• NATE SUMMARY FUNCTIONS, REPORTS, AND VISUALIZATIONS:
• The story of the kitchen sink

• Network

  • of Nodes

  • of Edges

  • of Isolates

  • of Subcomponents

  • of Subcomponents > 1

  • Average size of subcomponents
  • Average size of subcomponents that are NOT the giant component
  • Directed, Undirected, or Mixed?
  • Bipartite?
    -- If Directed: DAG?
    -- If Directed: Average Indegree, average Outdegree
    -- If Not Directed: Average degree
  • Node Attributes
  • Edge Attributes

** Social Network Visualizations:
-- Plot histogram of degree (indegree and outdegree if they exist)
-- Plot histogram of MGD
-- Plot histogram of ESP
-- Plot network
--- Edge thickness by weight
--- Edge thickness by similarity score

• Network SimSummary

• Average similarity
• Average dissimilarity

** Network SimSummary Visualizations

• Plot histogram of node-node similarity scores
• Plot histograms of all similarity scores calculated

• Semantic Summary Stuff

• Number of Discourse Categories

** Semantic Visualizations

• Alluvial Flow diagram

** Semantic Network

• Plot Semantic Network
  -- Colour by Discourse Categories(?)
  --

should be a method in the socnet, simnet, and docnet objects as all three have similarity analyses.

Sasha also want to be able to produce a histogram of all similarity scores calculated. This could be an argument for the method.

• Look into fixing layout (don't have to scroll left so much)
• Remove color from nodes
• Make nodes rectangular
• Filter for lower case and non-ascii characters
• Output to pause-able movie format
• Interactive, slider-based browser format (D3, perhaps?)
• HEATMAPS

A method that returns basic summary information about a Nate network object.

To include in output:

• number of nodes

• number of edges

• number of isolates

• number of subcomponents

• number of subcomponents > 1

• average size of subcomponents

• average size of subcomponents that are not the giant component

• modularity score

• number of communities

• directed / undirected

• n-mode

• if directed: average in degree, out degree

• if undirected: average degree

• list of node attributes

• list of edge attributes

Currently, all of the 'columns' passed into a nate object are stored in self.data, which can be very memory-inefficient.

If we keep the special columns (text, time, ID) separate, and package all of the non-special columns into the self.data object, we'll get the best of both worlds (easy compartmentalization in namespaces, memory efficiency)

Not critical unless we encounter memory bottlenecks.

Currently, they return ranges or individual records (based on an slice or integer subscript, respectively). It's more useful to return a random sample.

John said:

"It would be amazing to have, for example, a tidy dataframe with a datetime index for doing some simple time series stuff.
2:11 PM
Or the time stamps in a column that we can make a datetime object."

Quick visualization of degree distribution

Ur bad at code get better

histogram of edgewise shared partners

See: https://github.com/malcolmvr/backbone_network

Currently, it's necessary to run nate.preprocess to get the spaCy data necessary to instantiate a nate pipeline.

First, pipelines should elegantly check to see if the necessary preprocessing has been completed. This should be simple and is a logical endpoint.

For further user friendliness, though, it would be prudent to enable each of the pipeline-returning functions to also run preprocessing using defaults that will configure the preprocess function to meet their requirements.

Low priority.

Alluvial flow diagram for temporal networks

When working with end-point classes in the pipeline (e.g. svo_bursts), users are often called upon to submit a filtering term (either an SVO or a token pair) for visualization steps.

Currently, if users submit an SVO or token pair that isn't present in the svo_bursts class, an uninformative KeyError will result. Consider replacing these with informative errors.

Histogram of MGD

Right now, if a user runs a nate pipeline from command line, they won't know what's going on - there's no outward indication of what nate is working on, or even if it is still working.

We should include a judicious number of informative print statements designed to let users know what nate is up to.

edge width by weight OR similarity score

with community detection