Overview
This research line develops computational approaches that explain how individual adaptation behavior can scale to community-level language change. The objective is to build interpretable models that connect exposure, cognition, and social structure to observed shifts in phonetic and phonological patterns.
Active Workstreams
- Simulation of phonetic drift under different exposure and network conditions.
- Predictive modeling of adaptation magnitude from behavioral and acoustic features.
- Tooling that supports annotation, quality control, and reproducible analytics across projects.
Methods and Toolchain
- Statistical learning models and agent-based simulations.
- Feature engineering from acoustic and experimental datasets.
- Model comparison and cross-validation for explanatory and predictive goals.
- Reproducible workflows for data processing and reporting.
Recent Software Releases
- eeg-pipeline - EEG processing tools used in lab analyses.
- phonJSD - Software for working with variable phonetic data and related workflows.
Integration Across Lab Clusters
- Uses community and cognition datasets to test model generalizability.
- Provides quantitative support for hypotheses generated by field and lab studies.
- Feeds model outputs back into experimental design decisions.
Project Snapshot
- Cluster: Computation
- Inputs: behavioral, acoustic, EEG, and sociolinguistic features
- Outputs: simulation results, predictive models, reproducible analysis tools
- Current software portfolio: 2 public repositories
Current Data (2024-2026)
- Software: Active lab tooling includes eeg-pipeline and phonJSD.
- Recent methodological output: 2023 publication on analytic flexibility in speech analyses (Coretta et al.; Berry co-author).
- Computational conference output on record: 2022 COLING presentation and manuscript on verbosity-controlled translation workflows.
Last Updated
February 2026