Automatic cLasification d

Slides:



Advertisements
Similar presentations
A Comparison of Implicit and Explicit Links for Web Page Classification Dou Shen 1 Jian-Tao Sun 2 Qiang Yang 1 Zheng Chen 2 1 Department of Computer Science.
Advertisements

GMD German National Research Center for Information Technology Darmstadt University of Technology Perspectives and Priorities for Digital Libraries Research.
Structured Design The Structured Design Approach (also called Layered Approach) focuses on the conceptual and physical level. As discussed earlier: Conceptual.
Content-based Recommendation Systems
Data Mining for Web Personalization
Towards a framework that allows using a cognitive architecture to personalize recommendations in e-commerce Jordi Sabater-Mir, Joan Cuadros and Pere Garcia.
User modelling for adapted accessible interaction Julio Abascal #, Olatz Arbelaitz *, Myriam Arrue #, Javier Muguerza * # EGOKITUZ: Laboratory of HCI for.
Interception of User’s Interests on the Web Michal Barla Supervisor: prof. Mária Bieliková.
1 CS 501 Spring 2002 CS 501: Software Engineering Lecture 11 Designing for Usability I.
Web Mining Research: A Survey Authors: Raymond Kosala & Hendrik Blockeel Presenter: Ryan Patterson April 23rd 2014 CS332 Data Mining pg 01.
Recommender Systems Aalap Kohojkar Yang Liu Zhan Shi March 31, 2008.
WebMiningResearch ASurvey Web Mining Research: A Survey Raymond Kosala and Hendrik Blockeel ACM SIGKDD, July 2000 Presented by Shan Huang, 4/24/2007.
1 Chapter 10 Introduction to Machine Learning. 2 Chapter 10 Contents (1) l Training l Rote Learning l Concept Learning l Hypotheses l General to Specific.
Web Mining Research: A Survey
© Prentice Hall1 DATA MINING TECHNIQUES Introductory and Advanced Topics Eamonn Keogh (some slides adapted from) Margaret Dunham Dr. M.H.Dunham, Data Mining,
Web Mining Research: A Survey
Web Mining Research: A Survey
WebMiningResearchASurvey Web Mining Research: A Survey Raymond Kosala and Hendrik Blockeel ACM SIGKDD, July 2000 Presented by Shan Huang, 4/24/2007 Revised.
02 -1 Lecture 02 Agent Technology Topics –Introduction –Agent Reasoning –Agent Learning –Ontology Engineering –User Modeling –Mobile Agents –Multi-Agent.
Extracting Test Cases by Using Data Mining; Reducing the Cost of Testing Andrea Ciocca COMP 587.
12 -1 Lecture 12 User Modeling Topics –Basics –Example User Model –Construction of User Models –Updating of User Models –Applications.
Introduction to Data Mining Engineering Group in ACL.
CS Machine Learning. What is Machine Learning? Adapt to / learn from data  To optimize a performance function Can be used to:  Extract knowledge.
Wang, Z., et al. Presented by: Kayla Henneman October 27, 2014 WHO IS HERE: LOCATION AWARE FACE RECOGNITION.
Towards Ad-hoc Situation Determination Graham Thomson, Paddy Nixon and Sotirios Terzis.
The 2nd International Conference of e-Learning and Distance Education, 21 to 23 February 2011, Riyadh, Saudi Arabia Prof. Dr. Torky Sultan Faculty of Computers.
CHAPTER 12 ADVANCED INTELLIGENT SYSTEMS © 2005 Prentice Hall, Decision Support Systems and Intelligent Systems, 7th Edition, Turban, Aronson, and Liang.
Intrusion Detection Jie Lin. Outline Introduction A Frame for Intrusion Detection System Intrusion Detection Techniques Ideas for Improving Intrusion.
Personalization in Local Search Personalization of Content Ranking in the Context of Local Search Philip O’Brien, Xiao Luo, Tony Abou-Assaleh, Weizheng.
Copyright R. Weber Machine Learning, Data Mining ISYS370 Dr. R. Weber.
Distributed Networks & Systems Lab. Introduction Collaborative filtering Characteristics and challenges Memory-based CF Model-based CF Hybrid CF Recent.
