xel86 / tripletlearning

Creative Inquiry - using Triplet Constrained Representation Learning to group a corpus of text into latent groups

Triplet Constrained Representation Learning

Triplet Constrained Representation Learning is a method of organizing a corpus of text into latent semantic groups based on relavancy to a predetermined notion. This project is being lead through a creative inquiry at Clemson University, and the proper explantaion of this method is explained in full by the project lead, in this PDF

Roadmap

• Test our method on MNIST
  • Create triplet function based on Stochastic Triplet Embedding [✔]
  • Implement loss function based on triplet function in pytorch [ ]
  • Simulate researcher feedback with metacriteria for triplet labels [✔]
  • Compare our methods accuracy with others to determine usefulness [ ]
• Use refined method with textual dataset
  • Group based on latent semantic findings in the text
  • Researcher feedback loop with metacriteria to group uncertain groups
  • ?
• TBA