This is a presentation of a workshop on recommendation strategies for Personal Learning Networks, held on 11/7/2013 at the PLE conference in Berlin (DE). It discusses the matching on similarity and matching of dissimilarity.
4. Table
Introduce yourself to the others at your
table. Make a mental note of who you
want to meet up with later in the
conference.
5. Individual
Write down 10 tags - one tag per card -
that describe
• your work
• the topics you find important in the PLE
conference
6. Table
Pool all the tags in the centre of the table,
and read them all. Staple similar tags
together, and choose one marker tag for
the group.
Example: create, creativity and creation
8. One-to-one
Match on similarity with everyone at the
table
A = total number of used cards = 5 + 4 = 9
B = number of piles which have both your colours
= 1+1=2
9. One-to-One
Match on dissimilarity with everyone at
the table
Ex: Overlapping tagsets, with overlap 2:
[learning, writing, reading]
[learning, network, writing, blog, wiki]
10. Table
Who are your best matches? Share the
outcome with the others at the table. Do
the results match your initial ‘gut’ feeling?
11. What did we do?
Content
Relevance
Experience
of
Breakdown
Desire to
Connect
Similarity matching or Dissimilarity
matching?
12. Method: User profiles
• Scoop.IT profiles as the starting point
for tagsets
• Selection on the basis of Scoop.IT posts
with comments
• Keyword extraction + stemming
14. Method: User
Evaluation
1.Content Relevance: The Scoop.IT contains
new and relevant content for me
2.Experience of Breakdown: This Scoop.IT
feed makes me re-assess my thoughts about this
topic
3.Desire to Connect: I would like to engage in
a discussion with the curator of this Scoop.IT
feed
15. Results
1.Experience of Breakdown strongly
correlates with Desire to Connect
If you feel you have learnt something
from someone, you very likely want to
connect
2.Matching on dissimilarity is better at
predicting Experience of Breakdown
16. Plenary session
• Questions? Comments?
• User profiles: How can qualitative
profiles be improved?
• Matching: How can qualitative
matching be improved?