| 1. |
SMArch : Software Architecture for Smart Mobility (ITS). |
| 2. |
Unusual driving spots discovery from driving probe data using machine learning
Driving probe data are captured to analyze driving behavior of a car and the road conditions. Probe devices are realized by smartphone or AVL (Automatic Vehicle Location) with GPS (Global positioning system). Some researchers apply machine learning methods for analysis of driving probe data. In this research, we propose a method to distinguish unusual driving behavior (sudden braking) using machine learning method. We apply SVM (support vector machine) and feature selection method to the driving probe data provided by Aizuw-Wakamatsu City Open Data Utilization Verification Project. We report the effect of feature selection for unusual driving behavior detection, using F-measure and accuracy. We also show that selected features identify unusual driving condition (place and date).. |
| 3. |
Comment Mining to Estimate Junior High-school Student Performance toward Improvement of Student Learning. |
| 4. |
Estimation of Waiting Time of Elevators toward Ito-Campus Guide System. |
| 5. |
A Proposal of Information Sharing System for Improving Community Life. |
| 6. |
A Study on Learning Process in Embedded System Development Education aiming at Quad Pro. |
| 7. |
Case study of Project Based Learning on embedded system development for undergraduate students and analyze of learning effect. |
| 8. |
Communication behavior analysis using BLE and smartphone sensor information. |
| 9. |
A platform of sharing and recommending information at a local community area. |
| 10. |
Extraction of Current Actual Status and Demand Expressions in Community Complaint Reports. |
| 11. |
Booth and visitors feature analysis of considering the residence time based on Wi-Fi log data. |
| 12. |
Development of Adaptive User Interface with transportation information system(Patrash). |
| 13. |
Complaint Data Analysis to Estimate Issue Resolution Date. |
| 14. |
ShaymaaE.Sorour, Tsunenori Mine, Kazumasa Goda, Sachio Hirokawa, A Predictive Model to Evaluate Student Performance (Preprint), 情報処理学会論文誌, Vol.56, No.3, 2015.03, In this paper we propose a new approach based on text mining techniques for predicting student performance using LSA (latent semantic analysis) and K-means clustering methods. The present study uses free-style comments written by students after each lesson. Since the potentials of these comments can reflect student learning attitudes, understanding of subjects and difficulties of the lessons, they enable teachers to grasp the tendencies of student learning activities. To improve our basic approach using LSA and k-means, overlap and similarity measuring methods are proposed. We conducted experiments to validate our proposed methods. The experimental results reported a model of student academic performance predictors by analyzing their comments data as variables of predictors. Our proposed methods achieved an average 66.4% prediction accuracy after applying the k-means clustering method and those were 73.6% and 78.5% by adding the overlap method and the similarity measuring method, respectively.------------------------------This is a preprint of an article intended for publication Journal ofInformation Processing(JIP). This preprint should not be cited. Thisarticle should be cited as: Journal of Information Processing Vol.23(2015) No.2 (online)DOI http://dx.doi.org/10.2197/ipsjjip.23.192------------------------------In this paper we propose a new approach based on text mining techniques for predicting student performance using LSA (latent semantic analysis) and K-means clustering methods. The present study uses free-style comments written by students after each lesson. Since the potentials of these comments can reflect student learning attitudes, understanding of subjects and difficulties of the lessons, they enable teachers to grasp the tendencies of student learning activities. To improve our basic approach using LSA and k-means, overlap and similarity measuring methods are proposed. We conducted experiments to validate our proposed methods. The experimental results reported a model of student academic performance predictors by analyzing their comments data as variables of predictors. Our proposed methods achieved an average 66.4% prediction accuracy after applying the k-means clustering method and those were 73.6% and 78.5% by adding the overlap method and the similarity measuring method, respectively.------------------------------This is a preprint of an article intended for publication Journal ofInformation Processing(JIP). This preprint should not be cited. Thisarticle should be cited as: Journal of Information Processing Vol.23(2015) No.2 (online)DOI http://dx.doi.org/10.2197/ipsjjip.23.192------------------------------. |
| 15. |
Analysis of Purchase Behavior based on IC card usage record before and after Transportation
Capturing people's behavior patterns, such as purchase patterns or shop-around-behavior patterns, are big marketing concern since such patterns are good resources to invent new marketing services. To this end, a lot of studies using POS (Points Of Sales) data have been conducted to extract purchase behaviors for CRM (Customer Relation Management). In addition, since a lot of widespread sensor devices, smart cards or mobile phones have become available recently, anyone can try to analyze pedestrians' trajectories obtained as GPS data and extract people's behavior patterns. However, only few research has been conducted to analyze relationships between boarding and purchase behaviors based on usage histories of smart IC cards. This paper presents case studies of purchase behavior analysis based on usage histories of smart IC cards, especially focusing on purchase behaviors of passengers who just get on and off trains.. |
| 16. |
Noh Guide : Proposal Of Stage Art Noh-Guide with Personal Recommendation
