David A. Wroblewski. 1987. Nondestructive graph unification. In Howard Forbus, Kenneth; Shrobe, editor, Proceedings of the 6th National Conference on Artificial Intelligence (AAAI-87), pages 582--589, Seattle, WA, July. Morgan Kaufmann.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....of feat.re cxist,qccs and gesture address disagreements are Lrcatcd by tending term structures. This extension can he seen ; intuitionistic. Tim extension is classified into (1) above. Bed on this papcr s formaliz;rtion, unilicaion gorithms have been developed using graph unification techniques[23, 16]. Programs bqed on fimsc alg rihms have been implemented in Collllllon Lisp. 2 Requirements of Negative In describing linguistic ikrmation using (typed) rare structures, negative descripLions make the scripfion compac L intuitive, mid hence easy to mdcr stand. I br example, we want to describe ....

....yields A because of tag inconsisteucy. These computations require negligible additional computation. qb simplify he exphmation, the destrue;ire version of graph unification is used above. Other versions bed ou more efficient grapll uuilication methods such t Wroblewski s aud Kogurc s method[23, 16] have also been developed. 1,hrthermore, it is ey to modify other graph unificatiou nethods[21, 6] to allow augmented TFSs. 5 Conclusion This paper hs prol)ed au augmentation of fea ture structures (FSs) which introduces negative m formation into FSs m uuificatioud)sed tbrmalisms. U ....

David A. Wroblewski. Nondestructive graph unification. In Proceedings o] the 6th Natioonl Conference on Artificial Intelligence, pages 582-587, AAAI, Seattle, Washington, USA, 1987.

....than for unification itself. Hence, it is crucial to reduce the amount of copying, both in terms of the number and of the size of copies, in order to improve the efficiency of unification. A naive implementation of unification would copy the arguments even before unification starts. That is what [Wroblewski 87] calls early copying. Early copying is wasted effort in cases of failure. He also introduced the notion of over copying, which results from copying both arguments in their entirety. Since unification produces its result by merging the two arguments, the result usually contains significantly fewer ....

....END unify Figure 4: The unification procedure 328 approach methods early over redundant incr. lazy structure copying copying copying copying copying sharing naive yes yes yes no no no Peteira 85 no no no no no yes Karttunen Kay 85 no no yes no yes yes Karttunen 86 no no yes no no no Wroblewski 87 no yes yes yes no no Godden 90 no no yes no yes yes Kogure 90 no yes yes no yes yes LIC no yes no yes yes yes Figure 5: Comparison of unification approaches Both of these approaches suffer from difficulties of their own. In Godden s case, part of the copying is substituted traded for by the ....

David A. Wroblewski. Nondestructive graph unification. In Proceedings of the 6 National Conference on Artificial Intel- ligence, AAAI, pp. 582-587, Seattle, WA, 1987. 330

....# # A = b C = 1 # D = e # F = 1 # # Figure 1: Two ways to represent an identical graph. redundant. As copying is the most expensive part of unification, a great deal of research has gone in eliminating superfluous copying. Examples of these approaches are given in (Tomabechi, 1991) and (Wroblewski, 1987). In order to avoid superfluous copying, these algorithms incorporate control data in the graphs. This has several drawbacks, as we will discuss next. Memory Consumption To achieve the goal of eliminating superfluous copying, the aforementioned algorithms include administrative fields which we ....

....has to perform additional work. Running the same tests on machines with less memory showed a clear performance advantage for the algorithms using less memory, because paging could be avoided. 5 Related Work We reduce memory consumption of graph unification as presented in (Tomabechi, 1991) or (Wroblewski, 1987)) by separating scratch fields from node structures. Pereira s (Pereira, 1985) algorithm also stores changes to nodes separate from the graph. However, Pereira s mechanism incurs a log(n) overhead for accessing the changes (where n is the number of nodes in a graph) resulting in an O(n log n) ....

David A. Wroblewski. 1987. Nondestructive graph unification. In Howard Forbus, Kenneth; Shrobe, editor, Proceedings of the 6th National Conference on Artificial Intelligence (AAAI-87), pages 582--589, Seattle, WA, July. Morgan Kaufmann.

....of the dags that need not be copied are redundant. In the ideal case, only those parts that are actually needed for the new term are copied. early copying Copies are made before it is known whether a unification will fail. If a unification fails, then copies that were made serve no purpose. cf. [Wro87]: Wroblewski s definition of early copying indicates the redundant copying that is done before a failing unification starts, not the superfluous coping that is done to find out unification fails. Ideally, both types of copying are to be avoided. In addition, Tom91] gives two requirements that ....

....been implemented as such) We would like to note that by far not all algorithms have been covered. Instead we limited ourselves to algorithms that proved to be the most e#cient and that are best suited for parallelization. 4. 1 Destructive Unification The algorithm we give here is taken from [Wro87] and can be contributed to [Shi84, Kar84, Wit86] The nodes of the graphs contain a forward and a arc list field. The function of the latter should be obvious. The forward field is used to make a node identical to another. If it points to another node, the semantics is that the node it points to ....

[Article contains additional citation context not shown here]

David A. Wroblewski. Nondestructive graph unification. In Howard Forbus, Kenneth; Shrobe, editor, Proceedings of the 6th National Conference on Artificial Intelligence (AAAI-87), pages 582--589, Seattle, WA, July 1987. Morgan Kaufmann. 12, 17, 20, 21

....of the dags that need not be copied are redundant. In the ideal case, only those parts that are actually needed for the new term are copied. early copying Copies are made before it is known whether a unification will fail. If a unification fails, then copies that were made serve no purpose. cf. [Wro87]: Wroblewski s definition of early copying indicates the redundant copying that is done before a failing unification starts, not the superfluous coping that is done to find out unification fails. Ideally, both types of copying are to be avoided. In addition, Tom91] gives two requirements that ....

