SNLPSymposium on Natural Language Processing
SNLPStatistical Natural Language Processing
SNLPSadie Nash Leadership Project (Brooklyn, NY)
References in periodicals archive ?
The impetus for change came from three alternative approaches to planning, each of which appeared to dominate the SNLP family.
SNLP maintains a plan as a partially ordered set of actions, along with ordering constraints between steps, bindings for variables in steps, and explicit data structures, called causal links, that represent what the step is supposed to accomplish in the plan (Weld 1994).
BURIDAN (Kushmerick, Hanks, and Weld 1995) is a modified version of the SNLP planning algorithm that can create plans that meet a threshold probability of success when actions are nondeterministic, as are the last two actions in figure 1.
This section described decision-theoretic planners based on SNLP, skeletal refinement planning, PRODIGY, and compilation to satisfiability.
For example, is it possible to exploit utility ranges to search for optimal plans in SNLP or PRODIGY?
Algorithms such as TWEAK (Chapman 1987), SNLP (McAllester and Rosenblitt 1991), UCPOP (Penberthy and Weld 1992), and GRAPHPLAN (Blum and Furst 1995) can all be viewed as special-purpose theorem provers aimed at planning problems.
The causal encoding (Kautz, McAllester, and Selman 1996) is based on the causal-link representation used by partial-order planners such as SNLP (McAllester and Rosenblitt 1991).
Causal link planners, for example, SNLP (McAllester and Rosenblitt 1991) and UCPOP (Penberthy and Weld 1992), have received less attention in recent years because they are out-performed by GRAPHPLAN and SATPLAN in most domains.