We show the Noh Guide to understand the real stage performance, and it's system frameworks. We consider the audience who look the stage at the Noh theater with their mobile device, and use their device to get the information of performance more and more easily with personalized recommendation. At the same time, the system make the difficulty of Noh to decrease with the audience and actor. In this paper, we introduce outview of our prototype system and result of the experiment with 15 number of tester. Now, Only National Noh Theatre have such display system which are fitted to the chair before, but We will be able to use the same service with our Noh Guide System.. |
| 17. |
Reciprocal Recommendation for Job Matching : an Evaluation
This paper discusses reciprocal recommendation methods for job matching. We investigate how job matching factors are influenced by changing the importance rate between user condition and profile. In this paper, we report experimental results with real job-opening information and discuss how job matching factors such as success rate of matching, the number of matched pairs, satisfaction degrees of a matched pair are influenced when changing the importance rate between user condition and profile and the number of job seekers.. |
| 18. |
PATRASH : Personalized Autonomous TRAnsportation recommendation System considering user context and social History
Now, we are working on developing a transfer system which is based on information of user's position and so on to find out the best route so that users can get transfer information easily. The information include now time,user's position and bus stop/station position.Further, this system also provide information of delay,the bus is full or empty.The situation is calculate by the position of the bus and the time table.The system will give the transfer information with result of the calculation and give the best choice when users open up the App.And more,the system considered the user's using history,so the user can get information with the least operation. For this purpose, the system will determine if the user is using the routine route or not.This paper describes the verification of accuracy and the algorithm of this transfer system which considering the delay information.This paper also talks about by accumulating data of user's history, how did we determine if the user is using the routine route or not and the verification of accuracy is worth to looking forward.. |
| 19. |
Experimental Evaluation of Eleven Fault-Density Models
ソフトウェアテスト・保守において，限られたリソースで信頼性を確保することを目的として，モジュールに含まれるfaultの有無や数を予測するモデルが数多く提案されている．しかし，従来研究では，評価用のプロジェクト（データセット）の数が少ない，比較に用いたモデル数が少ない，モジュールの規模を考慮した評価を行っていない（fault密度を予測していない），モデル構築用と評価用のデータセットを同じサンプルから抽出している（バージョンをまたがる評価を行っていない），といった問題があり，いずれのfault予測モデルが優れているのか明らかでなかった．そこで本論文では，25のオープンソースプロジェクトからそれぞれ2バージョンずつ収集したデータセットを用いて，11種類のfault密度予測モデルの性能を実験的に評価した．実験の結果，樹木モデル（M5）が最も高い性能を有することが分かった．. |
| 20. |
Experimental Evaluation of Eleven Fault-Density Models
ソフトウェアテスト・保守において,限られたリソースで信頼性を確保することを目的として,モジュールに含まれるfaultの有無や数を予測するモデルが数多く提案されている.しかし,従来研究では,評価用のプロジェクト(データセット)の数が少ない,比較に用いたモデル数が少ない,モジュールの規模を考慮した評価を行っていない(fault密度を予測していない),モデル構築用と評価用のデータセットを同じサンプルから抽出している(バージョンをまたがる評価を行っていない),といった問題があり,いずれのfault予測モデルが優れているのか明らかでなかった.そこで本論文では,25のオープンソースプロジェクトからそれぞれ2バージョンずつ収集したデータセットを用いて,11種類のfault密度予測モデルの性能を実験的に評価した.実験の結果,樹木モデル(M5)が最も高い性能を有することが分かった.. |
| 21. |
Patrash : Personalized Autonomous Transit Recommender ad Assistant System with input Helper
This paper describes an ongoing project work and a work-in-progress system, which is a personalized transit recommendation system called Patrash: Personalized Autonomous Transit Recommender and Assistant System with input Helper. The Patrash system deals with time tableinformation of various public transportation companies and recommends transit information to users considering their context information and usage histories. To this end, the Patrash system tries to detect user's situation if the users want to transfer lines on a routine basis or not. In this paper, we introduce overviews of this project and the Patrash system, and show some of preliminary experimental results to illustrate effects of the Patrash system.. |
| 22. |
Possible Collaborators Search for Efficient Demand Respons
Due to the devastating earthquake in East Japan, saving electric power is urgently needed for economic development. Therefore, depending on supply and demand, one demand response which is electric power network for changing the quantity of electricity is being studied. In demand response model which consumers will be requested to reduce electric power. With the number of consumers requested growing, much more computations and request information traffic will be needed for searching the collaborator. Therefore, this is a research for realizing demand response efficiently on smart grids to search collaborators. Not only perform peak power cut but also decrease request information traffic effectively.. |
| 23. |
Maching between People and Job with Bidirectional Feedback
This paper proposes a novel job matching method with bi-directional feedback. The proposed method uses, as mutual feedback, bilateral messages between job seekers and recruiters, such as apply to a job, scout of a seeker, and reply to the offer, on the seekers-recruiters network. From the feedback messages, it computes the popularity value of each seeker and recruiter, and evaluation value of each other; a ranking list for recommendation of each seeker and recruiter is created based on the popularity and evalueation values. Preliminary experiments illustrate the validity of the proposed method.. |
| 24. |
Finding Improvable Students from Self-Evaluated Comments via PCN Method. |
| 25. |
Possible Collaborators Search for Efficient Demand Respons. |
| 26. |
Estimating Brownout effects for Demand Response considering brownout incentive caused by the Great East Japan Earthquake. |
| 27. |
A Model-Driven Development Case of Airship Auto Navigation System Using SysML. |
| 28. |
Improvement Case of Development Process with SysML in Airship Auto Navigation System
