Welcome to toxit.info
Toxit is a data exploration platform that centers around a weighted undirected graph representation of Reddit.com community connections. Graph edges are weighted by set intersection cardinality of either moderators or commentators for a dataset under investigation. Comments are passed to a natural language processing model which returns the toxicity scoring of the comment. The number of comments above a threshold are counted and displayed in the centre of each node. Our NLP model was trained on The Measuring Hate Speech Corpus to within 0.1132 MAE over [-1, 1].
Click a node to select it. By default, this opens the navigation tab with information about its relations to other nodes. You can also drag and interact with nodes and hover to view node information.
When middle mouse or button is clicked will reset the view of the graph.
By default the side menu is in a collapsed state. When the user clicks the it will open the side menu.
Sidebar tabs with the symbol can be reordered.
Data samples change over time, You can explore different snapshots using the dropdown. You can also toggle between common moderators and commenters when viewing connections.
Change the website styling to use a dark or light theme.
Provides aggregate descriptive statistics of the selected node as well as a link to the subreddit. 'Flagged comments' is a count of comments with a toxicity score above a threshold value.
Customize graph physics.
Change the physics solver, and fine tune parameters with sliders.
Download as JSON, CSV, or Python Binary Pickle File