2006 Fiscal Year Final Research Report Summary
Web Warehouse Construction Method for Aggregating On-line Spatio-temporal Information
| Project/Area Number |
16500048
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Media informatics/Database
|
|---|
| Research Institution | Nagoya University (2006)
University of Tsukuba (2004-2005) |
|---|
Principal Investigator |
ISHIKAWA Yoshiharu Information Technology Center, Nagoya University, Professor, 情報連携基盤センター, 教授 (80263440)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KITAGAWA Hiroyuki Graduate School of Systems and Information Engineering, University of Tsukuba, Professor, 大学院システム情報工学研究科, 教授 (00204876)
|
|---|
| Project Period (FY) |
2004 – 2006
|
| Keywords | moving objects / spatio-temporal databases / web mining / document clustering / novelty of information / information extraction / histograms / information retrieval interface |
|---|
| Research Abstract |
The research results can be summarized as follows:
1. Query processing and mining methods for moving object databases to analyze movement status: Vehicles equipped with GPS devices and walking people with mobile terminals are called moving objects. A database that stores and manages movement history data of a lot of moving objects is called a moving object database. First, we have developed an efficient query processing technique for a statistical query that requests the occurrence probability of the movement histories that obey the user specified movement pattern. Second, we have developed and evaluated a mobility histogram method that accumulates continuously delivered movement trajectory data and represents movement statistics in a compact manner.
2. Document clustering method based on the notion of novelty: One of the features of the documents delivered on the Internet such as news articles is that the value of a document is the highest when it is obtained and decreases as time passe
… More
s. The document clustering technique proposed by ours has the feature that it weights high importance on novel documents to generate a clustering result. First, we have constructed the clustering system and performed evaluation based on it. Second, we have constructed a user interface system called T-Scroll on the clustering stem. T-Scroll visualizes each cluster set constructed at each time period and generates links between clusters of neighboring time periods then represents the overall trends of the evolving topics.
3. Discovery of spatial information sources on the Web/Record extraction from a large document set that cooperates with databases: In recent years, research on searching and extracting useful information from large document sources such as the Web is quite active. In this work, first we have developed a method to discover pages and/or sites that are highly related to a specific region or area based on the link analysis technique. Second, we have developed a record extraction method from a large document set cooperating with existing databases. In this approach, the user selects records in databases using queries then the records are provided to the record extraction system as samples. The record extraction system performs selective record extraction based on the sample records and achieves efficient record extraction. Less
|