SysML can describe the hardware and the software of the system. SysML is useful in a development of embedded systems. However development process using SysML is not defined. In this study, we develop an airship auto navigation system using SysML and improve the process by agile. We analyze our review and the number of bugs, and examine the effect of the process improvement.. |
| 29. |
User Customized Data Protection using Policy Rule Repository. |
| 30. |
Towards Clustering Learners by Analyzing Sentences from Time-Series Comments of Classroom : Proposal for a PCN Method
In this paper we analyze writings in the classroom such as summaries, reviews, or comments on the class, and propose a method for guessing the learning status of learners (named PCN Method). After applying PCN Method to analysis of classes on introductory education for information processing at Kyushu University, we are sure that PCN Method is effective for grasping the learning status of learners.. |
| 31. |
User Feedback Utilization and Query Destination Learning Methods for Peer-to-Peer Information Retrieval - an Evaluation
This paper proposes a novel Peer-to-Peer Information Retrieval method using user feedback and query-destination-learning. The method utilize negative feedback information actively so that other agents can filter out irrelevant documents with it. Using query-destination-learning, the method can accumulate relevant information from all the member agents in a community. It can also reduce communication loads by caching queries and their sender-responder agent addresses in the community. The experimental results illustrated the validity of the proposed methods.. |
| 32. |
A Method for Relevance Calculation between Browsed Web Pages toward Personalized Recommendation System
This thesis proposes "Relative-Webpage-Recommendation" as a support of browsing. Using this method, user can get webpages that has relevance to browsing one. And we think that this method provides more personalized web browsing. Relevance between webpages is calculated from user's browsing history. In this thesis, we propose the method for calculating relevance from two elements. One is that user's browsing history is mappable to forest structure. Another is a hypothesis that there is some co-occurrence between pages which browsed at the same time by tab browser. Additionally, we consider experimental result, and show issue.. |
| 33. |
The ACP2P Method using User Feedbacks and Query Learning in Multiple Communities - An Evaluation
This paper proposes refinement methods of the ACP2P using user feedbacks and query-learning methods. The proposed methods are effective in increasing retrieval accuracy in communities and for decreasing communication loads required for documents retrieval. The paper especially focuses on user feedbacks and query learning in multiple communities. The experimental results illustrated the validity of our proposed methods.. |
| 34. |
A practical approach to foster self-directed ICT human resources through intimate relationship between academia and business industries
The progress of Information Technology, which is the infrastructure of an advanced information society, is remarkable and has the enormous impact on our daily life. On the other hand, it has been pointed out by the industry that there is a lack of highly skilled ICT personnel who can lead the next generation. In order to address this issue, the Graduate School of Information Science and Electrical Engineering in Kyushu University has established Social Information System Engineering Course. In this education course, we have been running the practical education program with an objective to foster world class leader who has extraordinary technical skill and sense of ethics. This program is supported by various companies through Nippon Keidanren and by Ministry of Education, Culture, Sports, Science and Technology. This paper describes the content, method, result and evaluation for our two years experience. We also discuss the issues and concerns that need to be resolved.. |
| 35. |
A Method for Query Generation Using the Related Word Extraction Algorithm
Web検索では,検索エンジンによって得られた結果がユーザの必要としている情報ではないことがしばしばある.この問題を解決する一つの方法として検索エンジンに与えるクエリを改善するクエリ拡張がある.クエリに含まれる語としてユーザの意図する適切な語を生成し追加することでWeb検索結果の改善に大きな効果がある.本論文では,ユーザの意図する語を生成するための方法として,センテンス間の距離に注目した関連単語抽出アルゴリズムを提案する.このアルゴリズムは重要な語の近くに出現する単語は重要であるという考えに基づいている.更に,このアルゴリズムを適合性フィードバックの一手法であるRSVと組み合わせることで,検索エンジンを利用した初期検索結果の検索精度の改善に役立つこと,その結果はRSV単独で行った場合よりも検索精度が向上することを実験により示す.. |
| 36. |
Advantageous Effects of Internship Program for Fostering Highly Skilled ICT Personnel through Large Scale Business-academia Collaboration
高度情報化社会の基礎である情報技術の発達は我々の生活を一変させるほどのインパクトを持っており，その進歩の速度は目覚ましいものがある．一方で産業界からは次世代を担う能力を備えた高度ICT（Information and Communication Technology）人材の不足が指摘されている．このような技術の進歩と社会的要請に応えるためには，高度な技術力を持ち，基礎知識と社会的倫理観を兼ね備えた世界に通用するリーダの育成が不可欠であり，そのための実践的教育プログラムが求められる．特に，インターンシップは，学生にICT産業の現実を体感させ，さらには，学生自身の実力と期待されている能力との差を実感させることができるとともに，教育カリキュラム自体にフィードバックさせるため，実践的教育プログラムの中でも重要な位置を占める．ところが，これまでのわが国のインターンシップはその約6割が2週間未満という短期間で実施されていたため，企業紹介の一環あるいは早期採用活動の一部という色合いが強く，実践的教育という目的からは不十分であった．そこで，九州大学大学院システム情報科学府（以下，本学府）では，学生教育のために産学が連携し，実践的な長期インターンシップの実施を開始することとなった．本論文では，本学府が行っている産学が連携したインターンシップの取り組みについて，その具体的な内容を学生・教員・企業という3者の立場からの考察を交えながら論じるとともに，今後の課題について述べる．The progress of Information Technology, which is the infrastructure of an advanced information society, is remarkable and has the enormous impact on our daily life. On the other hand, it has been pointed out by the industry that there is a lack of highly skilled ICT (Information and Communication Technology) personnel who can lead the next generation. In order to address this issue, practical education program is required to foster world class leader who has extraordinary technical skill, basic knowledge and sense of ethics. Especially, Internship program is very important to have students experience the real environment in ICT industry, recognize the gap between the required skills and their current skills, and feedback to education program itself. However, traditional Internship is not enough to realize these objectives because the duration has been less than 2 weeks for 60% of traditional Internship, where the main purpose is the part of either promotion of the company's image or early recruitment activity. So, the Graduate School of Information Science and Electrical Engineering in Kyushu University has started the practical and long term Internship program through large scale business-academia collaboration. This paper describes the content and evaluation of this collaborative Internship program that are viewed from students, university and corporate staffs. We also discuss the issues and concerns that need to be resolved.. |
