Research Direction

[mantzaris]

Potential research questions:

• Produce the bipartite network with tags to emotion clusters (tag<->emotion)

• Produce two separate networks, one for the emotion clusters as nodes with memes as edges between them, and then a network for the tags with the description co-location (co-occurrence) edges between them.

• Produce a network with both types of nodes (emotion cluster centers and tags) and all 3 types of edges (tag<->tag / emotion<->emotion / tag<->emotion).

• Produce a graph where there are the different images as nodes and tags as nodes, where the edges are between images and tags (image<->tag), and then also another graph where the tags also link between themselves (image<->tag / tag<->tag). Do the same with the emotion cluster centers in place of the tag as nodes (image<->emotion) and also (image<->emotion / emotion<->emotion (via memes)). Then all types together (image<->tag / image<->emotion) and finally with all types of edges (image<->tag / image<->emotion / tag<->tag / emotion<->emotion).

• What are the use cases of these graphs? (for the user)

• What unique behaviors can be uncovered?

• If these networks are of any value, should they be available to the user as a tool within Tagasaurus?

Summary of possible network visualizations
(tag<->emotion) (bi-partite)
(tag<->tag / emotion<->emotion)
(tag<->tag / emotion<->emotion / tag<->emotion)
(image<->tag)
(image<->tag / tag<->tag)
(image<->emotion)
(image<->emotion / emotion<->emotion)
(image<->tag / image<->emotion)
(image<->tag / image<->emotion / tag<->tag / emotion<->emotion)

@randyll @MichaelHopwood @phuongpho