Generating Context-Free Grammars using Classical Planning

Authors: Javier Segovia-Aguas, Sergio Jiménez, Anders Jonsson

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We designed two types of experiments: (1) Generation, for computing CFGs compliant with a set of input strings and (2), Recognition for parsing unseen input strings given a CFG. All the experiments are run on an Intel Core i5 3.10 GHz x 4 with 4 GB of RAM, using the classical planner Fast Downward [Helmert, 2006], with the LAMA-2011 configuration, and a planning time limit of 600 seconds. We created six domains that correspond to CFGs with different structure and alphabet. [...] Table 1 shows the results of the Generation tasks. [...] Table 2 shows the results for the Recognition tasks.
Researcher Affiliation Academia Javier Segovia-Aguas1, Sergio Jim enez2, Anders Jonsson 1 1 Universitat Pompeu Fabra, Barcelona, Spain 2 University of Melbourne, Parkville, Australia
Pseudocode Yes Figure 2: Example of a planning program Π = {Π0, Π1} with one auxiliary procedure that produces the string aabbaa. Figure 3: Planning program that represents the CFG in Figure 1(a).
Open Source Code No The paper does not provide any explicit statement about releasing its source code or a link to a code repository for the methodology described.
Open Datasets No The paper states, "We created six domains that correspond to CFGs with different structure and alphabet." These are described as
Dataset Splits No The paper describes generating CFGs from "input strings" and then performing "Recognition for parsing unseen input strings". While this implies distinct sets of data for learning and testing, it does not specify a conventional training/validation/test split with percentages, absolute counts, or references to predefined splits.
Hardware Specification Yes All the experiments are run on an Intel Core i5 3.10 GHz x 4 with 4 GB of RAM
Software Dependencies Yes using the classical planner Fast Downward [Helmert, 2006], with the LAMA-2011 configuration
Experiment Setup Yes All the experiments are run on an Intel Core i5 3.10 GHz x 4 with 4 GB of RAM, using the classical planner Fast Downward [Helmert, 2006], with the LAMA-2011 configuration, and a planning time limit of 600 seconds. The compilation takes as input a CFG generation task Σ, E, m such that |et| z for each et E, a number of program lines n and a stack size ℓ.