5 September 2017

Graph-Based Taxonomy Finder

There are many ways of building out a semantic taxonomy of terms. One approach is using a graph representation with a machine learning approach. The below algorithm highlights a high-level approach to using a Word2Vec and Minimum Spanning Tree for building out a weighted graph for a taxonomy.

  • Use Gensim to process data model which provides for generalizable context for terms
  • Create distance file for terms using Gensim
  • Use output file to feed into minimum spanning tree by edge relationships between vertex terms
  • Visualize the weighted output graph in Gephi