I made these slides for 1st round of 2nd semester M Tech seminars. These are based on the work done by DOCEAR team and research papers by them. I also referred other material on mind maps to understand the concept.
2. • Mind-mapping is a technique to record and organize
information, and to develop new ideas [Holland et al.
2004]
• Mind-maps are similar to outlines and consist of three
elements, namely nodes, connections, and visual clues.
• To begin mind-mapping, users create a root node that
represents the central concept that the users are
interested in [Davies 2011]. To detail the central concept,
users create child-nodes that are connected to the root
node. To detail the child-nodes, users create child-nodes
for the child-nodes, and so on.
3.
4. • Faste and Lin [2012] evaluated the effectiveness of mind-
mapping tools and developed a framework for
collaboration based on mind-maps.
5. • Kudelic et al. [2012] created mind-maps from texts
automatically.
AND
• Bia et al. [2010] utilized mind-maps to model semi-
structured documents, i.e. XML files and the
corresponding DTDs, schemas, and XML instances.
6. • Jamieson [2012] researched how graph analysis
techniques could be used with mind-maps to quantify the
learning of students.
AND
• Somers et al. [2014] used mind-maps to research how
knowledgeable business school students are.
7. • By Joeran Beel, Stefan Langer, Marcel Genzmehr, Bela
Gipp
• Published in UMAP 2014
• Presented 8 ideas on how mind mapping can be used in
IR applications
• User modelling was the most feasible use case
• Proposed to implement a prototype- Research paper
recommender system
8. • By Joeran Beel, Stefean Langer, Bela Gipp, Andreas
• Published in D-lib magazine of Digital Libraries 2014
AND ACM/IEEE Joint Conference on Digital Libraries
2014
• Introduced 4 datasets which contains metadata about
research articles, details of Docear’s users and their
mind-maps and recommendations they received.
9. • By Stefan Langer and Joeran Beel
• Published in a workshop: Dimensions and Design at the
ACM RecSys 2014 Conference
• Proved that user characteristics affect the performance of
recommender system.
10. • 3 Approaches of Content Based Filtering to build user
models:
• Use the terms from last edited node (CTR 0.2% - 1%)
(MindMeister)
• All terms in User’s current Mind-map
• All terms from all Mind-maps user has ever created.
11. • 28 variables
• Number of mind-maps
• Number of nodes
• Size of user model
• Whether to use only visible nodes
• Weighting schemes
• Docear’s mind-map specific user modelling approach
12. • Mind-map and node selection
• Mind-map selection
• Node selection
• Node extension
• Node and feature weighting
• User model size
13.
14.
15.
16.
17.
18.
19. • Based on depth of a node
(weighted stronger the deeper they are- improved
CTR of 5.62% from 5.15%)
• Based on number of children (weighted stronger for more
number of children- improved CTR of 5.17% from 5.01%)
• Based on number of siblings (More siblings-Higher
weights – improved CTR of 5.41% from 5.01%)
20.
21. • Combined optimal values of all variables in single
algorithm
• Used 75 Most Recently MOVED nodes from past 90
days.
• Nodes Expansion
• Term Weighting (TF-IDuF)
• 35 highest weighted nodes were User Models
• Comparison with 4 baselines
• Stereotype
• Recently modified node
• All nodes of current mind-map
• All nodes of all mind-maps
22.
23. • BEEL, J., LANGER, S., GENZMEHR, M. AND GIP, B., 2014.
Utilizing Mind-Maps for Information Retrieval and User
Modelling. Proceedings of the 22nd Conference on User
Modelling, Adaption, and Personalization (UMAP
• BEEL, J., LANGER, S. AND GIPP, B., 2014. The Architecture
and Datasets of Docear’s Research Paper Recommender
System. In Proceedings of the 3rd International Workshop on
Mining Scientific Publications (WOSP 2014) at the ACM/IEEE
Joint Conference on Digital Libraries (JCDL 2014).
• STEFAN LANGER, BEEL, 2014. Comparability of
Recommender System evaluations and characteristics of
docear’s users. In ACM RecSys 2014 conference
• STEFAN LANGER, BEEL, GIP 2014. Mind-Map Based User
Modeling and Research Paper Recommender Systems in
ACM Transactions
• www.docear.org