Detour: Graph Analysis [4]
Background
In previous post, I have covered my approach to create a sentence-based graph for extracting relationship.
In this post, I am going to change some of the construction steps to see if I can get better result.
Approach
I applied following schema to update node and edge weights to skew the random walk.
I am switching gear to create a see if certain tokens are clustered in way that is consistent with our understanding.
Then I remove any tokens that are not ‘verb’ or ‘noun’.
From the output, you will notice the community output matches our understanding of the sentence.
Intermission
This post I further analyse the properties of the graph generated from the sentence. There is way to isolate communities of the graph that match the partitioning of the sentence of how human beings would perceive the sentence.
In next post, I will focus on analysis on the original graph.
Stay tuned!