J.e.L. Ambite and C.A. Knoblock, "Planning by Rewriting: Efficiently Generating High-Quality Plans," in Proc. 14 National Conf. on Artificial Intelligence, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....all possible kinds of changes in the constraint graph during the search process not only after the current graph has been completely searched, and not only along an increasing plan length. Unlike other planning systems that use productions rules to change the plan as part of the search (e.g. [Ambite Knoblock 1997]) modeling planning as an SCSP allows us to specify the problem in a declarative manner and enables the corresponding productions to be deduced automatically. The automatic method guarantees that local search methods can potentially find all valid plans, which is not normally the case with manual ....

Ambite, J. L., and Knoblock, C. A. 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. AAAI-97, 706--713.

....identified as the involved agent sends a detailed message about it to the monitor. However, the execution still proceeds as mentioned before assuming an empty result set. Figure 4 shows a screen shot of the IPMs log table. Figure 4: IScape Processing Monitor (IPM) 5. Related Work SIMS [AK97, AKS96], TSIMMIS [CMH 94] Information Manifold [LRO96] and OBSERVER [MIKS00] are some of the other efforts to integrate information from multiple heterogeneous sources. The goal of InfoQuilt is to provide an environment where users can query, analyze, reason about inter domain relationships and ....

....They then use these models to translate a query specified by the user into an execution plan that specifies the actual available sources that will be used and how information retrieved from them will be integrated. We compare our vision and approach with other systems in this section. SIMS [AK97, AKS96] adopts the approach of creating a model of the domain using a knowledge representation system establishing a fixed vocabulary and accepting queries specified in the same language. A major limitation of the system is that its mediator is specialized to a single application domain [AHK96] ....

J. Ambite, and C. Knoblock. Planning by Rewriting: Efficiently generating highquality plans. Proceedings of the 14th National Conference on Artificial Intelligence, Providence, RI, 1997.

....the development of the Web. This information is difficult to handle because it is heterogeneous, dynamic and distributed in nature. However, there are very few approaches that try to integrate a set of different and specialized information sources and reuse the data retrieved to solve problems [1, 2]. This is especially true if the goal of the system is to solve complex problem solving tasks, like finding complete travel plans by gathering information available on the Web. In order to use Web information to solve complex problems, a framework called MAPWeb (MultiAgent Planning in the Web) ....

Ambite J.L., Knoblock C.A. Planning by rewriting: Efficiently generating high-quality plans. In proceedings of the Fourteenth National Conference on Artificial Intelligence. 1997.

....[6, 4, and as well as quality of the plans they produce [12, 5] Traditionally, this knowledge is encoded as search control rules to limit the search for generation of the first viable plan. Recently, Ambite and Knoblock have suggested an alternative approach called planning by rewriting [1]. Under this approach, a partial order planner generates an initial plan, and then a set of rewrite rules are used to transform this plan into a higher quality plan. Unlike the search control rules for partial order planners (such as those learned by UCPOP EBL [6] and PIPP [14] that are defined ....

....that are defined on the space of partial plans, rewrite rules are defined on the space of complete plans. In addition, it has been argued that plan rewrite rules are easier to state than search control rules, because they do not require any knowledge of the inner workings of the planning algorithm [1]. That may partially explain why most of the search control systems have been designed to automatically acquire search control rules, whereas existing planning by rewriting systems use manually generated rewrite rules. To date, there has been no comparison of these two techniques to study their ....

[Article contains additional citation context not shown here]

J. Ambite and C. Knoblock. Planning by rewriting: Efficiently generating highquality plans. In AAAI-97, Menlo Park, 1997. AAAI Press.

....all possible kinds of changes in the constraint graph during the search process not only after the current graph has been completely searched, and not only along an increasing plan length. Unlike other planning systems that use productions rules to change the plan as part of the search (e.g. [Ambite Knoblock 1997]) modeling planning as an SCSP allows us to specify the problem in a declarative manner and to derive the corresponding productions automatically. This automatic method guarantees that local search techniques 5 See also other articles in this book. can potentially find all valid plans, which is ....

