Learning Combinatory Categorial Grammars for Plan Recognition
Authors: Christopher Geib, Pavan Kantharaju
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | This work is motivated by past work on CCG learning algorithms for natural language processing, and is evaluated on five well know planning domains. |
| Researcher Affiliation | Collaboration | Christopher W. Geib SIFT LLC. 319 1st Ave. South, Suite 400 Minneapolis, MN 55401 cgeib@sift.net Pavan Kantharaju Department of Computer Science, Drexel University 3141 Chestnut St Philadelphia, PA 19104 pk398@drexel.edu |
| Pseudocode | Yes | However, Algorithm 1 is the high level pseudo code for our learning algorithm, LEXlearn, where Λcur refers to the current lexicon. |
| Open Source Code | No | The paper does not provide any specific statements or links indicating that its source code is open or publicly available. |
| Open Datasets | Yes | The Monroe domain captures plans for disaster relief including, providing medical aid, plowing snow, and clearing debris from roads. We base our domain off the publicly-available Monroe dataset generated by the SHOP2 planner (Nau et al. 2003). |
| Dataset Splits | Yes | For all learning runs, we split the plan traces into 80% training and 20% testing instances. |
| Hardware Specification | No | The paper does not specify any hardware used for running the experiments. |
| Software Dependencies | No | The paper mentions using 'HTNSolver2 planner (Nau et al. 1999)' but does not provide a specific version number for this or any other software component. |
| Experiment Setup | Yes | The parameter τ, used to prune low probability categories, is set to 0.0625 for all our experiments. ... For all experiments, we set the prior root probabilities for all atomic categories and goal category to 0.01 and 0.99, respectively. ... The third parameter is the number of iterations for gradient ascent, Max GA, which is set to five, and finally, the learning parameter, α, was set to 0.001 based on initial experiments. |