artificial Intelligence Programming, Lawrence Erlbaum (1987).

....when other techniques cannot reduce the size of the search space to reasonable proportions. Under such circumstances, guesses have to be made to guide the search engine to the area of the search space most likely to contain acceptable answers. Best first search (see, among many others, (Charniak et al. 1987)) is an example of how to use heuristics. The hunting techniques applied in this research are most closely related to the field of constraint satisfaction problems (CSP) Beale 1996) overviews this field and (Tsang 1993) covers it in depth. Further references include (Mackworth 1977) ....

Charniak, E; Riesbeck, C.K.; McDermott D.V. and Meehan, J.R. 1987. Artificial Intelligence Programming.

....(frame) decision (RMM) Agent Bayesian belief update Information Act Environment and Other Agents capability (frame) commitment Figure 2: The components of agent s mental state. To implement the components of an agent s mental state depicted in Figure 2 we use a version of the framebased system [2]. Thus, frames are used to represent the agent s beliefs, capabilities, and preferences, and to organize his knowledge using the object oriented paradigm into hierarchy of concepts and classes. Our rational agent processes his knowledge to determine optimal decisions, and to perform belief update ....

E. Charniak, C. K. Riesbeck, D. V. McDermott, and J. R. Meehan. Artificial Intelligence Programming. Lawrence Erlbaum Associates, Publishers, Hillsdale, New Jersey, 1987.

....and Ourston, 1991b] and a complete description is given in [Ourston, 1991] 3 THE DEDUCTIVE COMPONENT The deductive component of Either is a standard backward chaining, Hornclause theorem prover similar to Prolog. Our particular implementation is based on the deductive retrieval system from [Charniak et al. 1987]. Deduction is the first step in theory revision. The system attempts to prove that each example is a member of each of the known categories. Failing positives (examples that cannot be proven as members of the correct category) indicate overly specific aspects of the theory and are passed on to ....

E. Charniak, C. Riesbeck, D. McDermott, and J. Meehan. Artificial Intelligence Programming (2nd Ed). Lawrence Erlbaum and Associates, Hillsdale, NJ, 1987.

No context found.

Charniak, E., Riesbeck, C. and McDermott, D., Artificial Intelligence Programming (Erl- baum, Hillsdale, N J, 1980).

....action, it caches it in the table, so the next lookup will be much faster. This table does not depend on the particular problem being solved, so it is saved from run to run. The indexes in the list, and the action difference table, are implemented as discrimination trees on symbolic expressions [8, 28]. Each node discriminates on a particular position in the expression (CAR, CADR, etc. and partitions the expressions it is storing into buckets depending on whatever content it finds at that position. When a bucket gets to be too large, it is further discriminated. These indexes are used to ....

Eugene Charniak, Christopher Riesbeck, Drew McDermott, and James Meehan. Artificial Intelligence Programming. Lawrence Erlbaum Associates, 1987. second edition.

No context found.

Charniak, E., Riesbeck, C., and McDermott, D., 1979. Artificial Intelligence Programming, Baltimore: L. E. Erlbaum.

No context found.

E. Charniak, C.K. Riesbeck, D.V McDermott and J.R. Meehan. 1987. Artificial IntelligenceProgramming. Erlbaum, Hillsdale, NJ.

No context found.

Charniak, E., C.K. Riesbeck, and D.V. McDermott. Artificial Intelligence Programming. Lawrence Erlbaum Associates, Hillsdale, New Jersey, 1980.

No context found.

Charniak, E., Riesbeck, C. K., and Drew, V. (1987). Artificial Intelligence Programming. Lawrence Erlbaum Associates, Hillsdale, N J.

No context found.

Charniak, E; Riesbeck, C.K.; McDermott D.V. and Meehan, J.R. 1987. Artificial Intelligence Programming. Hillsdale, NJ: Erlbaum.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC