Design

We aim to visualize the diffusion of the topic among the users, with respect to the time.

Three views:

Geo-based View Semantic-based View Link-based View

Geo-based View

The color of an area is determined by the posts in that area

The diffusion along the geo-dimensions can be visaulized in this way

Local topics, like games, the weather

Show the distinction of different regions' interest of the same topic, like 春晚

Sematic-Based View

Use topic models: Given a user’s weibo, we can calculate related topics and similarity matrix.

Use MDS(Multi-Dimensional Scaling)to put these points on the view according their similarity matrix.

Implemented using Java-R interface(JRI), take the advantage of R to compute the similarity matrix.

Combination of visualization, machine learning and natural language processing.

The following linkage is the natural linkage of the Weibo community

Webpage Design

Top: Graph Chart, one of the three views

Below: a line chart of the repost counts against time. Linked to the previous chart.

Side bar: statistics charts about the related topic

Problems

Dataset scale: The API quota is the bottle neck

Snapshot

 
public_course/visclass_f12/project/group_5.txt · Last modified: 2013/01/22 23:11 by f12_zhanghantian