Discussed which dataset to use. Outcome: use Reddit data
We will have teem meetings every Wednesday after the lecture and every Thursday after our TA meeting.
- Use Python and Flask as back-end
- Front-end to use will be decided base don features we want to visualize
- Use git branching model, where master=stable, all other features are developed in a separate branch. Changes go through a feature review via pull request.
- Focus on analytics primarily
Every team member creates 2 paper models of a feature or workflow we want to visualize. Next week we go over the features and discuss any changes
This list is non-exhaustive. If you have other idea, please add it below or into the board.
- Visualize how communities evolved over the years -> timeline
- During election cycle, explore relationship between communities
- Sentiment analysis
- Which ones link the most to each other
- Most popular ones
- Visual link explorer -> seo backlinks
- Wordcloud (realtime, get body content and generate)
- (Related) Feature vector analysis: look at the feature vector of a post, find other posts with similar percentages and where they link
- Chain links analysis: how many levels deep
- Echo chamber detection
- Find related subreddits by looking where they link the most
- Thread brigading detection based on sentiment
The dataset contains post id. The content of the post can be retrieved by the following API call: https://www.reddit.com/comments/commentIDhere/.json
- Discussed high level paper prototyping for 2 views
- Discussed architecture of the solution and how we will implement it: use Python (Flask) with Pandas for the data processing. Frontend will be up to the implementation teams per view, including D3 and Bokeh for interactive views.
- Discussed what to ask Georgi for next day and get initial design approved.
- Started working on data processing with Pandas to create an API to retrieve the data
- Make functional prototype by next week
- Visualization design
- Navigate over time in discreet time intervals on the bottom, support playing animation over time
- Main screen displays graph of interactions between subreddits, where line thickness and color represent different features. Optionally the placement of the graph nodes could convey additional information.
- Hover or click on subreddit shows a radar plot for a given subreddit, showing relationships between these features
- After meeting: continued work on API and started with visualization skeleton first screen
- Finalized static API to return data for frontend and began working on producing results in the backend
- Created frontend skeleton structure with D3.js
- Prototype looking fine.
- Important to focus on design: colors, user usability/interaction
- Combine the invidual parts of the prototype and design this nicely
- Fix the bar chart with time line soon
- Progress was fine
- Having a complete demo ready for next week
- Prepare some guidance for the user
- Use an appropriate color scheme
- Prepare the user stories
- Add feedback actions, like zoom in, highlight selected nodes etc
- Check the app for possible delays
- Bar charts need to be refined and integrated
- Final presentation, report, and video content was discussed
- motivations
- how our vizes are different from a regular graph
- identify passive elements (e.g. background) and active elements (e.g nodes, seach bar)
- justify color use, element positions, font used, etc.
- what are the findings of our vizes
- what we have archived
- how components communicate with each other