Ambite, J. L., and Knoblock, C. A. 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. AAAI-97, 706--713.

....the depth of the search tree (c) planning cases that can be used to solve similar problems in future. Recent work [4, 11] has shown that most of the techniques used for learning to improve planning efficiency can be extended to learn meta knowledge to improve plan quality. Planning by rewriting [1, 10] is another technique that has been proposed for improving plan quality. This approach suggests using a planner to produce a (possibly low quality) plan and then applying a set of rewrite rules to convert it into a higher quality plan. The ultimate aim of our work is to develop a domain ....

....produce good quality solutions is essential if AI planners are to be widely applied to the real world situations. However, conventional wisdom in AI has been that domain independent planning is a hard combinatorial problem. Taking into account plan quality makes the task even more difficult [1]. This paper has presented a mixedinitiative planning technique for learning local search control rules for partial order planners that improve plan quality without sacrificing much in the way of planning efficiency. We believe that these ideas will contribute towards making AI planning applicable ....

J. Ambite and C. Knoblock. Planning by rewriting: Efficiently generating high-quality plans. In Proc. of AAAI, Menlo Park, 1997. AAAI Press.

.... search strategies have been used, including special purpose algorithms and combinations of branch and bound search with random techniques such as random walk, iterative improvement and simulated annealing [38] SIMS also uses rewriting techniques to optimise its partially ordered query plans [39]. When queries need to be split across multiple databases there are two additional considerations in query optimisation. First, the query processor may have information about the capabilities of the individual component databases that can be used to optimise the subqueries sent to them. For ....

J. L. Ambite and C. A. Knoblock. Planning by rewriting: Efficiently generating high-quality plans. In Proceedings of the 14th National Conference on Artificial Intelligence (AAAI'97), 1997.

....all possible kinds of changes in the constraint graph during the search process not only after the current graph has been completely searched and not only along an increasing plan length. Unlike other planning systems that use productions rules to change the plan as part of the search (e.g. (Ambite Knoblock 1997)) modeling planning as an SCSP allows us to specify the problem in a declarative manner and to derive the corresponding productions automatically. This automatic method guarantees that local search techniques can potentially find all valid plans, which is not normally the case with manual ....

Ambite, J. L., and Knoblock, C. A. 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. AAAI-97, 706--713.

....constraints. If the conjunction of all the user constraints with all the resource constraints is satisfiable, then the resource contains data which are relevant to the user request. A somewhat related research area is the research on information mediators among heterogenous information systems[21, 1]. Each local information system is wrapped by a so called wrapper agent and their capabilities are described in two levels. One is what they can provide, usually described in the local data model and local database schema. Another is what kind of queries they can answer; usually it is a subset of ....

....matching can result from utilizing different combinations of these filters. Selection of filters to apply is under the control of the user (or the requester agent) Acknowledgement: We thank Davide Brugali, Somesh Jha and Anandeep Pannu for their helpful discussions in this project. References [1] J.L. Ambite and C.A. Knoblock. Planning by Rewriting: Efficiently Generating High Quality Plans. Proceedings of the Fourteenth National Conference on Artificial Intelligence, Providence, RI, 1997. 2] K. Decker, K. Sycara, M. Williamson. Middle Agents for the Internet. Proc. 15th IJCAI, pages ....

[Article contains additional citation context not shown here]

J.L. Ambite and C.A. Knoblock. Planning by Rewriting: Efficiently Generating High-Quality Plans. Proceedings of the Fourteenth National Conference on Artificial Intelligence, Providence, RI, 1997.

