Sequential Action Patterns in Collaborative Ontology Engineering Projects: A Case-study in the Biomedical Domain

Simon Walk's talk at CIKM '14 about our paper titled "Sequential Action Patterns in Collaborative Ontology Engineering Projects: A Case-study in the Biomedical Domain"

Transcript

1. 1  Graz University of Technology CIKM2014 S C I E N C E  P A S S I O N  T E C H N O L O G Y  Graz University of Technology CIKM2014 Sequential Action Patterns in Collaborative Ontology-Engineering Projects: A Case-Study in the Biomedical Domain Simon Walk1, Philipp Singer2 and Markus Strohmaier2,3 1 Graz University of Technology 2 Gesis – Leibniz Institute for the Social Sciences 3 University of Koblenz
2. 2  Graz University of Technology CIKM2014 2 Introduction & Motivation The importance of collaborative ontology-engineering projects increased over recent years due to an increase in • complexity of the modeled domains • requirements for the resulting ontology No individual is able to single-handedly cover the increased complexity and requirements. Hence, it is crucial to better understand and steer the underlying processes of how users collaboratively work on an ontology (i.e., via predictive models).
3. 3  Graz University of Technology CIKM2014 3 Approach & Objective To that extend we analyzed five collaborative ontology- engineering projects from the biomedical domain to: 1. explore regularities and common patterns in user action sequences 2. fit and select models using Markov chains of varying order 3. predict user actions via the fitted Markov chains Our main objective is to predict future user actions in collaborative ontology-engineering projects.
4. 4  Graz University of Technology CIKM2014 4 Datasets Five collaborative ontology-engineering projects from the biomedical domain with varying sizes of features. Note that all ontologies were created with WebProtégé or derivatives of WebProtégé!
5. 5  Graz University of Technology CIKM2014 5 Types of Action Paths
6. 6  Graz University of Technology CIKM2014 6 Types of Action Paths
7. 7  Graz University of Technology CIKM2014 7 Types of Action Paths
8. 8  Graz University of Technology CIKM2014 8 Types of Action Paths
9. 9  Graz University of Technology CIKM2014 9 Extracted Action Paths 1. Users for Classes  Sequences of users that changed a class. 2. Change Types for Users & Classes  Sequences of change types performed by a user / on a class. 3. Properties for Users & Classes  Sequences of properties changed by a user / for a class.
10. 10  Graz University of Technology CIKM2014 10 Exploring Regularities and Sequential Patterns
11. 11  Graz University of Technology CIKM2014 11 Exploring Regularities Randomness & Regularities Wald-Wolfowitz runs test  Adapted by O’Brien and Dyck (1985)  For ~60% of our paths, regularities could be detected.1 Sequential Pattern Mining PrefixSpan to investigate commonly used sequential patterns. Only immediately succeeding states build patterns.  E.g., “A B C” contains “A B” and “B C” but not “A C” 1https://github.com/psinger/RunsTest
12. 12  Graz University of Technology CIKM2014 12 Results for the Sequential Pattern Analysis Users for Classes Paths
13. 13  Graz University of Technology CIKM2014 13 Results for the Sequential Pattern Analysis Users for Classes Paths
14. 14  Graz University of Technology CIKM2014 14 Model Fitting & Selection
15. 15  Graz University of Technology CIKM2014 Modeling Fitting  Markov chains are stochastic processes representing transition probabilities between a countable number of known states.  A state space: listing all possible states  A transition matrix: listing all transition-probabilities between states  A Markov chain of n-th order means that n previous states contain predictive information about the next state.
16. 16  Graz University of Technology CIKM2014 16 Modeling Fitting & Selection We fitted Models from orders of zero to five.2  Lower order models are nested within higher order models.  Higher orders need exponentially more parameters and may result in overfitting. Bayesian model selection (Singer et al. 2014)2  Higher order models receive a penalty due to higher complexity. 2 https://github.com/psinger/PathTools
17. 17  Graz University of Technology CIKM2014 17 Results Bayesian Model Selection
18. 18  Graz University of Technology CIKM2014 18 Predicting User Actions
19. 19  Graz University of Technology CIKM2014 19 K-Fold Cross-Fold Prediction Experiment 1. Fit Markov chain model.  Split Paths into training and test set (stratified).  Rank transitions for each row in the transition matrix. 1. Determine position of test set transition in the fitted Markov chain model. 1. Calculate average over all positions.  Average Position of 1 equals best prediction accuracy.
20. 20  Graz University of Technology CIKM2014 20 K-Fold Cross-Fold Prediction Results
21. 21  Graz University of Technology CIKM2014 21 Results for the Prediction Task
22. 22  Graz University of Technology CIKM2014 22 Conclusions  A number of sequences were produced in a non- random way and frequent patterns can be extracted.  Memory effects (serial dependence) can increase prediction accuracy.  The resulting prediction models can (potentially) be used for  the creation of various recommendations as well as  to assess the impact of potential changes on the ontology and the community.
23. 23  Graz University of Technology CIKM2014 23 Future Work  Include additional data sources (e.g., Semantic MediaWikis).  Analyze higher order patterns and compare patterns of different data sources  Conduct live-lab experiments with generated prediction-models (recommendations).
24. 24  Graz University of Technology CIKM2014 24 Questions?
25. 25  Graz University of Technology CIKM2014  Graz University of Technology CIKM2014 Thank you for your attention!
26. 26  Graz University of Technology CIKM2014 26 References Wald and J. Wolfowitz. On a test whether two samples are from the same population. The Annals of Mathematical Statistics, 11(2):147–162, 1940. P. C. O’Brien and P. J. Dyck. A runs test based on run lengths. Biometrics, pages 237–244, 1985. P. Singer, D. Helic, B. Taraghi, and M. Strohmaier. Detecting memory and structure in human navigation patterns using markov chain models of varying order. PloS one, 9(7):e102070, 2014.