....been implemented as such) We would like to note that by far not all algorithms have been covered. Instead we limited ourselves to algorithms that proved to be the most efficient and that are best suited for parallelization. 4. 1 Destructive Unification The algorithm we give here is taken from [Wro87] and can be contributed to [Shi84, Kar84, Wit86] The nodes of the graphs contain a forward and a arc list field. The function of the latter should be obvious. The forward field is used to make a node identical to another. If it points to another node, the semantics is that the node it points to ....

[Article contains additional citation context not shown here]

David A. Wroblewski. Nondestructive graph unification. In Howard Forbus, Kenneth; Shrobe, editor, Proceedings of the 6th National Conference on Artificial Intelligence (AAAI-87), pages 582--589, Seattle, WA, July 1987. Morgan Kaufmann.

....than for unification itself. Hence, it is crucial to reduce the amount of copying, both in terms of the number and of the size of copies, in order to improve the efficiency of unification. A naive implementation of unification would copy the arguments even before unification starts. That is what [Wroblewski 87] calls early copying. Early copying is wasted effort in cases of failure. He also introduced the notion of over copying, which results from copying both arguments in their entirety. Since unification produces its result by merging the two arguments, the result usually contains significantly ....

....ENDIF ENDIF END unify Figure 4: The unification procedure approach methods early over redundant incr. lazy structure copying copying copying copying copying sharing naive yes yes yes no no no Pereira 85 no no no no no yes Karttunen Kay 85 no no yes no yes yes Karttunen 86 no no yes no no no Wroblewski 87 no yes yes yes no no Godden 90 no no yes no yes yes Kogure 90 no yes yes no yes yes LIC no yes no yes yes yes Figure 5: Comparison of unification approaches Both of these approaches suffer from difficulties of their own. In Godden s case, part of the copying is substituted traded for by the ....

David A. Wroblewski. Nondestructive graph unification. In Proceedings of the 6 th National

....is a costly operation in unificationbased parse systems, because it involves a lot of copying of feature structures. In many implementations of parsing systems it takes more than 80 of the total parse time. Several algorithms have been devised to do unification efficiently (Tomabechi 1991; Wroblewski 1987; Sikkel 1993) The efficiency of unification can be increased by minimizing the amount of copying in cases that unification fails, while on the other hand the overhead costs to do this should be as small as possible. Up to now Tomabechi s algorithm seems to be the fastest. With this algorithm the ....

Wroblewski, D. (1987). Nondestructive graph unification.

....regression model results in a negative value for the constant. Thus, we did not allow for constant coefficients in the best models. Related Work In the unification grammar research paradigm, several efficient unification algorithms have been proposed. Non destructive Unification proposed in (Wroblewski, 1987) copies nodes incrementally as the DAGs are traversed during unification. Although this approach solves over copying problem, it does not support structure sharing. The approach taken in (Pereira, 1985) supports structure sharing by having the result DAG share some nodes with the original DAGs. ....

Wroblewski, D. (1987). Non-destructive graph unification, in Proceedings of the 6th National Conference on Artificial Intelligence, Seattle, WA, pp. 582-587.

....on f structures. For example, if you are performing destructive post processing on strings you should consider copying the strings before changing their contents. There is a great deal of research on more efficient copying policies than the nave copying scheme currently implemented in t [Wroblewski, 1987, Godden, 1990, Emele, 1991, Tomabechi, 1991] We hope to address this in the next version of t . 3 t F Structures This section describes what t f structures may contain, and how t handles various operators for disjunction, negation, etc. This section also provides examples which specify ....

Wroblewski, D. (1987). Nondestructive graph unification. In AAAI-87, pages 582--589.

.... as versions of them together with environments binding previously uninstantiated DAG nodes to values (Pereira, 1985; Emele, 1991) and secondly, destructively updating (sometimes incrementally constructed copies of) portions of input feature structures to produce the results (e.g. Karttunen, 1986; Wroblewski, 1987; Kogure, 1990; Tomabechi, 1991) The first approach often results in less copying of DAGs, though Pereira s technique involves practical overheads (see below) and Emele s requires a parser using the unification module to perform chronological backtracking. The second approach is prone to ....

....no space was wasted in an append operation. If the unification succeeded, then any new bindings produced, a copy of the first list, and the second list are destructively concatenated together 16 . This approach to minimising the space consumed by failed unifications is similar in spirit to Wroblewski s (1987) unification technique involving updating feature structures destructively. The algorithm can in some cases add active edges to the chart which have no chance of being incorporated into a complete analysis. In particular, ones ending at the last vertex when there are no daughter less rules in the ....

Wroblewski, D. (1987) "Nondestructive graph unification." In Proceedings of the 6th National Conference on Artificial Intelligence (AAAI-87), Seattle, WA, 582--587.

No context found.

David A. Wroblewski. 1987. Nondestructive graph unification. In Howard Forbus, Kenneth; Shrobe, editor, Proceedings of the 6th National Conference on Artificial Intelligence (AAAI-87), pages 582--589, Seattle, WA, July. Morgan Kaufmann.

No context found.

D. Wroblewski, `Nondestructive graph unification', Proc. AAAI-87, 1987, pp. 582--587.

No context found.

David A. Wroblewski. 1987. Nondestructive graph unification. In Proceedings of AAAI-87, pages 582--587, Seattle, WA.

No context found.

David A. Wroblewski. 1987. Nondestructive graph unification. In Proceedings of AAAI-87, pages 582--587, Seattle, WA.

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