Data-Driven Educational Data Mining ---- the Progress of Project Presenter: Jun Ge Supervisors: Jimmy Huang, Marina G Erechtchoukova Partner: Shore Consulting Group Inc. Joseph Siahou, Scott McPhee
Agenda Introduction Content based Semantic Enhanced Recommendation System Basic Approaches Methodology Challenges Query Expansion based on Domain Taxonomy General process and key point Trial result Challenges and further steps Reference Acknowledgement
Introduction Basic info Data: Curricula in several Canadian provinces Course resources, e.g. textbook, PPT, and pdf file etc., Teachers’ demographic information The main idea is to connect the government, teachers and students with a system which can help teachers to find resources that meet government curricula standards and also can help students to improve their performance.
Recommendation System----Basic Approaches Content-based filtering Recommend items similar to those a given user has liked in the past Collaborative filtering Rrecommendation based on the ratings or behavior of other users in the system Hybrid approach
Recommendation System----Methodology Using semantics and a taxonomy to enhance personalized recommending process Key Component Item Representation User Profile Strategy for matching the two above
Recommendation System Advantages: Broadening the recommendation set by integrating semantics in the personalization process New items have higher possibilities to be recommended Personalized Problems and Challenges: Lack of data -- user profile and user interacts Unstructured data: Raw pdf files, no attributes with well-defined values Complexity of natural language Not a state-of-the-art technology
Query Expansion Based on Domain Taxonomy General information retrieval process Use Google as search engine Focus on new query construction
Query Expansion Key point: Context: system records( e.g. history queries, viewed documents) user background( e.g. domain specific info) task-related context( e.g. subject domain) Strategies: according to how the contextual info acquired Relevance feedback: users make relevance judgments Knowledge models: ontology( e.g. domain specific)
Query Expansion----Bloom’s Taxonomy What is Bloom’s Taxonomy? Created in 1956 Promote higher forms of thinking in education used when designing educational, training, and learning processes
Query Expansion----Methodology Implement result: implemented
Query Expansion Advantages: Challenges: Task-related Heuristic, easy to implement Challenges: How to assign weights to each term in the new query? How to utilize the categories of Bloom Taxonomy? How to evaluate the performance?
Reference Lops, P., De Gemmis, M., & Semeraro, G. (2011). Content-based recommender systems: State of the art and trends. In Recommender systems handbook (pp. 73-105). Springer US. Eirinaki, M., Vazirgiannis, M., & Varlamis, I. (2003, August). SEWeP: using site semantics and a taxonomy to enhance the Web personalization process. In Proceedings of the ninth ACM SIGKDD international conference on Knowledge discovery and data mining (pp. 99-108). ACM. Pazzani, M. J., & Billsus, D. (2007). Content-based recommendation systems. In The adaptive web (pp. 325-341). Springer Berlin Heidelberg. Schafer, J. B., Frankowski, D., Herlocker, J., & Sen, S. (2007). Collaborative filtering recommender systems. In The adaptive web (pp. 291-324). Springer Berlin Heidelberg. Guo, X., Zhang, G., Chew, E., & Burdon, S. (2005, December). A hybrid recommendation approach for one-and-only items. In Australasian Joint Conference on Artificial Intelligence (pp. 457-466). Springer Berlin Heidelberg. Bhogal, J., MacFarlane, A., & Smith, P. (2007). A review of ontology based query expansion. Information processing & management, 43(4), 866-886. Wu, J., Ilyas, I., & Weddell, G. (2011). A study of ontology-based query expansion. Technical report CS-2011–04.
Acknowledgement The project is conducted in collaboration with the industrial partner - Shore Consulting Group Inc. We are thankful to Big Data Research and Analytics Information Network for continuing financial support. This work is also supported by an ORF-RE (Ontario Research Fund - Research Excellence) award in BRAIN Alliance.1 ------------------------------- 1. http://brainalliance.ca