Skip to content

Latest commit

 

History

History
44 lines (31 loc) · 1.62 KB

README.md

File metadata and controls

44 lines (31 loc) · 1.62 KB

AutoSense

AutoSense Model for Word Sense Induction

This code was used in the experiments of the research paper

Reinald Kim Amplayo, Seung-won Hwang, and Min Song. AutoSense Model for Word Sense Induction. AAAI, 2019.

The src/models folder contains one Java file containing the GAS class. The GAS (Granularity-Agnostic Sense Model) refers to the AutoSense model. To use the model, create an object of GAS using the following line:

GAS gas = new GAS(data, target, numSenses, numTopics, alpha, beta, gamma);

where

  • data: is a list of data instances
  • target: is the target word
  • numSenses: is the number of senses hyperparameter
  • numTopics: is the number of topics hyperparameter
  • alpha: is the Dirichlet prior of the topic distribution (set to 0.1 in the paper)
  • beta: is the Dirichlet prior of the sense distribution (set to 0.01 in the paper)
  • gamma: is the Dirichlet prior of the switch distribution (set to 0.3 in the paper)

Then, you would need to run the Gibbs sampler using the following lines of code:

gas.initialize();
gas.estimate(numIters);

where numIters is the number of iterations (set to 2000 in the paper).

To print the results, use the line:

gas.printSemEval(filename, target);

To cite the paper/code, please use this BibTex:

@inproceedings{amplayo2019granularity,
	Author = {Reinald Kim Amplayo and Seung-won Hwang and Min Song},
	Booktitle = {AAAI},
	Location = {Honolulu, HI},
	Year = {2019},
	Title = {AutoSense Model for Word Sense Induction},
}

If you have questions, send me an email: reinald.kim at ed dot ac dot uk