Modeling Issues Encountered from Building a Taxonomy from a Biology Textbook

1. Modeling Issues Encountered from Building a Taxonomy from a Biology Textbook A. Patrice Seyeda, John Pachecob, Andrew Goldenkranzb, and VinayChaudhribaDepartment of Computer Science and Engineering, University at Buffalo, Buffalo, New York, bSRI International, Menlo Park, CA Introduction • Task to create a taxonomy from an AP Biology textbook’s glossary terms (Campbell and Reece, 8th Edition) Materials and Methods • Imported ~2400 glossary terms and definition strings from electronic glossary into Collaborative Protégé in OWL format, as classes and comment strings • Use of just the subclass-of relation • Team consisted of biologists and KR specialists • Biologists attempted initial classifications, where modeling issues were identified and solutions provided Subclass/Subevent • There was a strong initial tendency to use the hierarchy to organize sub-parts or sub-processes • E.g., Telophase is a subclass of Mitosis, instead of a sub-process Entity/Process Dichotomy • Biologists wanted to classify Light-Microscope under the subclass Technology • Biologists also wanted to classify Technology under the subclass Inquiry • These two uses of the term `Technology’ refer to two different senses • Glossary definition for Technology: “The application of scientific knowledge for a specific purpose, often involving industry or commerce but also including uses in basic research.” Linnaean Classification • Biologists wanted to classify the different kingdoms under the class Kingdom • However, there are 5 instances of Kingdom Solution • Move parts or sub-processes to appropriate locations whenever they are found • In training sessions, reinforce the subclass of relationship Solution • In this case it was simple to notice the problem and refactor the taxonomy • Some terms are polysemous • Definitions including “; also” are a warning sign Example: Wild Type is ``An individual with the phenotype most commonly observed in natural populations; also refers to the phenotype itself.’’ Results Entity/Role Dichotomy • Initial encoding of organic molecules included classes defined by structure and classes defined by function • Proteins and steroids are defined by their chemical composition • Hormones are defined by the function they perform • There is an overlap between Steroidand Hormone • Some hormones are steroids, others are proteins • Biologists also wanted to relate the different levels of classification of the Linnaean taxonomy to each other in the familiar hierarchical way • However, an instance of Phylumis not an instance of Kingdom Solution • Built Linnaean taxonomy under organism • Used common English names for simplicity • “Moss is a Plant” is clearer than “Moss is a Plantae” • Made the classes that represent the tiers in that hierarchy into “Classification-Units” Solution • Hormones are defined as roles that certain chemicals play • Steroid-Hormone is still a class in the taxonomy which represents a useful class of biologists’ intuitive thinking Classifying Areas of Research • Initial tendency to classify areas of research (Genetics, Anatomy, Ecology) under Inquiry • Areas of research are complex, involving research activities and educational institutions constituted of departments, faculty members, programs and curricula. • However, the definitions of the terms for each area of research is prefixed by ``the scientific study of’’ • Given commitment to processes, classification under Inquiry is appropriate. Conclusions • Initially, biologists relied on prior knowledge and definitions for organizing class hierarchy, classes treated as organizational “buckets” • Ontological principles were applied to identify modeling issues and provide a foundation during taxonomy building process • The biologists now have a much better sense for these types of modeling issues Potential Refinements • Add the latin-named classes as instances of their Classification-Unit Examples: • Animalia instance of Kingdom • Plantae instance of Kingdom • Treat Classification Units as meta-classes • Animal would be an instance of the meta-class Kingdom • Animalia would be a synonym for that class Acknowledgements This work was funded by Vulcan Inc. References Campbell, N. A. and Reece, J. B. (2007). Biology: AP Edition. Pearson, 8th edition.