The similarity matrix visualization is for displaying the similarity between topics. It can be ordered based on name or based on cluster by selecting the appropriate option in the drop-down menu. The topics are clustered based on similarity of users talking about them.

Since our visualization’s main objective was to effectively convey the links between users and the topics they talk the most about, we looked into different forms of graphs and node-link representations. Initially we considered implementing a bipartite graph but decided against it as it would not be effective to use it to show links between 50 topics and 280 users. We then decided that the best way to represent this data will be with node-link clusters, topics and users being different clusters with different shapes. We used size and color to encode the popularity of the users and the topics. The similarity matrix was unanimously chosen as the most appropriate way to represent the similarity between the topics.

• To minimize scrolling as much as possible, we decided to place the cluster on the left and the information pertaining to the node(s) of interest will be displayed on the right.
• To have the maximum amount of screen space to display the information without cluttering the screen and causing a cognitive overload to the user, we used an overlay to have the description of our data and a link to the documentation.
• Information on the cards are displayed in a nested manner and not all at once for the same reason.
• Tooltips are used to get a quick preview of the node’s information without having to click on it.
• On the similarity matrix, hovering over a particular clustered cell on the matrix will highlight the topics to give the user feedback of their position on the matrix.

U240 is the biggest user bubble with the highest rank and most number of comments made and received so he must be the Physician

There are few topics that have been repeated; Complementary-alternative medicine(2), Diet interacting with medications(2), Disease course-monitoring-testing(5), Goals of treatment(4), Hope and support - General(2), Side effects due to steroid drugs(2), Talking to the doctor(2), Using last resort medications(2.) Despite the topic name being the same, they have entirely different users talking about them.

Because of the same reason topics with the same name are in different clusters.

Topics with the same name have different keywords

Hope and Support is a largely spoken topic which did not come under the Physician’s top 5 most spoken topics

What can do you with this visualization?

Clicking on the user nodes will give you:
• The top 5 topics they spoke about
• Card on the right displays information of
• Rank of user
• Posts created by the user
• Comments received
• Comments made by the user
• On clicking the down arrow symbol you can get a list of top 5 topics the user spoke about

Clicking on the topic nodes will give you:
• All the users that have the topic in their top 5 most spoken topics
• Card on the right displays information of
• Rank of topic
• User popularity of topic
• On clicking the down arrow symbol you can get a list of top 10 users and keywords associated to that topic