Unnatural Feature Engineering: Evolving Augmented Graph Grammars for Argument Diagrams

Created by W.Langdon from gp-bibliography.bib Revision:1.8051

@InProceedings{Xue:Unnatural:2016,

• author = "Linting Xue and Collin F. Lynch and Min Chi",
• title = "Unnatural Feature Engineering: Evolving Augmented Graph Grammars for Argument Diagrams",
• booktitle = "Proceedings of the 2016 Conference on Educational Data Mining, EDM16",
• year = "2016",
• editor = "Tiffany Barnes and Min Chi and Mingyu Feng",
• pages = "255--262",
• address = "Raleigh, USA",
• month = jun # " 29-" # jul # " 2",
• publisher = "International Educational Data Mining Society",
• keywords = "genetic algorithms, genetic programming, Evolutionary Computation, Augmented Graph Grammars, Argument Diagramming, Feature Engineering",
• URL = "http://www.educationaldatamining.org/EDM2016/proceedings/paper_137.pdf",
• size = "8 pages",
• abstract = "Graph data such as argument diagrams has become increasingly common in EDM. Augmented Graph Grammars are a robust rule formalism for graphs. Prior research has shown that hand-authored graph grammars can be used to automatically grade student-produced argument diagrams. But hand-authored rules can be time consuming and expensive to produce, and they may not generalize well to novel contexts. We applied Evolutionary Computation to automatically induce empirically-valid graph grammars for argument diagrams that can be used for automatic grading or provide the basis for hints. Our results show that our approach can generate more relevant rules than experts or other state of the art algorithms, and that these evolved rules outperform the alternatives.",
• notes = "http://www.educationaldatamining.org/EDM2016/proceedings.html",
• cv-category = "Peer-Reviewed Conference Paper",

}

Genetic Programming entries for Linting Xue Collin Lynch Min Chi

Citations