README.org 3.47 KB
Newer Older
1
* CSEDU-2019
2
** Resources and annexes to the paper "[[file:tex_draft.pdf][Towards Visual Explorations of Forums' Collective Dynamics in Learning Management Systems]]"
3

4
5
6
*** Images
 The images in the paper are sometime small.  We offer high enough resolution images to see the details.
**** Figure 1
Malik Koné's avatar
Malik Koné committed
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
[[file:images/discussion.png]]

 Illustration of how the strength of actors' ties (or links) varies as a function of time and topic overlap.  Thread 1 corresponds to actor-topic dynamic (\ref{eq:1a}) where <img src="https://latex.codecogs.com/svg.latex?B"/>'s late post after <img src="https://latex.codecogs.com/svg.latex?A"/>'s \nth{1} publication does not correlate strongly enough to create the link from <img src="https://latex.codecogs.com/svg.latex?A"/> to <img src="https://latex.codecogs.com/svg.latex?B"/>.  But <img src="https://latex.codecogs.com/svg.latex?A"/>'s \nth{2} post is timely enough, although not exactly on the same topic as <img src="https://latex.codecogs.com/svg.latex?B"/>'s message, to create the tie <img src="https://latex.codecogs.com/svg.latex?A \dashrightarrow B"/> drawn as a dashed arrow.  In thread 2, in addition to the tie <img src="https://latex.codecogs.com/svg.latex?B \dashrightarrow A"/>, we have a topic overlap and time proximity between <img src="https://latex.codecogs.com/svg.latex?C"/> and <img src="https://latex.codecogs.com/svg.latex?A"/>.  This makes a the strong tie <img src="https://latex.codecogs.com/svg.latex?A \to C"/>.

**** Figure 2
[[file:images/fu.png]]

   iForum's Dashboard \citep{Fu2017} showing (a) overall changes of post in the forum, (b) a thread representation, (c) discussions in packed forms, (d) the social network and (e) the details of a discussion.

**** Figure 3
[[file:images/convis.png]]

   Convis Dashboard \citep{Hoque2016} helps explore conversations.  On the left, the topics found in the forum using \gls{lda} and organized hierarchically.  In the middle, the colored rectangles show a sentiment analysis for each message.  Each message is linked to his author place on a semi-circle and thus creating a social network.  Finally, on the right, Convis display the detail of the conversation.

**** Figure 4
[[file:images/func.png]]
 Graphical proposition for the function <img src="https://latex.codecogs.com/svg.latex?I(s,r)"/> of the messages' interaction strength.  <img src="https://latex.codecogs.com/svg.latex?s"/> is the ``internal'' strength of two messages based on content, time and social network structure.  <img src="https://latex.codecogs.com/svg.latex?r"/> is ``external'' requirement, it is a parameter set by the observer.
 
25

Malik Koné's avatar
Malik Koné committed
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
**** Figure 5
[[file:images/cycles.png]]

   Interactions cycles built from a bi-party actor-topic graph Thread 3 and 4 are transformed to an actor-actor graph.  Dotted arrows denotes weaker links.

**** Figure 6
[[file:images/pipeline.png]]
Data Analysis cycle.

**** Figure 7
[[file:images/dynco_portrait.png]]
   Visualizations from the FFL dataset.

**** Figure 8
[[file:images/uvci_portrait.png]]
   Detail of \gls{vuci} conversation taking place in 2 hours.  Each circle is a message, and hovering over them brings up its content.  The circle's size is proportional to their content's length.  Message are layered vertically by actors.  On the left is an indication of the actors total messages count.

**** Figure 9
[[file:images/evolution.png]]
   At the top (a) is half of compound yearly actor-actor network.  The three bottom images (b), (c) and (d) are closeup around actor 642 during the quarters of the year.