| 37. |
Introduction to the Special Issue on Agent. |
| 38. |
An Approach to Foster Highly Skilled ICT Personnel through Large Scale Business-academia Collaboration
高度情報化社会の基礎である情報技術の発達は我々の生活を一変させるほどのインパクトを持っており，その進歩の速度は目覚ましいものがある．一方で産業界からは次世代を担う能力を備えた高度ICT（Information and Communication Technology）人材の不足が指摘されている．このような技術の進歩と社会的要請に応えるため，九州大学大学院システム情報科学府では平成19年4月から新しい修士課程教育コースである社会情報システム工学コースを設置した．本コースでは，文部科学省の支援を受け，日本経団連傘下企業との大規模な連携体制の下で，高度な技術力を持ち，基礎知識と社会的倫理観を兼ね備えた世界に通用するリーダの育成を目的とした実践的教育を開始した．この教育プログラムの計画実施過程の中で，1) PBL（Project Based Learning），2) オムニバス形式講義，3) 長期インターンシップ，4) カリキュラム内容検討，のあり方や実施方法に関していくつかの知見が得られた．本論文では，当コースの教育内容と方法，修士課程1年前期終了時点における実績と評価，および今後の課題について述べる．The progress of Information Technology, which is the infrastructure of an advanced information society, is remarkable and has the enormous impact on our daily life. On the other hand, it has been pointed out by the industry that there is a lack of highly skilled ICT (Information and Communication Technology) personnel who can lead the next generation. In order to address this issue, the Graduate School of Information Science and Electrical Engineering in Kyushu University has established Social Information System Engineering Course. Since April 2007, we have started practical education program with an objective to foster world class leader who has extraordinary technical skill, basic knowledge and sense of ethics. This effort is gradually progressing by the collaboration with various companies through Nippon Keidanren with support of Ministry of Education, Culture, Sports, Science and Technology. There have been several findings in 1) PBL (Project Based Learning), 2) omnibus classes, 3) long term internship, and 4) curriculum improvements during the planning and execution of this program. This paper describes the content, method, result and evaluation at the end of the first semester of this course. We also discuss the issues and concerns that need to be resolved.. |
| 39. |
A Method to Make User Profiles using ODP for Personalized Search
The user often uses search engines to search his intended pages in the Internet. However, the intended pages do not always exist on the higher rank of the results. The Personalized Search is promising approach to solve this problem. In the Personalized Search, User Profiles, which are often created from the documents of the user to present his/her interests, are well used. This paper proposes (1) a method for generating User Profiles based on Open Directory Project (ODP) and shows (2) a Personalized Search system using Book Marks as the documents for profiling. Some of experimental results show the validity of the method for creating User Profiles and the precision enhancement of this system.. |
| 40. |
A Method to Make User Profiles using ODP for Personalized Search
The user often uses search engines to search his intended pages in the Internet. However, the intended pages do not always exist on the higher rank of the results. The Personalized Search is promising approach to solve this problem. In the Personalized Search, User Profiles, which are often created from the documents of the user to present his/her interests, are well used. This paper proposes (1) a method for generating User Profiles based on Open Directory Project (ODP) and shows (2) a Personalized Search system using Book Marks as the documents for profiling. Some of experimental results show the validity of the method for creating User Profiles and the precision enhancement of this system.. |
| 41. |
Applying a related word extraction algorithm to Wikipedia for query extention
In this paper, we propose a method to extend the original query by applying a related word extraction algorithm to Wikipedia that is an encyclopedia. The algorithm extracts words related to given key words from any text and calculates a score for each extracted word according to distance between words in the text. The method helps users to obtain their desired documents in web search by extending a query with highly scored words.. |
| 42. |
Applying a related word extraction algorithm to Wikipedia for query extention
In this paper, we propose a method to extend the original query by applying a related word extraction algorithm to Wikipedia that is an encyclopedia. The algorithm extracts words related to given key words from any text and calculates a score for each extracted word according to distance between words in the text. The method helps users to obtain their desired documents in web search by extending a query with highly scored words.. |
| 43. |
Agent-community-based peer-to-peer information retrieval improved with user feedback: an evaluation
In this paper, we discuss a method for improving the Agent-Community-based Peer-to-Peer (ACP2P) Information Retrieval based on the user feedback. The ACP2P assumes an agent works as a delegate of its user and search for information that the user wants by communicating with other agents. In the ACP2P, using user feedback is effective to search for information highly-acclaimed by users in a community as the information relevant to a user query, and thus sharing the feedback among the community helps to make retrieval accuracy increase in all the community. We also present how to use relevance judgements provided by the user when calculating the similarity between a query and a document, and show how to reduce communication loads in searching for information relevant to a query. Experimental results illustrate the validity of our approach.. |
| 44. |
Agent-community-based peer-to-peer information retrieval improved with user feedback: an evaluation