....also uses a branch and bound approach in which it iteratively builds, compares, and eliminates partial plans to find the optimal plan. It differs from Drips and Streamer in that partial plans are not constructed using abstraction and it considers simpler utility models. Planning by rewriting (Ambite Knoblock 1997) finds approximately optimal plans by generating an initial solution plan, then rewriting the current solution plan in order to improve its quality using a set of declarative plan rewriting rules. All of the planning approaches described above address a different problem from ours: they focus on ....

Ambite, J., and Knoblock, C. 1997. Planning by rewriting: Efficiently generating high-quality plans. In AAAI '97.

....minimal perturbation. Ideally, a process management system should provide a spectrum of plan repair mechanisms ranging from the correct but costly minimal perturbation, correctnesspreserving methods to transformational approaches that employ domain specific transformation rules (in the spirit of (Ambite Knoblock 1997; Howe 1995) that may trade correctness for efficiency. Complementing the work on plan repair are efforts to develop plans that are less likely to fail in the first place. This work is in the early stages but will become increasingly important as plans are generated that are to be executed in ....

Ambite, J. L., and Knoblock, C. A. 1997. Planning by rewriting: Efficiently generating high-quality plans. In Proceedings of the Fourteenth National Conference on Artificial Intelligence.

....[7, 5, 8] and as well as quality of the plans they produce [14, 6] Traditionally, this knowledge is encoded as search control rules to limit the search for generation of the first viable plan. Recently, Ambite and Knoblock have suggested an alternative approach called planning by rewriting [1]. Under this approach, a partial order planner generates an initial plan, and then a set of rewrite rules are used to transform this plan into a higher quality plan. Unlike the search control rules for partial order planners (such as those learned by UCPOP EBL [7] and PIPP [17] that are defined ....

....that are defined on the space of partial plans, rewrite rules are defined on the space of complete plans. In addition, it has been argued that plan rewrite rules are easier to state than search control rules, because they do not require any knowledge of the inner workings of the planning algorithm [1]. That may partially explain why most of the search control systems have been designed to automatically acquire searchcontrol rules, whereas existing planning by rewriting systems use manually generated rewrite rules. To date, there has been no comparison of these two techniques to study their ....

[Article contains additional citation context not shown here]

J. Ambite and C. Knoblock. Planning by rewriting: Efficiently generating high-quality plans. In Proceedings of the Fourteenth National Conference on Artificial Intelligence, Menlo Park, CA, 1997. AAAI Press.

....of the information in the KS, including: a fine grained semantic description and an associated cost for each type of data function available (cost information could be learned and recorded by the agent) domain knowledge about the relations between data; and a reasoning system. As an example, (Knoblock Ambite 1997) present a query that retrieves ship types whose range is greater than 10,000 miles. Given the presence of domain knowledge that ships with this range have a draft over twelve feet, and cost knowledge indicating that getting draft information is cheaper than range information, SQO converts the ....

....or logical information that geo loc(x) resort loc(x) resort loc(x) but again, this computation is expensive, and could be prohibitively so depending on the amount of domain knowledge present. Attempts are being made to reduce the negative impacts of query re mappings. Planning by rewriting (Knoblock Ambite 1997) generates low cost but possibly sub optimal plans. Another effort, recursive information gathering (Duschka Levy 1997) generates a maximally contained set of query rewritings (the set contains information subsets of the query, not just equivalents) By exploiting functional dependencies in ....

Knoblock, C., and Ambite, J. 1997. Planning by rewriting: efficiently generating high-quality plans. In Proceedings of the Fourteenth National Conference on Artificial Intelligence.

....planners, during the last couple of years we witnessed the occurrence of new type of planning systems: the stochastic planners. This new approach to AI planning trades in the completeness of the planner for the speed up of the search process. Planners like SatPlan [Kautz and Selman 1996] or PBR [Ambite and Knoblock 1997] are at least one order of magnitude faster than the classic planning systems, and they are also capable of handling significantly larger problem instances. In this paper we present SINERGY, which is a general purpose, stochastic planner based on the genetic programming paradigm [Koza 1992] ....

