Modeling language change
In Fall 2015 I wrote a Java program using the MASON toolkit to simulate five different language change models. The main idea is to take theoretical explanations of language change and build them into an agent-based simulation, with agents playing the role of individual speakers.
In this project it's all about stress patterns in noun/verb pairs, like address (n) versus address (v). Over time, which syllable is stressed has changed. In a paper from 2010 mathematical models were used to reproduce different observed changes. I implemented those models in an agent-based approach, and added two features the mathematical models couldn't represent:
- Analogical change - N/V pairs with same prefix tend to have the same stress pattern
- Dialectical divergence - N/V pair trajectories can diverge between distant groups of speakers
The group of speakers in the blue box has 0.56 probability of saying address as a noun, while the group in the red box has an average of 0.75.