In this paper, we discuss a method for improving the Agent-Community-based Peer-to-Peer (ACP2P) Information Retrieval based on the user feedback. The ACP2P assumes an agent works as a delegate of its user and search for information that the user wants by communicating with other agents. In the ACP2P, using user feedback is effective to search for information highly-acclaimed by users in a community as the information relevant to a user query, and thus sharing the feedback among the community helps to make retrieval accuracy increase in all the community. We also present how to use relevance judgements provided by the user when calculating the similarity between a query and a document, and show how to reduce communication loads in searching for information relevant to a query. Experimental results illustrate the validity of our approach.. |
| 45. |
Agent-Community-based Peer-to-Peer Information Retrieval Improved with User Feedback-an Evaluation. |
| 46. |
Agent-Community-based Peer-to-Peer Information Retrieval Improved with User Feedback-an Evaluation. |
| 47. |
A Method of Query Generation Using the Related Word Extraction Algorithm
When searching for information a user wants, search engines often return lots of results unintended by the user. Query expansion is a promising approach to solve this problem. In the query expansion research, one of big issues is to generate appropriate keywords representing the user's intention. This paper proposes the related-word-extraction-algorithm (RWEA) which pays attention to the distance between sentences where keywords in the original query appear. The RWEA is based on the idea that a word nearby important words is also important. We conducted several experiments to illustrate the validity of the RWEA. The results promise the effectiveness of the RWEA for improving search results.. |
| 48. |
A Method of Query Generation Using the Related Word Extraction Algorithm
When searching for information a user wants, search engines often return lots of results unintended by the user. Query expansion is a promising approach to solve this problem. In the query expansion research, one of big issues is to generate appropriate keywords representing the user's intention. This paper proposes the related-word-extraction-algorithm (RWEA) which pays attention to the distance between sentences where keywords in the original query appear. The RWEA is based on the idea that a word nearby important words is also important. We conducted several experiments to illustrate the validity of the RWEA. The results promise the effectiveness of the RWEA for improving search results.. |
| 49. |
Web page recommendation by using the user's schedules. |
| 50. |
The use of schedule information in the web retrieval. |
| 51. |
A Network Infrastructure Construction Method for Building Multi-Agent-Based System
各ユーザに割り当てられたエージェントによって構成されるマルチエージェントシステムを基盤としたアプリケーションでは,エージェントの起動と停止が頻繁に生じることが想定される.そのため,エージェント間通信の頑健性と,エージェント間を接続するネットワークの管理コストを軽減することが強く求められる.一方,通信相手であるエージェントのネットワーク上の位置と,そのエージェントまでの経路探索も効率良く行う必要がある.本論文では,これらの目的を満たすためのネットワークトポロジー,房付順序木(Ordered Tree with Tuft:OTT)を提案するとともに,このOTTを,マルチエージェントシステムKODAMAの利用するミドルウェアAgent Communication Zone (ACZ)のエージェント間経路探索.メッセージ配送制御のために実装し,シミュレーション実験を行った結果について議論する.. |
| 52. |
Query Term Selection for User-Schedule-based Web Page Recommendation
Personalized services on Web information retrieval and recommendation have become more and more attractive. Realizing such services is a crucial issue for ubiquitous computing environment where users can easily access what they want at any time. An important clue for the services is to consider user's view points and situations. This paper presents a personalized information recommendation system that makes use of a user schedule. The schedule is an important resource for grasping his/her view points and situations along the time sequence of his/her activities. We show some of the preliminary experimental results to illustrate the validity of our approach.. |
| 53. |
Query Term Selection for User-Schedule-based Web Page Recommendation. |
| 54. |
Active Browsing : A Novel Personalized Web Browsing Support System
This paper presents the Active Browsing system which supports user's Web browsing activity. The system observes user's Web browsing behaviors, learns his/her ways or preferences and tries to save his/her labor for reaching the pages he/she wants by recommending them. The system recommends not only the pages the user browsed in the past, but also the pages which are new and interesting to him/her by collecting from other users with a collaborative-filtering technique. The Active Browsing system is implemented using Multi-Agent KODAMA.. |
| 55. |
Tsunenori Mine, Akihiro Kogo, Makoto Amamiya, Agent-community-based peer-to-peer information retrieval and its evaluation, Systems and Computers in Japan, 10.1002/scj.20625, Vol.37, No.13, pp.1-10, 2006.11, The Agent-Community-based Peer-to-Peer (ACP2P) information retrieval method uses agent communities to manage and look up information of interest to users. An agent works as a delegate of its user and searches for information that the user wants by communicating with other agents. The communication between agents is carried out in a peer-to-peer computing architecture. Retrieving information relevant to a user query is performed with content files which consist of original and retrieved documents, and two histories: a query/retrieved document history and a query/sender agent history. Making use of the histories has a collaborative filtering effect, which gradually creates virtual agent communities, where agents with the same interests stay together. Our hypothesis is that a virtual agent community reduces communication loads necessary to perform a search. As an agent receives more queries, then more links to new knowledge are acquired. From this behavior, a "give-and-take" (or positive feedback) effect for agents seems to emerge. However, we have only shown parts of the effects of reducing communication loads and making "give-and-take" through preliminary experimental results. This paper discusses more detail of experimental results and shows that the two histories help in reducing communication loads among agents, facilitating bidirectional communications between them and thus creating virtual agent communities, where agents with the same interests stay together. © 2006 Wiley Periodicals, Inc.. |
