Conversation Visualization
The Graph Conversation Nodes is a visual representation of the data and the conversation flow in the form of Network graphs. Network graphs, also known as node graphs, display relations between elements (nodes) using links and semantic metadata. Network graphs are particularly useful for visualizing a conversational node path or tree. We can use a network graph to answer questions like:
What path did a user take through a conversation?
What nodes were most frequented over a series of sessions?
What conversational themes were most popular?
Note: In order to view the Graph Conversation Nodes, you will first need to annotate the corpus conversations.
Demo video
Features
Nodes are colour coded by the node category they belong to
Graph legend can be flexed to display any number of node groups
Node names and edge hit values can be surfaced by hovering over the node/edge
Zoom, click and drag functionality
Better able to read and maneuver overlapping edges by zooming into detailed areas and dragging/rotating nodes
Node and edge labels remain constant when zooming
Enables 360° view of the network
Size of the node represents the number of hits, the larger the node means more hits
Lines between indicate the number of times going between the nodes in all conversations-
thicker line means more linksCategorisation done by taking the first section of the name that is separated by full stops
Clicking on a node will load the Sankey diagram below this has more detailed information on the relationships of the nodes
Hover over node will show how many hits, similar to the conversation node graph
Hover over link will show how many times there has been a link between the two nodes
Entries is the number of times a conversation has included this node
Entries % is the percentage of times a conversation has included this node
Ends is the number of times it has gone from this node to the END node
Ends % is the percentage of times a conversation has gone from this node to the END node
during the lifetime of the DP. You can drag the nodes in order to see more informationFollows same color coding as the node graph
Contact