Ambite, J.L., Knoblock, C.: Planning by Rewriting: Efficiently Generating HighQuality Plans. In Proceedings of AAAI-97 706-713, 1997.

....minimal perturbation. Ideally, a process management system should provide a spectrum of plan repair mechanisms ranging from the correct but costly minimal perturbation, correctness preserving methods to transformational approaches that employ domain specific transformation rules (in the spirit of [2, 40]) that may trade correctness for efficiency. 6.3 Robust Process Design Complementing the work on plan repair are efforts to develop plans that are less likely to fail in the first place. This work is in the early stages but will become increasingly important as plans are generated that are to be ....

J. L. Ambite and C. A. Knoblock. Planning by rewriting: Efficiently generating highquality plans. In Proceedings of the Fourteenth National Conference on Artificial Intelligence, 1997.

....will be stored separately by extending each concept name by the identifier of the agent who did send this concept. To summarize, the context matching consists of two consecutive steps: 1. For every pair of words u; v given in the context slots compute the real valued word distances dw (u; v) 2[0,1]. Determine the most similar matches for any word u by selecting words v with the minimum distance value dw (u; v) These distances must not exceed a given threshold. 2. For every pair of most similar matching words, check that the semantic distance among the attached concepts does not exceed a ....

....constraints. If the conjunction of all the user constraints with all the resource constraints is satisfiable, then the resource contains data which are relevant to the user request. A somewhat related research area is the research on information mediators among heterogenous information systems[22][1]. Each local information system is wrapped by a so called wrapper agent and their capabilities are described in two levels. One is what they can provide, usually described in the local data model and local database schema. Another is what kind of queries they can answer; usually it is a subset ....

J.L. Ambite and C.A. Knoblock. Planning by Rewriting: Efficiently Generating High-Quality Plans. Proceedings of the Fourteenth National Conference on Artificial Intelligence, Providence, RI, 1997.

....If the conjunction of all the user constraints with all the resource constraints is satisfiable, then the resource contains data that are relevant to the user request. Broker and matchmaker agents can be seen as a kind of so called mediator agents among heterogeneous information systems[21, 1]. Each local information system is wrapped by a wrapper agent, and their capabilities are described in two levels. The first is what they can provide, which is usually described in the local data model and local database schema. The second is what kind of queries they can answer, which is ....

J.L. Ambite and C.A. Knoblock. Planning by Rewriting: Efficiently Generating High-Quality Plans. Proceedings of the Fourteenth National Conference on Artificial Intelligence, Providence, RI, 1997.

....method is formally incomplete, but in practice it can efficiently solve problems that are very hard to solve by more traditional systematic methods. 1 Though local search techniques have been applied with success to many combinatorial problems, they have only recently been applied to planning (Ambite Knoblock 1997; Kautz Selman 1998; 1996; Serina Gerevini 1998) In particular, Kautz and Selman experimented the use of a stochastic local search algorithm (Walksat) in the context of their planning as satisfiability framework, showing that Walksat outperforms more traditional systematic methods on ....

Ambite, J. L., and Knoblock, C. A. 1997. Planning by rewriting: Efficiently generating high-quality plans.

....that in [5] Given a plan oe, while Foulser, Li, and Yang would look for the optimal, e.g. B minimal subplan of oe, we are only concerned with obtaining a B minimal subplan, i.e. for us, one B minimal plan is as good as any other ones. Our work is also related to that by Ambite and Knoblock [1] who propose a planning strategy that divides the planning process into two steps: initial plan generation, and plan refinement. While much of [1] is about domain specific rules that would transform a given plan into a better one, we are mainly interested in domain independent measures here. It ....

....obtaining a B minimal subplan, i.e. for us, one B minimal plan is as good as any other ones. Our work is also related to that by Ambite and Knoblock [1] who propose a planning strategy that divides the planning process into two steps: initial plan generation, and plan refinement. While much of [1] is about domain specific rules that would transform a given plan into a better one, we are mainly interested in domain independent measures here. It would be an interesting future research topic to see how useful a domain independent notion of plan quality such as one of those studied here ....

