Discovering, Visualizing and Sharing Knowledge through Personalized Learning Knowledge Maps The AWAKE Project personal knowledge structure (user A) A m E ?? y map grid semantische ? C d achsen o netz awarenes cs kuns s multi- Inte t Jasminko Novak, user rnet mixed C reality collabora distribute S tive d interactiv C Michael Wurst m inter e W a e activ s r d e art virtual t t reality i a a g l e s t ? ? Data a Modification Agents Sources COR A MyAgent Agent B Agent C Agent D Agent E g Learned User Data D a t a b DIS a s e D a a D t t b a b a a a s s e e E i S e m i a u t o n o m d u r c S S e h m m e i i a a u u t t n o o n o o m m d d u u r c r c h h Cluster Splitter Mixer Cross Delineate Profiles Agent Adapters n C Data Entry Modifiers g
Context and Challenge Support for heterogeneous expert communities Discovering connections across different domains Exchange of knowledge between individual experts Challenge: Capture, visualize and exchange existing knowledge of user groups 2
Practical Approach Situation: Group of users explores an 1) Creating a context for the information space interpretation of users’ Goal: Capture user knowledge reflected in their structuring actions interaction with information 2) Acquisition and visualization of similarity relations between users, documents and topics Knowledge Maps 3
Agent Architecture Topic Map Agent Clustering Search Editor Agent Agent Map Personalization Agent Personalized Navigation (online) Shared Data Space Data Analysis Heterogeneous (offline) Document Sources Learning Feature Extraction Agents Agents 4
Personal Agents System generated Knowledge Maps Art • System generated Knowledge Maps Interactive • Detailed document information • Related (similar) documents Video Sound • Personal Map Editor • Search for Personal Maps Agents Show • Document pool personalization • Topic network 5
Personal Agents System generated Knowledge Maps Art • System generated Knowledge Maps Document title Interactive • Detailed document information Author(s) Abstract • Related (similar) documents Keywords Video Sound ... • Personal Map Editor • Search for Personal Maps Agents Show • Document pool personalization • Topic network 6
Personal Agents System generated Knowledge Maps Art • System generated Knowledge Maps Interactive • Detailed document information • Related (similar) documents Video Sound • Personal Map Editor • Search for Personal Maps Agents Show • Document pool personalization • Topic network 7
Personal Agents Personal Knowledge Maps • System generated Knowledge Maps Video • Detailed document information Sound • Related (similar) documents • Personal Map Editor New Cluster • Search for Personal Maps • Document pool personalization • Topic network 8
Personal Agents Search for Personal Knowledge Maps • System generated Knowledge Maps • Detailed document information Video Video Sound Sound • Related (similar) documents New Sound Cluster New • Personal Map Editor Cluster New Cluster • Search for Personal Maps • Document pool personalization • Topic network 9
Personal Agents Personalized Knowledge Maps • System generated Knowledge Maps Learn&Apply to Document Pool • Detailed document information • Related (similar) documents • Personal Map Editor Video • Search for Personal Maps • Document pool personalization Sound • Topic network New Cluster 10
Personal Agents Creating a Collaborative Topic Net • System generated Knowledge Maps multi-user • Detailed document information • Related (similar) documents distributed • Personal Map Editor • Search for Personal Maps interactive • Document pool personalization theatre virtual • Topic network 11
Personal Agents 12
Agent Architecture Topic Map Agent Clustering Search Map Editor Agent Agent Personalization Agent Personalized Navigation (online) Shared Data Space Heterogeneous Data Analysis Document (offline) Sources Feature Extraction Learning Agents Agents 13
Shared Data Space Relationships between documents, topics and users Visualization Client Personal Agents Raw data documents Extracted Meta users Information topics Feature Extraction Shared Data Space Learning Agents Agents 14
Content and Context Analysis Learning (similarity) relations between documents, topics and us ers Learned relationships Reference Analysis Agent Combined Similarity Text Measure Analysis Agent content 15
Content and Context Analysis Learning (similarity) relations between documents, topics and us ers Learned relationships Reference Analysis Agent Context Analysis Combined Similarity Text Agents Measure Analysis Agent context content 16
Content and Context Analysis 1 0,9 0,8 0,7 avg. err. Text Context Combination 0,6 0,5 0,4 0,3 1 3 5 7 9 11 13 15 17 19 Number of maps 17
Content and Context Analysis Some Relevant questions from a Machine Learning Viewpoint • Amount of user interaction data needed to learn the desired concepts? • Robustness of the extracted knowledge against noise or malicious manipulation • Local/Global Pattern (e.g. how are minor opinions reflected in the resulting structures?) 18
Application Domain netzspannung.org • Knowledge portal connecting media art, design and technology • Collaborative information pool (projects, events...) • Very heterogeneous content • different categorization schemes • constantly growing 19
Summary and Future Work Summary • Model and prototype for capturing, visualizing and exchanging knowledge • Agent technology as software architecture and data integration model • Machine Learning methods for combining context and content data Future work • Evaluation and user testing • Improvement of visualization and data analysis methods • Support for the Topic Map Format 20
Project & Partners Project partners � Fraunhofer Institute for Media Communication MARS Exploratory Media Lab (Project Leader) � Fraunhofer Institute for Industrial Engineering Competence Center Softwaretechnologie & interactive Systems � University of Dortmund Artificial Intelligence Unit � University of Siegen Dept. for Parallel Systems Thank you for your attention! 21
Recommend
More recommend