| 56. |
D_024 Efficient Web Retrieval Based on User's Intention. |
| 57. |
秋恩炯, 峯 恒憲, 雨宮 真人, クラスタリングと内部ネットワーク処理を利用した経路制御アルゴリズムに基づくデータ収集の仕組み, 情報処理学会研究報告モバイルコンピューティングとユビキタス通信（MBL）, Vol.2006, No.50, pp.45-50, 2006.05, ネットワークの生存期間を延ばすため，センサーネットワークでは，エネルギーに敏感で，効率的なデータ収集手法が要求される．この論文では エネルギーを効率的に扱うデータ収集手法を提案する．この手法では，まずセンサーノードをクラスタリングし，作成された各クラスタ毎に，分散されたデータ経路木を，ネットワーク内部でのデータ統合手法に基づいて，形成する．この手法では，センサーノードのＩＤを利用して，順番に，１つのクラスタヘッドを選択する．更に，クラスタの形成とデータ経路木の構成を同時に実行する．これにより，観測されたデータの経路木を構成するのに従来必要であったエネルギーを削減できる。また、各ノードでデータの集約を行うことで、転送するデータの総量も削減する。実験結果により、提案手法が、ダイレクトスキームやLEACHと比べて、ネットワークの生存期間の観点で性能を上回ることがわかった。Sensor networks require energy-aware,efficient data-collecting methods to extend network lifetime. In this paper，we propose an enelgy-efficient data gathering mechanism which clusters sensor nodes and forms a distributed data-routing tree based on in-network data fusion within each cluster. In our mechanism，the cluster formation and the data-routing tree construction are simultaneously carried out so that they reduce their energy required to organize a multi-hop routing tree of sensed data. The mechanism also performs data aggregation at each member node to reduce the amount of transmission data. Moreover our work distributes energy load to each node to avoid the intensive energy consumption of a cluster-head. Experimental results show that our data gathering mechanism outperforms the direct scheme protocol and the LEACH protocol on the point of view of the network lifetime.. |
| 58. |
秋 恩炯, 峯 恒憲, 雨宮 真人, クラスタリングと内部ネットワーク処理を利用した経路制御アルゴリズムに基づくデータ収集の仕組み, 電子情報通信学会技術研究報告. MoMuC, モバイルマルチメディア通信, Vol.106, No.44, pp.45-50, 2006.05, ネットワークの生存期間を延ばすため,センサーネットワークでは,エネルギーに敏感で,効率的なデータ収集手法が要求される.この論文では,エネルギーを効率的に扱うデータ収集手法を提案する.この手法では,まずセンサーノードをクラスタリングし,作成された各クラスタ毎に,分散されたデータ経路木を,ネットワーク内部でのデータ統合手法に基づいて,形成する.この手法では,センサーノードのIDを利用して,順番に,1つのクラスタへツドを選択する.更に,クラスタの形成とデータ経路木の構成を同時に実行する.これにより,観測されたデータの経路木を構成するのに従来必要であったエネルギーを削減できる。また、各ノードでデータの集約を行うことで、転送するデータの総量も削減する。実験結果により、提案手法が、ダイレクトスキームやLEACHと比べて、ネットワークの生存期間の観点で性能を上回ることがわかった。. |
| 59. |
Agent-Community-Based Peer-to-Peer Information Retrieval and Its Evaluation
階層化コミュニティ構造をもつマルチエージェントシステムを利用する情報検索手法ACP2P法は, コミュニティ内の各エージェントがもつデータの内容に対して検索を行うほか, 他のエージェントから受けた検索履歴をもとに情報のありかの特定や同じトピックに関心をもつエージェント同士のグルーピングを実現する. これにより, 必要な検索結果を得るために行う通信の量を徐々に削減していくことができる. 更に検索要求に多くこたえられるエージェントほど自分の求める情報源へのパスを増やすことができ, その結果, 検索効率が向上するという「give and take」効果の実現が期待できる. しかしこれまでは簡単なシミュレーション実験によりACP2P法が仮定していた効果の一部を確認したにすぎず, 十分にその効果を示したとはいえなかった. 本論文では, 特に, エージェント間通信量の削減効果と同じトピックに関心をもつエージェントから構成される仮想エージェントコミュニティの生成効果の点について行った評価実験の結果について議論し, これらの効果があることを示す.. |
| 60. |
B-017 An Approach towards Autonomous Agent-Community Generation. |
| 61. |
D-041 Agent-Community-based Peer-to-Peer Information Retrieval and Its Retrieval Accuracy Evaluation. |
| 62. |
Automatic Exercise Generating System That Dynamically Evaluates both Students' and Questions' Levels
Popularization of computers and the Internet enable us to hold lectures using Web contents as a teaching material and even develop new lecture methods using the technologies. Although teachers have prepared a lot of Web contents, most of them are utilized so as only to be browsed by students. If we arrange some exercises according to lecture notes and prepare an answering mechanism for the exercises via the Internet, every student can attempt the exercises any time. This paper proposes a Web-based self teaching system: AEGIS (Automatic Exercise Generator based on the Intelligence of Student) that generates exercises of various difficulty levels according to each student's achievement level, marks his/her answers and returns them to him/her. Once the documents are marked up with tags by teachers to specify the positions of questions, AEGIS can generate three question types: multiple-choice question, fill-the-gap question, and error-correcting question from the same tagged document. It is very important for AEGIS to estimate both an achievement level of each student and a difficulty level of each question. AEGIS dynamically evaluates these levels. In this paper, we discuss the design of the tags for AEGIS, the re-estimation of the levels and generating method of AEGIS.. |
| 63. |
AEGIS : an Automatic Exercise Generating System That Dynamically Evaluates both Students' and Questions' Levels
Popularization of computers and the Internet enable us to hold lectures using Web contents as a teaching material and even develop new lecture methods using the technologies. Although teachers have prepared a lot of Web contents, most of them are utilized so as only to be browsed by students. If we arrange some exercises according to lecture notes and prepare an answering mechanism for the exercises via the Internet, every student can attempt the exercises any time. This paper proposes an Web-based self teaching system : AEGIS (Automatic Exercise Generator based on the Intelligence of Student) that generates exercises of various difficulty levels according to each student's achievement level, marks his/her answers and returns them to him/her. Once the documents are marked up with tags by teachers to specify the positions of questions, AEGIS can generate three question types : multiple-choice question, fill-the-gap question, and error-correcting question from the same tagged document. It is very important for AEGIS to estimate both an achievement level of each student and an difficulty level of each question. AEGIS dynamically evaluates these levels. In this paper, we discuss the design of the tags for AEGIS, the re-estimation of the levels and generating method of AEGIS.. |
| 64. |
Topics Extraction and Clustering from Call for Papers Files about International Conference