J. L. Ambite and C. A. Knoblock. Planning by rewriting: efficiently generating high-quality plans. In Proceedings of the 14th National Conference on Artificial Intelligence (AAAI--97), AAAI Press, Menlo Park, CA., pages 706--713, 1997.

....as well as quality of the plans they produce (Perez 1996; Iwamoto 1994) Traditionally, this knowledge is encoded as search control rules to limit the search for generation of the first viable plan. Recently, Ambite and Knoblock have suggested an alternative approach called planning by rewriting (Ambite Knoblock 1997). Under this approach, a partial order planner generates an initial plan, and then a set of rewrite rules are used to transform this plan into a higher quality plan. Unlike the search control rules for partial order planners (such as those learned by UCPOP EBL (Kambhampati, Katukam, Qu 1996) and ....

....that are defined on the space of partial plans, rewrite rules are defined on the space of complete plans. In addition, it has been argued that plan rewrite rules are easier to state than search control rules, because they do not require any knowledge of the inner workings of the planning algorithm (Ambite Knoblock 1997). That may partially explain why most of the search control systems have been designed to automatically acquire searchcontrol rules, whereas existing planning by rewriting systems use manually generated rewrite rules. To date, there has been no comparison of these two techniques to study their ....

[Article contains additional citation context not shown here]

Ambite, J., and Knoblock, C. 1997. Planning by rewriting: Efficiently generating high-quality plans. In Proc. of the Fourteenth National Conference on Artificial Intelligence. Menlo Park, CA: AAAI Press.

....of the information in the KS, including: a fine grained semantic description and an associated cost for each type of data function available (cost information could be learned and recorded by the agent) domain knowledge about the relations between data; and a reasoning system. As an example, (Knoblock and Ambite, 1997) present a query that retrieves ship types whose range is greater than 10,000 miles. Given the presence of domain knowledge that ships with this range have a draft over twelve feet, and cost knowledge indicating that getting draft information is cheaper than range information, SQO converts the ....

....or logical information that geo loc(x) resort loc(x) resort loc(x) but again, this computation is expensive, and could be prohibitively so depending on the amount of domain knowledge present. Attempts are being made to reduce the negative impacts of query re mappings. Planning by rewriting (Knoblock and Ambite, 1997) generates low cost but possibly sub optimal plans. Another effort, recursive information gathering (Duschka and Levy, 1997) generates a maximally contained set of query rewritings (the set contains information subsets of the query, not just equivalents) By exploiting functional dependencies in ....

C. Knoblock and J. Ambite. 1997. Planning by rewriting: efficiently generating high-quality plans. In Proceedings of the Fourteenth National Conference on Artificial Intelligence, Providence, Rhode Island, July.

.... a wide variety of solutions, from search control rules [3] to hierarchical planning [8] to skeletal planning [5] More recently, we witnessed the occurrence of the stochastic planners, which trade in the completeness of the planner for the speed up of the search process (e.g. SatPlan [7] or PBR [1] are at least one order of magnitude faster than the classic planning systems) In this paper we present SYNERGY, which is a general purpose, stochastic AI planner based on the genetic programming paradigm [9] Genetic programming (GP) is an automatic programming technique that uses evolution like ....

Ambite, J.L., Knoblock, C. "Planning by Rewriting: Efficiently Generating High-Quality Plans." In Proceedings of AAAI-97 706-713, 1997.

....planning (Friedland and Iamasaki 1985) More recently, we witnessed the occurrence of a new type of planning systems: the stochastic planners. The new approach trades in the completeness of the planner for the speed up of the search process. Planners like SatPlan (Kautz and Selman 1996) or PBR (Ambite and Knoblock 1997) are at least one order of magnitude faster than the classic planning systems, and they are also capable of handling significantly larger problem instances. In this paper we present SINERGY, which is a general purpose, stochastic AI planner based on the genetic programming paradigm (Koza 1992) ....

