A5018935948- Web Data Mining and Analysis
- Advanced Database Systems and Queries
- Data Management and Algorithms
- Semantic Web and Ontologies
- Topic Modeling
- Advanced Text Analysis Techniques
- Advanced Image and Video Retrieval Techniques
- Recommender Systems and Techniques
- Image Retrieval and Classification Techniques
- Natural Language Processing Techniques
- Information Retrieval and Search Behavior
- Text and Document Classification Technologies
- Caching and Content Delivery
- Graph Theory and Algorithms
- Video Analysis and Summarization
- Speech and dialogue systems
- Cloud Computing and Resource Management
- Advanced Data Storage Technologies
- Algorithms and Data Compression
- Advanced Computational Techniques and Applications
- Distributed systems and fault tolerance
- Machine Learning in Bioinformatics
- Complex Network Analysis Techniques
- Electronic Health Records Systems
- Geographic Information Systems Studies
Doshisha University
2015-2024
National Center of Neurology and Psychiatry
2022
Tokyo Institute of Technology
2014
Nara Institute of Science and Technology
2001-2006
Aichi Institute of Technology
2004
University of Tokyo Hospital
2003
Kobe University
1997-2003
Web search engines help users find useful information on the World Wide (WWW). However, when same query is submitted by different users, typical return result regardless of who query. Generally, each user has needs for his/her Therefore, should be adapted to with needs. In this paper, we first propose several approaches adapting results according user's need relevant without any effort, and then verify effectiveness our proposed approaches. Experimental show that systems adapt preferences...
If there is a bias in the number of instances that make up class dataset, predicted results will be affected when applied to machine learning as training data. A method called resampling, which adjusts majority and minority instances, usually used solve imbalance Although resampling can eliminate imbalances, it may cause data complexity deteriorates classification accuracy. Noise overlap are well-known factors complexity. mixture with features classified into other classes at time training,...
In IR (information retrieval) systems based on the vector space model, TF-IDF scheme is widely used to characterize documents. However, in case of documents with hyperlink structures such as Web pages, it necessary develop a technique for representing contents pages more accurately by exploiting their hyperlinked neighboring pages. this paper, we first propose several approaches refining target page using its and then compare retrieval accuracy our proposed approaches. Experimental results...
We propose an efficient distributed query processing method for large XML data by partitioning and distributing to multiple computation nodes. There are several steps involved in this method; however, we focused particularly on dynamic relocation of partitioned our research. Since the efficiency depends both size its structure, these factors should be considered when is partitioned. Each nodes so that CPU load can balanced. In addition, it important take account workload among each because...
One of the research methods food computing that utilizes large amounts recipe data is construction ingredient embeddings. Ingredient embedding expected to be applied a wide range food-related applications because it enables computers handle characteristics ingredients. In existing method contract embedding, was calculated using graph structure ingredients and compounds created based on co-occurrence probability in each recipe. However, this must consider which are used recipes, whether...
We propose an effective classification view mechanism for hypertext data such as Web documents based on Kohonen's self-organizing map (SOM) and search engines. collected by engines are automatically classified SOM the obtained SOMs incrementally modified according to interaction between users SOMs. At present, various designed retrieve documents. When we use get a lot of answers have examine each document. Therefore, in order make up engines, need function classify corresponding user's point...
The TF-IDF scheme is widely used to characterize documents in an information retrieval (IR) system based on the vector space model. However, for having a hyperlink structure such as Web pages, page contents can be characterized more accurately by using of hyperlinked neighboring pages. Therefore, this paper, we propose several techniques pages improve and then verify effectiveness our techniques. © 2005 Wiley Periodicals, Inc. Syst Comp Jpn, 36(14): 56–68, 2005; Published online InterScience...
Purpose Nowadays there are a large number of XML documents on the web. This means that information retrieval techniques for searching very important and necessary internet users. Moreover, it is often said users search engines want to browse only relevant content in each document. Therefore, an effective element aims produce elements or portions Based demand by users, purpose this paper propose evaluate method obtaining more accurate results search. Design/methodology/approach The existing...
We propose a new Web page scoring method based on the link analysis among sets of pages. Conventional analyses such as PageRank and HITS calculate importance degree each page; however, authors pages often create multiple to describe specific topic. The degrees cannot be derived by conventional accurately. To cope with this problem, we need treat same contents edited author set (WPS). After constructing structure WPSs, their using schemes. In paper, compared our approach NTCIR test...
In this paper, we propose a cross-language information retrieval (CLIR) method based on estimating for domains of the query using hierarchic structures Web directories. To get most appropriate translation queries, utilize directories written in many different languages as multilingual corpus disambiguating and domain search results From experimental evaluations, found that there is an advantage accuracy our proposal CLIR system. We it effective to restrict target fields lower level merged...
Currently, many Web sites that provide cooking recipes are available on the Internet. Now, people can upload their original to such sites, so a large number of users able search them and satisfy with information needs. However, it is difficult for find they would like cook because similar have already been uploaded by different people. The fact do not capability recognize difference between reason behind this issue. Therefore, key requirement be well-known recipe from among ones, therefore,...
The use of recommender systems to support users' search and selection items in an information overloaded environment is widespread. Usually, precision recall are utilized for evaluating a system. However, alternative measure should be considered, because user's satisfaction the most important factor considered when constructing Briefly, there exists novel technique, serendipitous recommendation that considers this factor. In paper, we propose new method technique. our method, utilize map...
Database systems have recently utilized both row-and a column-oriented storage systems, also termed as hybrid storage, their devices for large scale data management. The based database should ideally be operated on distributed computing environments query optimization, however, studies optimization of such are not available in literature. Therefore, the selection type and accurate estimation workload critical factors efficient processing environments. In this paper, we describe novel method...