Social networks have become extremely popular, and with their popularity has come a desire to learn and extract information from these networks. However, the traditional node-link diagrams often used to portray social networks is often misleading because they are absolute when there is often uncertainty in such networks. For instance, consider the following diagram of, say, four email accounts.
Here every node is an account and an edge exists between two nodes if an email was exchanged between these two accounts. If what we want to see is a graph of "who contacted who," this diagram is sufficient; however, often we want to know more information about the relationship between nodes and the role/roles a particular node plays in a network. For example, say we want to find out what pairs of nodes constitute "friends" in the above diagram. A single email between two people probably does not mean these two people are friends. To try to infer this information, we need to know more about the emails and the people. Now, each node and link in our diagram no longer has a single meaning, but rather has a set of attributes attached to it. With this added information, we might get something like the following picture:
In this case, we might have mapped the thickness of an edge to the number of emails exchanged between two people and its opacity to the "quality" of the emails. While the quantity of emails we can be certain about, there is nothing certain about the quality of these emails. But, say we have some methods for modeling the quality of emails, maybe "friendship" quality and "professional" quality. Now, not only do we have uncertainty information in our graph, but also multiple graphs: a "friendship" graph and a "professional" graph. How do we represent a single "uncertain graph" as well as visually comparing two or more of them? This project will try to answer that question.
The end result of this project is to create a software tool to:
| Milestone | Expected Completion Date | Possible Challenges |
|---|---|---|
| Finish Preliminary Research | 4/19 | None |
| Present a set of 3 or more visualization techniques for viewing a single graph with uncertainty attached to the nodes and edges | 4/26 | Node |
| Present a set of 3 or more new visualization techniques for viewing ego networks and compare with current techniques for visualizing ego networks. | 5/3 | None |
| Present a set of one or more techniques for visually comparing 2 or more "uncertain graphs." | 5/10 | None |
| Present a tool which combines the 3 types of visualizations thus far (i.e. single graph, ego networks, and comparison of multiple "uncertain graphs." | 5/17 | None |
| Search for users to get feedback on techniques thus far. | 5/24 | Finding users |
| Revise visualizations based on feedback. | 5/31 | None |