Ambite, J.L. and C. Knoblock 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. To appear in Proceedings of the Fourteenth National Conference on Artificial Intelligence (AAAI97).

....time pressured environment where goals can change due to information gathered during the dialog. Batch planning is inappropriate because, at any single time, too much information about the conversation is lacking to construct a useful complete plan. Recent planning systems relying on local search [1] and action selection [2] provide further evidence for the validity of this approach. 3 The Next action Problem Our main goal is to develop algorithms for enabling an agent to choose which action it should pursue next. Previously, we have studied this in the context of conversational turn taking, ....

Jos'e Ambite and Craig Knoblock. Planning by rewriting: Efficiently generating highquality plans. In Proceedings of AAAI-97, pages 706--713, 1997.

....also uses a branch and bound approach in which it iteratively builds, compares, and eliminates partial plans to find the optimal plan. It differs from Drips and Streamer in that partial plans are not constructed using abstraction and it considers simpler utility models. Planning by rewriting (Ambite Knoblock 1997) finds approximately optimal plans by generating an initial solution plan, then rewriting the current solution plan in order to improve its quality using a set of declarative plan rewriting rules. All of the planning approaches described above address a different problem from ours: they focus on ....

Ambite, J., and Knoblock, C. 1997. Planning by rewriting: Efficiently generating high-quality plans. In AAAI '97.

....axiom lattice. Planning By Rewriting. When queries are posed to the system, Ariadne reasons about the domain model and source descriptions in order to develop an efficient plan for retrieving and integrating the data. The method used to accomplish this is called Planning by Rewriting (PBR) (Ambite and Knoblock 1997). Under PBR, an initial, sub optimal plan is quickly generated and then iteratively improved by applying a series of rewriting rules. Rewriting relies on local search algorithms that can alter both the sources used to resolve portions of a query as well as the ordering of operations performed by ....

Ambite, J.L. and Knoblock, C.A. 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. AAAI-97, Providence, RI.

....such as replanning after failures and information gathering actions. In this paper, we present an approach to query planning in mediators that addresses these challenges. Our query planner is based on a general paradigm for efficient high quality planning called Planning by Rewriting (PbR) [4]. This planning style uses declarative plan rewriting rules and efficient local search techniques to transform an easy to generate, but possibly suboptimal, initial plan into a high quality plan. PbR was designed to address planning efficiency and plan quality, while providing the benefits of ....

....Fig. 8. Optimized query plan. seaport(country code country name geoloc code port name) # seaport(geoloc code port name) # location(country code geoloc code) # country(country code country name) Fig. 9. Sample integration axiom. 4. Review of planning by rewriting Planning by Rewriting [4] follows the iterative improvement style of many optimization algorithms. The framework works in two phases: 1) Efficiently generate an initial solution plan. 2) Iteratively rewrite the current solution plan in order to improve its quality using a set of declarative plan rewriting rules until ....

[Article contains additional citation context not shown here]

J.L. Ambite, C.A. Knoblock, Planning by rewriting: Efficiently generating high-quality plans, in: Proc. AAAI-97, Providence, RI, 1997.

No context found.

Ambite, J.L. and Knoblock, C.A. Planning by Rewriting: Efficiently Generating High-Quality Plans. Proc of the 14 th Natl Conf on Artificial Intelligence, Providence, RI. 1997

....on the operations to process the data. Moreover, the choice of sources, data processing operations, and their ordering, strongly affects the plan cost. In this paper, we present an approach to query planning in mediators based on a general planning paradigm called Planning by Rewriting (PbR) [4]. Our work yields several contributions. First, our PbR based query planner combines both the selection of the sources and the ordering of the operations into a single search space in which to optimize the plan quality. Second, by using local search techniques our planner explores the combined ....

