Knowledge generation and use based on probabilistic model estimation from large-scale structured data

2014 Fiscal Year Final Research Report

• Project/Area Number: 23300039
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Media informatics/Database
• Research Institution: Kobe University
• Principal Investigator: EGUCHI Koji 神戸大学, システム情報学研究科, 准教授 (50321576)
• Co-Investigator(Kenkyū-buntansha): TAKASU Atsuhiro 国立情報学研究所, コンテンツ科学研究系, 教授 (90216648) ; OHKAWA Takenao 神戸大学, 大学院・システム情報学研究科, 教授 (30223738)
• Project Period (FY): 2011-04-01 – 2015-03-31
• Keywords: 統計モデリング / 混合メンバシップモデル / トピックモデル / 統計的ネットワークモデル / 潜在変数モデル / ギブスサンプリング / パーティクルフィルタ

Outline of Final Research Achievements

We aim to estimate probabilistic latent structure underlying collections of text data and network data with internal structure or external structure. The internal structure indicates, for instance, the attributed word tokens in text data or attributed nodes or edges in network data. The external structure indicates, for instance, the case when each node of network data is associated with a set of text data. We extract "knowledge" that can apply for various real-world problems, by estimating low-dimensional latent structure from a large amount of complexly structured data. We apply our techniques to the problems of information retrieval, recommendation, prediction, and time-series analysis.

Free Research Field

情報学