In this project, I developed an interactive Shiny application, MaxEnt with Hidden Structure in R, designed to assist linguists in generating phonological grammars (weights) using a Maximum Entropy model. The app was built based on the existing HGR model developed by Staubs (2014). HGR finds solutions for learning problems (with or without hidden structure), generate distributions over forms, and performs online learning simulations.
You can access the Shiny app here.
input
: Input formsoutput
: Output formsprobability
: Observed probabilities or raw frequenciesThe Shiny app includes three sample input files, which you can download from here. These files demonstrate different scenarios for using the MaxEnt model, including cases with or without hidden structures and cases of having output candidates with raw frequencies instead of normalized probabilities. These sample files can be uploaded directly into the Shiny app to see how the app handles different types of data.
This research was supported by the National Science Foundation grant BCS-2140826 to the University of Massachusetts Amherst.