Adaptive News Access Daniel Billsus Presented by Chirayu Wongchokprasitti.
IE 585 Introduction to Neural Networks. 2 Modeling Continuum Unarticulated Wisdom Articulated Qualitative Models Theoretic (First Principles) Models Empirical.
Recommendation system MOPSI project KAROL WAGA
PAUL ALEXANDRU CHIRITA STEFANIA COSTACHE SIEGFRIED HANDSCHUH WOLFGANG NEJDL 1* L3S RESEARCH CENTER 2* NATIONAL UNIVERSITY OF IRELAND PROCEEDINGS OF THE.
Context-Aware Interactive Content Adaptation Iqbal Mohomed, Jim Cai, Sina Chavoshi, Eyal de Lara Department of Computer Science University of Toronto MobiSys2006.
Testing & modeling users. The aims Describe how to do user testing. Discuss the differences between user testing, usability testing and research experiments.
Chapter 12: Web Usage Mining - An introduction Chapter written by Bamshad Mobasher Many slides are from a tutorial given by B. Berendt, B. Mobasher, M.
10th International Baltic Conference on Databases and Information Systems July 8-11, 2012, Vilnius, Lithuania Learner Model’s Utilization in the e-Learning.
Personalized Course Navigation Based on Grey Relational Analysis Han-Ming Lee, Chi-Chun Huang, Tzu- Ting Kao (Dept. of Computer Science and Information.
Data Mining: Knowledge Discovery in Databases Peter van der Putten ALP Group, LIACS Pre-University College LAPP-Top Computer Science February 2005.
Initial literature survey M. Shuaib Karim Institute of Software Technology and Interactive Systems Vienna University of Technology – Austria.
Web-Mining …searching for the knowledge on the Internet… Marko Grobelnik Institut Jožef Stefan.
Digital Libraries1 David Rashty. Digital Libraries2 “A library is an arsenal of liberty” Anonymous.
Introduction to Data Mining by Yen-Hsien Lee Department of Information Management College of Management National Sun Yat-Sen University March 4, 2003.
User Modeling and Recommender Systems: Introduction to recommender systems Adolfo Ruiz Calleja 06/09/2014.
Information Design Trends Unit Five: Delivery Channels Lecture 2: Portals and Personalization Part 2.
Detecting Sequences and Cycles of Web Pages Narayan L. Bhamidipati and Sankar K. Pal Indian Statistical Institute Kolkata.
Identifying “Best Bet” Web Search Results by Mining Past User Behavior Author: Eugene Agichtein, Zijian Zheng (Microsoft Research) Source: KDD2006 Reporter:
Adaptive Faceted Browsing in Job Offers Danielle H. Lee
Learning Kernel Classifiers 1. Introduction Summarized by In-Hee Lee.
Bringing Order to the Web : Automatically Categorizing Search Results Advisor : Dr. Hsu Graduate : Keng-Wei Chang Author : Hao Chen Susan Dumais.
Adaptivity, Personalisation and Assistive Technologies Hugh Davis.
Tutoring & Help Systems Deepthi Bollu for CSE495 10/31/2003.
Introduction to Machine Learning, its potential usage in network area,
Data Mining: Concepts and Techniques
User configurable advertising profiles applied to Web page banners
Improving searches through community clustering of information
DATA MINING © Prentice Hall.
Introduction Characteristics Advantages Limitations
Web Mining Ref:
Agenda About Screenpeaks The technology Content interfaces.
CH. 1: Introduction 1.1 What is Machine Learning Example:
Web Content Accessibility Beata M. Ofianewska (DG COMM) 7 December 2006 December 2006 COMM C2.
Research Areas Christoph F. Eick
Boštjan Kožuh Statistical Office of the Republic of Slovenia,
Towards a Personal Briefing Assistant
Pilar Orero, Spain Yoshikazu SEKI, Japan 2018
Testing & modeling users
Web Mining Research: A Survey
Presentation transcript:

