Transcript Document
Review of CIKM 2009 Da Zhou 2009-11-08 Outline • • • • Flash-based Database Poster Demo Conclusion Flash-based Database • RS-Wrapper: Random Write Optimization for Solid State Drive • A Flexible Simulation Environment for Flash-aware Algorithms • Dynamic In-Page Logging for Flash-Aware B-Tree Index – Gap-Joo Na (Sungkyunkwan University) Sang-Won Lee (Sungkyunkwan University) Bongki Moon (University of Arizona) – Experiments – People Outline • • • • Flash-based Database Posters Demo Conclusion Posters Information Retrieval: Knowledge Management: • • • • • • • • • • • • • • • • • • • • • • • • • • • • Citation Analysis, Social Networks for IR Domain-specific IR Filtering Foundation of Information Retrieval IR Architectures, Scalability and Efficiency IR Evaluation Language Specific IR Machine Learning for IR Multimedia IR Semi-Structured Information Retrieval User Modeling for IR, Search Personalization Web Search, Advertising, Adversarial Advertising and Optimization Classification and Clustering Data pre- and post-Processing Information Extraction Information Filtering and Recommender Systems Knowledge Synthesis and Visualization Large-scale statistical techniques Link and Graph Mining Semantic Techniques Temporal and Spatial data Mining Text Mining Posters • Acronym Extraction – – – – Training data: full name, context Existing method: context-based New idea: Context-based + Links between pages Dataset: web pages in intranet • Review assignment – Motivation: According to research topic leads too much papers are assigned to a few persons – Greedy algorithm – Integer linear programming algorithm – Dataset: Gold standard data Posters • Construct Wikis – Time consuming, Laborious – Snippet extraction and selection • Distance and influence for snippet • What make categories difficult to classify? – Data itself – The characteristics of data Posters • A word clustering approach for language model-based sentence retrieval in question answering system – Word-based model – Class-based model – N-gram model – Add a variance into the formula Posters • A Collaborative Filtering Approach to Ad Recommendation using the Query-Ad Click Graph Discount shoes A.com Cheap shoes B.com Running shoes C.com Posters • URL Normalization for de-duplication of web page • URL 1: t1, t2, t3, t4 • URL 2: t1, t2, t3, t_title • T4= t_title Demo • OfCourse: Web Content Discovery, Classification and Information Extraction for Online Course Materials • XQGen - An Algebra-based XPath Query Generator for MicroBenchmarking Conclusion • Limited improvement on old topics – More factors are considered – Minor improvement for existing method – Applying existing method in wider applications • DB is a small group • DB is closer to system than IR • IR is closer to daily life than DB