....such as replanning after failures and information gathering actions. In this paper, we present an approach to query planning in mediators that addresses these challenges. Our query planner is based on a general paradigm for efficient high quality planning called Planning by Rewriting (PbR) [4]. This planning style uses declarative plan rewriting rules and efficient local search techniques to transform an easy to generate, but possibly suboptimal, initial plan into a high quality plan. PbR was designed to address planning efficiency and plan quality, while providing the benefits of ....

[Article contains additional citation context not shown here]

J. L. Ambite and C. A. Knoblock. Planning by rewriting: Efficiently generating high-quality plans. In Proceedings of the Fourteenth National Conference on Artificial Intelligence, Providence, RI, 1997.

....based on comparing initial and optimal plans. We report results for several planning domains showing that the learned rules are competitive with manually specified ones, and in several cases the learning algorithm discovered novel rewriting rules. Introduction Planning by Rewriting (PbR) (Ambite Knoblock 1997; 1998; Ambite 1998) is a planning framework that has shown better scalability than other domainindependent approaches. In addition, PbR works with complex models of plan quality and has an anytime behavior. The basic idea of PbR is to first generate a possibly suboptimal initial plan, and then, ....

Ambite, J. L., and Knoblock, C. A. 1997. Planning by rewriting: Efficiently generating high-quality plans. In Proceedings of the Fourteenth National Conference on Artificial Intelligence.

....1. INTRODUCTION Gathering information from the World Wide Web is a research problem that has received substantial attention in recent years. There now exist a number of systems [7, 10, 13] and approaches towards automating this process, including work on data extraction [11, 14] query planning [1, 12], data materialization [2] and methods for handling data inconsistency [4] Today, it is possible to construct useful agents that rely on these technologies as tools to perform automatic and intelligent data integration [3] Although these individual technologies may each be efficient, overall ....

Ambite, J.L. and Knoblock, C.A. 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. AAAI-1997.

....1. INTRODUCTION Gathering information from the World Wide Web is a research problem that has received substantial attention in recent years. There now exist a number of systems [7, 10, 13] and approaches towards automating this process, including work on data extraction [11, 14] query planning [1, 12], data materialization [2] and methods for handling data inconsistency [4] Today, it is possible to construct useful agents built on these technologies to perform automatic and intelligent data integration [3] Although these individual technologies may each be efficient, overall end to end ....

....or automatically generated by the userinterface) may be sub optimal. In future work, we would like to explore the improvement of these suboptimal plans. In previous work on Ariadne, a highly efficient and scalable approach to plan optimization was demonstrated using Planning By Rewriting (PBR) [1], a local search approach to anytime plan refinement. We are investigating the applicability of PBR to our optimization needs. We are also exploring ways to improve the scalability and throughput of Theseus in a global context. Specifically, we would like to be able to deploy Theseus as an ....

Ambite, J.L. and Knoblock, C.A. 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. AAAI-97.

....Gathering information from the World Wide Web is a research problem that has been receiving substantial attention in recent years. There now exist a number of promising systems [9, 13, 14] and approaches towards automating this process, including work on data extraction [15, 17] query planning [1, 16], data materialization [2] and methods for handling data inconsistency [3] While gathering data is unquestionably an important task, there are also challenges related to the effective management and use of this data. We believe that information gathering is a piece of a larger puzzle called ....

Ambite, J.L. and Knoblock, C.A. 1997. Planning by Rewriting: Efficiently Generating High-Quality Plans. Proceedings of the Fourteenth National Conference on Artificial Intelligence.

....are on flexible and efficient query processing in mediator systems. In the context of the SIMS (Arens et al. 1993; Arens, Knoblock, Shen 1996) and Ariadne (Knoblock et al. 1998) projects, we have applied a general framework for efficient highquality planning, called Planning by Rewriting (Ambite Knoblock 1997), to the problem of generating query plans in distributed and heterogeneous environments (Ambite Knoblock 1998) SIMS and Ariadne are mediator systems that provide integrated access to heterogeneous sources in an application domain by building a model for the domain and mapping the contents of ....