Automatic cLasification d a PArallelism Data Mining based User Modeling Systems for Web Personalization applied to people with disabilities J. Abascal, O. Arbelaitz, J. Muguerza, I. Perona Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Contribution of Data Mining Schedule Introduction Contribution of Data Mining Profile generation Exploitation Summary Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Objective in any context: Introduction Objective in any context: To adapt web pages to the need of the users Adaptation becomes especially critical when the users have special needs Blind people: needs sounds (speech synthesis)‏ Low vision: large images and font sizes, bright colours … Motor problems: simple links or buttons Cognitive disabilities: simpler texts We need to: Build a model of the user that aggregates its main characteristics Use this model to perform actions that make easier its information acquisition in the Web Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

How would we like to do this adaptation? Introduction This way the adaptive system can adapt the interaction to the concrete users How would we like to do this adaptation? Without the user doing an explicit demand Automatically and dynamically adapting to the general characteristics of the user and the moment situation (mood, physical state, used device…) Different kinds of features can be modeled to adjust the interface User Navigational behavior Preferences Physical sensory or cognitive restrictions Context Moment interest Device Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Adaptive systems composed of Introduction Adaptive systems composed of Modeling component profiles and stereotypes to make assumptions about the characteristics of the user Usually built by means of ontologies (concept hierarchies)that allow to store, manipulate and extract assumptions from data about the user, its context, tasks, etc. Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Two options to acquire information: General schema Ontology User Profiles Adapted interface User’s information Two options to acquire information: Manually Automatically Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Directly designed by experts in the area. Rule based approach Manual option Directly designed by experts in the area. Rule based approach Rules generated based on the experts’ knowledge of the needs of different kinds of users Drawbacks: Profiles are artificially generated Need of the expert each time the interface needs to be adapted to a new kind of user Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Automatic option When the user is a person with disabilities, data mining is a way to automatically process information about the uses of the person Data mining for web personalization is based on statistical data obtained from real navigation data When the characteristics of the user change, collected data allows the automatic change of the interaction schema Learning from the own interaction allows maintaining a dynamic profile of the user, avoiding the application of all assumptions when the interest, characteristics or circumstances of the user change Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Different approaches to build models Automatic option Different approaches to build models Based on previous information from that user such as the logs of previous navigations (content-based approach) Information about groups of users with similar characteristics (collaborative approach) Our aim: to combine them to find a trade-off high specialization/computationally too expensive Obtaining user information in automatic approaches Two main sources: Obtained in the client part: desktop applications… Obtained in the server: browsing history… Using some of them can violate the user’s privacy Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Techniques: unsupervised learning techniques or clustering techniques Automatic option Techniques: unsupervised learning techniques or clustering techniques Output: sets of users with similar characteristics or needs Need of a distance metric In a determined dimensional space: Manhattan, Euclidean distance, cosine similarity … For sequences (click streams in a web, visited web pages, etc.) Edit distance Normalized Compression Distance… The used clustering techniques (SAHN, Fixed-width or Leader algorithm, k-means) will depend on: The nature of the data (vector like, sequence like) The selected distances Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Automatic option To establish the profiles related to each cluster or group of similar users we propose: Meta-learning techniques based for example in classification trees to profile the built clusters Paradigms such as association rules or frequent episodes so that we can predict the most probable transitions between links for a kind of user Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Automatic option Advantages: The learners can automatically generate profiles for new kinds of users Profiles are generated based in real browsing data (not artificial) New profiles can appear if the users change (incremental learning, enrich the ontology) Profiles based on the concepts in the ontology (different combinations) Profiles with new concepts. Machine learning used to validate new concepts Need of an significance analysis Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Exploitation . A user navigating in the web needs to be matched with one of the previously generated profiles or a new one Ontology User Profiles Adapted interface ?? Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Two main approaches to decide the profile of a user Exploitation Two main approaches to decide the profile of a user Explicit: using some questionnaire (game??) Disturbs the user with an explicit information demand making him/ her feel uncomfortable. There is high probability of the user answering with not the whole true about his/her characteristics Will not adapt to the conditions of a user in an concrete moment (tired or not, in a hurry…) just adapted to one of the existing profiles Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Implicit: based on the browsing information of the user Exploitation Implicit: based on the browsing information of the user This approach has the problem of the ti (cold start) The user does not feel uncomfortable There is no lying possibility Using machine learning can adapt to the conditions of a user in an concrete moment (tired or not, in a hurry…) Techniques: supervised classification based techniques Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Data Mining techniques: Summary Goals: Acquisition of user information to build profiles Classification of new users on previously defined profiles or stereotypes or a new one Data Mining techniques: Unsupervised techniques combined with classification trees or similar Supervised techniques Data Mining contributions: From the user point of view: comfortable Incremental learning capacity. Adapting to new kinds of users of new situations Enrichment of ontologies (hybrid solutions) Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores

Thanks for your attention!!! http://www.sc.ehu.es/aldapa/ olatz.arbelaitz@ehu.es Konputagailu Arkitektura eta Teknologia Saila Departamento de Arquitectura y Tecnología de Computadores