We are aiming to develop a Web mining system that gathers Web pages, filters them according to user's interests, extracts some entities specified by the users and clusters the entities extracted to discover some relationships between them. In this paper, we propose a Web mining method and discuss preliminary experimental results. The system at first filters the Web pages related to call for papers with SVM(Support Vector Machine) from Web pages gathered through a search engine. Next, it extracts conferences' topics and clusters them. The results clustered are shown through an Explorer like graphical user interface.. |
| 65. |
Development and Operation of a Document Database for University Research and Education Activities
九州大学教官の研究教育活動等報告書データベースは, 各教官の研究教育活動の報告書に関して, インターネットを通じた各教官による報告書の入力や更新, 管理者による報告書の検査等の管理作業, 閲覧者による報告書の検索や閲覧を行うためのシステムである.教官が入力を行うための方法は, 複数の方法からの選択が可能である.また, 報告書は各教官ごとに異なったフォーマットを選択することが可能で, データベース構造にリレーショナルデータベースとSGMLを混在させることでその機能を実現している.システム設計や運用体制についての評価をシステムのログや開発者に送られた電子メールに基づいて行った結果, システム運用時における管理者を複数設けて分担する分散管理による負担軽減が機能していることや, システム更新時の修正箇所が複数のモジュールにわたることが少ないことが確認できた.. |
| 66. |
Automatic Exercise Generator with Tagged Documents
Popularization of computers and the Internet enable us to hold lectures using Web contents as a teaching material. We should, however, make better use of the new computer and the Internet technologies, and have to develop a new lecture method using them. If we arrange some exercises which follow a lecture note and prepare an answering mechanism for the exercises through the Internet, every student can try the exercises any time. This system aims to help the students understand the lecture. In this paper, we propose AEGIS(Automatic Exercise Generator based on the Intelligence of Student)system which generates some kinds of exercises automatically based on the tagged documents with considering both the ability of a student and the abstrusity to solve the exercises. It realizes a feedback mechanism by restricting the exercise generation to the well-known three types.. |
| 67. |
AgenTV : A Natural Language interface using AAOSA. |
| 68. |
Dynamic information retrieval system on Web using AAOSA
As information on Web is getting increase more and more, it is more important to extract information which one needs. But traditional information retrieval systems can not keep up with increasing and changing of information. To solve this problem, we propose a new style of retrieval system, which consists of distributed and cooperative agents on Web. In our system, each agent is initialized by content of a Web page, gets keywords, and processes an user's query by communicating each other. We use agents proposed on Adaptive Agent Oriented Software Architecture (AAOSA). Now we are developing the system on a local server.. |
| 69. |
Japanese Information Retrieval System using Syntactic Dependency Relations on the Web
This paper proposes a Japanese Information Retrieval(IR) system which uses syntactic dependency relations between words. We believe this method makes good use of the grammatical conditions that have been ignored. Experimental results say that our IR system obtains 11% higher precision rate than a convertional one adapting tf・idf method, and also, the precision rate reached about 94% in the case that candidate pairs in a query were included more than 70% in a sentence of a retrieved document.. |
| 70. |
Dynamic Visualization of Pascal Programs on WEB
我々は、毎年2,300人の学生を対象とする一般情報処理教育をホームページを用いて行っている. しかし、プログラムの動作は文章や図だけでは初心者には分かりにくい。そこで、パスカルプログラムをWWWブラウザで見せるためのシステムを開発している. これは、パスカルプログラムをJavaアプレットヘ変換する。本稿ではシステム開発の背景と実現方式の概略を述べる。. |
| 71. |
Case Frame Acquisition of SAHEN Verbs by Using Precreated Case Frames of Japanese Basic Verbs
A large amount of sample sentences are required in order for case frame acquisition. However, collecting sentences that include Japanese SAHEN verb is hard work because the number of SAHEN verbs are huge and lots of coined words are created as such SAHEN verbs. Most SAHEN verbs consist of several elements such as a noun word, an adjective word, and a Japanese basic verb called a WAGO verb. Taking account of this feature, we assume that a case frame of a SAHEN verb is similar to one of a WAGO verb that is a component of the SAHEN verb. This paper proposes a method to acquire case frames of SAHEN verbs using case frames of WAGO verbs that have already obtained before. Experimental results show the effectiveness of our method.. |
| 72. |
Case Frame Acquisition of SAHEN Verbs by Using Precreated Case Frames of Japanese Basic Verbs
A large amount of sample sentences are required in order for case frame acquisition. However, collecting sentences that include Japanese SAHEN verb is hard work because the number of SAHEN verbs are huge and lots of coined words are created as such SAHEN verbs. Most SAHEN verbs consist of several elements such as a noun word, an adjective word, and a Japanese basic verb called a WAGO verb. Taking account of this feature, we assume that a case frame of a SAHEN verb is similar to one of a WAGO verb that is a component of the SAHEN verb. This paper proposes a method to acquire case frames of SAHEN verbs using case frames of WAGO verbs that have already obtained before. Experimental results show the effectiveness of our method.. |
| 73. |
On-Web-Teaching 2300 Students Information Processing and Computer Literacy at Kyushu University
We are developing a Web system for the introductory course of programming and computer literacy. 2300 students take the course with this Web system at Kyushu University. Our main object of this system is to reappear our lecture in front of students in their private studies. Teaching how programs run is a hard problem to deal with in information processing educations. In order to deal with such a difficult subject easily, we propose a program animation compiler for assisting students to understand a motion of programs. This compiler systematically translates tiny Pascal programs into Java Applets. In this paper, we state our idea and future plans of the system.. |
| 74. |
Sense Discrimination of Verbs Using EDR Dictionary