....plan is small enough, even if a cheaper one could be found. Finally, the generality of the PbR framework has allowed the design of a novel combination of traditional query optimization and source selection. I am also interested in the scalability of systems with multiple mediators (Knoblock Ambite 1997), maintaining accurate source descriptions in a mediator (Ambite Knoblock 1995) and general issues of knowledge representation and reasoning for information integration. ....

Ambite, J. L., and Knoblock, C. A. 1997. Planning by rewriting: Efficiently generating high-quality plans.

....efficiently and dynamically select sources based on the classes and attributes mentioned in the query. A paper on this topic is under submission to SIGMOD. Once the sources have been selected, Ariadne generates a plan using a method called Planning byRewriting, developed by Ambite and Knoblock (Ambite and Knoblock 1997). This approach takes an initial, suboptimal plan and then attempts to improve it by applying rewriting rules. In the case of query planning, producing an initial, suboptimal plan is straightforward; we can generate an initial plan in O(n) time, where n is the length of the query, based on a ....

J. Ambite and C.A. Knoblock. Planning by rewriting: Efficiently generating high-quality plans. In AAAI-97, 1997.

....finding any valid plan is not enough, plan quality is also critical. Finally, mediators need to incorporate traditional techniques for query planning in databases and extend them with new capabilities, such as replanning after failures and information gathering actions. The Planning by Rewriting (Ambite Knoblock 1997) paradigm is designed to address planning efficiency and plan quality, while providing the benefits of domain independence. Its characteristics make it especially well suited for query planning. First, PbR Copyright c fl1998, American Association for Artificial Intelligence (www.aaai.org) All ....

....processing operators and the selection of relevant information sources for terms in a given query. Mediators need to provide mechanisms to resolve the semantic heterogeneity among the different sources. Our approach follows that of the SIMS mediator system (Arens, Knoblock, Shen 1996; Knoblock Ambite 1997). Briefly, SIMS assumes that a set of information sources such as databases, knowledge bases, web servers, etc. supply data about a particular application domain. The system designer specifies a global model of the application domain and defines the contents of the sources in terms in this global ....

[Article contains additional citation context not shown here]

Ambite, J. L., and Knoblock, C. A. 1997. Planning by rewriting: Efficiently generating high-quality plans.

....et al. 1998) preprocesses the domain model so that the system can efficiently and dynamically select sources based on the classes and attributes mentioned in the query. In the second phase, Ariadne generates a plan using a method called Planning by Rewriting, developed by Ambite and Knoblock (Ambite and Knoblock 1997; 1998) This approach takes an initial, suboptimal plan and then attempts to improve it by applying rewriting rules. In the case of query planning, producing an initial, suboptimal plan is straightforward; we can generate an initial plan in O(n) time, where n is the length of the query, based on ....

Ambite, J.L. and Knoblock, C.A. 1997. Planning by rewriting: Efficiently generating high-quality plans.

No context found.

J.e.L. Ambite and C.A. Knoblock, "Planning by Rewriting: Efficiently Generating High-Quality Plans," in Proc. 14 National Conf. on Artificial Intelligence, 1997.

No context found.

J.L.Ambite, C.A.Knoblock. "Planning by rewriting: Efficiently generating high-quality plans". In proceedings of the Fourteenth National Conference on Artificial Intelligence. 1997.

No context found.

J. L. Ambite and C. A. Knoblock. Planning by rewriting: Efficiently generating highquality plans. In Proceedings of the 14th National Conference on Artificial Intelligence (AAAI'97), 1997.

No context found.

J.L.Ambite, C.A.Knoblock. "Planning by rewriting: Efficiently generating high-quality plans". In proceedings of the Fourteenth National Conference on Artificial Intelligence. 1997.

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