Case frames are useful to discriminate verb senses. However, enough number of them required for analyzing general texts are not available. Therefore, it is important to develop a case frames acquiring method in the NLP field. Some case frames whose semantic features consist of a set of word concepts can be constructed from classifying lots of sentences by a verb sense. A key of classifying those sentences is a semantic distance (similarities) between the sentences. In this paper, we propose a method for constructing case frame that is based on similarities between words or sentences. The similarity is calculated using the EDR conceptual dictionary. We also reports. about experimental results for evaluating our method and features of EDR dictionaries.. |
| 75. |
Sense Discrimination of Verbs Using EDR dictionaries and EDR Corpora
CO-locational data like a case frame is useful for dissolving syntactic ambiguity. In order to acquire those data,it is required to gather a lot of simple sentences whose verb stands for a same meaning. One of the way to acquire those sentences is classify simple sentences to the sets of sentences which have a same verb which stands for a same meaning. We proposed before a way to classify simple sentences by calculating similarity degrees between words or between sentences with EDR dictionaries. However, it has a problem such that the meaning of a verb included in classified sentences can not be specified, and then, attaching a semantic labeling to the sentences has to be done by hand. In this paper, we propose an improved method with EDR corpora. Experimental results show that extracted data from EDR corpora is useful for attaching a semantic label to the classified sentences and for improving an accuracy .of similarity degrees.. |
| 76. |
Assessment Report of Doctoral Theses. |
| 77. |
The Evaluation for Coordinated Morphological and Syntactic Analysis Method of Japanese Language. |
| 78. |
Parallel semantic analysis by spreading and struggling activities. |
| 79. |
An Implementation of Parallel Morphological and Syntactic Analysis Algorithm in Shared-Memory Multiprocessors and Its Evaluation
我々は、実時間自然言語理解システムの構築を目指して、まず解析時間ο(n)・プロッサ数ο(n^2)の性能を持つ並列構文解析アルゴリズムを提案し、次に構文解析と形態素解析の協調動作を行うシステムの提案を行ってきた。今回は、この協調型の並列形態素・構文解析システムを共有メモリ型の並列計算機上に実装し、特に、共有メモリに対する排他制御法についての実験を行ったので、その報告を行う。. |
| 80. |
Parallel semantic analysis using a force of spreading activities of concepts and holding a relations between concepts
自然言語処理の研究において、現在求められているのは、言葉の字面の解釈よりも、文の内容自体の解釈、更には、文の意図する内容の解釈である。文の内容や意味は、使われる状況と密接に結び付く概念であるため、自然言語文の解釈を計算機で行う為には、状況や意味概念の表現、扱い方が重要である。我々は、この様な意味表現や操作を規定するにあたり、意味を事物や事象のもつ概念間の関係とその関係を保つための強さと捉えている。この強さによって概念間を活性値が伝搬し、状況に応じた意味解釈を行うと考える。本稿では、それらの表現方法についての考察を行い、それによる意味解析モデルを提案する。. |
| 81. |
Parallel Discouse Processing Model Controlled by Priority
談話処理においては情報は常に不完全であり、一意に決めることの出来ない瞹昧な処理が必ず含まれている。これらの瞹昧さに対して並列に処理を行う場合、組み合わせ爆発を避けるだけでなく、マシンの能力に応じた並列性に抑えることを考えなければならない。この為には、重要性などを表す指標を用いることにより、重要度の低い処理の動的な遅延や除去が有効である。本稿では、フレームで表現した知識に対して処理要求の活性と知識の活性という概念を導入し、活性化の度合いによる制御(priority control)を用いて、必要かつ重要な処理を優先して行う方法を提案する。. |
| 82. |
An Efficient Parallel Parsing Algorithm for General Context-Free Grammars
自然言語を解析するためには 構文情報や意味情報 文脈情報など各種の膨大な量の知識情報を用いて処理することが必要であるしかし これらの多量な知識情報を計算機で処理するためには 莫大な解析時間が必要となるため 並列処理などの手法を導入して 解析時間の短縮を計る必要がある一般に文脈自由言語の構文解析には長さが"の入力に対してO(n^2)の手数を必要とするしたがって実時間自然言語理解システムを実現するためには まず文脈自由言語をO(n)時間で解析できる並列構文解析アルゴリズムを開発する必要があるその時 使用するプロセッサ数は実現の可能性上O(n^2)以下であることが望ましいこれまでに並列構文解析のアルゴリズムがいくつか提案されてきたが これらのアルゴリズムには 解析時間やプロセッサ数が多くなりすぎるという問題がある本論文では 一般の文脈自由文法を対象とする並列構文解析アルゴリズムを提案このアルゴリズムの性能は解析時間O(n)・プロセッサ数O(n^2)でその積がO(n^3)である本アルゴリズムでは 一般の文脈自由文法からLR法に類似した手法を用いて作成したLR状態遷移図を解析制御表として使用し すべての可能性を並列に試しながら入力に非同期に解析を行う. |
| 83. |
Parallel LR Parsing for General Context-free Grammars. |
| 84. |
Parallel Lexical Analyzer for Japanese sentences
従来,形態素解析は構文解析の前処理として,構文解析から何の情報も受けずに独立に一文単位で品詞の木を作成し,それを構文解析に渡すものが多かった.しかし構文解析から文節の予測情報を受け取ることによい辞書などの探索空間が狭まり,曖昧性の削減に役立てることができるため,形態素解析は,構文解析と協調しながら解析を行う方式が良い.しかし,両解析を融合した場合には,文法全体の煩雑化や解析自体の複雑化の恐れが生じるため,各解析は,独立した構成をとるほうが良い.また,解析時間の短縮化をはかるため,両解析ともそれぞれ並列に効率よく協調しながら動作する方式を開発する必要がある.我々は,並列に効率よく動作する並列構文解析のアルゴリズムを提案してきたが,本稿では,この並列構文解析と協調しながら並列に解析を進める並列形態素解析について提案する.. |
| 85. |
Parallel Parsing for Context-Free-Grammar
The paper presents an efficiently parallel parsing algorithm for arbitrary contextfree grammars. This algorithm uses an LR transition network which is similar to a shift reduce table used by standard LR parsing algorithm. It runs in linear time for n (where n is a length of the string being parsed) ; it needs O (n2) processes, and O (n2) spaces.. |
| 86. |
The Evaluation of Datarol Programs
いままでに並列構文解析アルゴリズムについての報告が多数なされているが、それらには(1)入力に同期した解析を行うために解析時間が0(n^2)(nは入力文の長さ)となる、(2)プロセス数が指数オーダーになる、といった問題があった。我々のアルゴリズムは一般の文脈自由文法を対象とし、一般のLR法で使用されるshift reduceテーブルに改良を加えたLR状態遷移図を使用して、全ての可能性を並列に試しながら解析を行う。このアルゴリズムは解析時間が0(n)ですみ、プロセスの数と使用メモリ領域をどちらも0(n^2)に抑えている。. |
| 87. |
A Parallel Parsing for Context _Free_ Grammar
本稿では、文脈自由文法に対する並列構文解析アルゴリズムを提案する。このアルゴリズムは、LR法で使用するshift reduceテーブルに類似したLR状態遷移図を使用し、入力に対して非同期な解析を行う。まず全プログラムを非同期に並列動作させ、全ての構文木をO(n)時間で作成するアルゴリズムを示す。この時プロセスは構文木の数だけ必要となる。次に、プロセスをshift/reduceの各操作に対して同期させることでプロセス数をO(n^2)に削減するアルゴリズムを示す。このアルゴリズムの解析時間もO(n)で抑えられ（ここで、nは入力文の長さとする）、メモリー空間はO(n^2)である。The paper presents a parallel parsing algorithm for arbitrary context-free grammars. This algorithm uses a LR transition networks which is similar to a shift reduce table used by standard LR parsing algorithm. No process is not synchronized with input strings. First, we describe a basic algorithm. A number of processes are operated in parellel, but not process is synchronized. Then, we propose an efficient parallel parsing algorithm. It runs in linear time for n(where n is the length of the string being parsed); it needs O(n^2) processes, and O(n^2) spaces.. |
| 88. |
Parallel Parsing Algorithm for Context Free Grammar
自然言語処理では、解析のある時点で部分句の構文構造やその意味構造を唯一のものに確定することは不可能のため、いくつかの可能性を残しながらそれらの処理を並行させ、徐々に適当なものに絞り込んでいくという方法を開発することが必要である。そのためには構文処理と意味処理とを並列に協調させる意味理解のプロセスをモデル化することが必要であり、また構文処理、意味処理自体を並列に行う方法を開発しなければならない。このような観点から、本稿では意味処理と協調しながら構文解析を行うための、並列構文解析アルゴリズムを提案する。. |
