https://github.com/sannetenoever/STiMCON
Raw File
Tip revision: 873a2bf5c79fe2f828e72e14ef74db409d387854 authored by Sanne ten Oever on 23 July 2021, 08:31 UTC
Update README.md
Tip revision: 873a2bf
README.md
# STiMCON 
This repository describes the scripts from the paper <i>Ten Oever & Martin (2021), An oscillating computational model can
track pseudo-rhythmic speech by using linguistic predictions, Elife, 10:e68066. DOI: https://doi.org/10.7554/eLife.68066 </i> 

## Explanation
The repository consists of scripts belonging to the Corpus Gesproken Nederlands (CGN), simulations and fitting with STiMCON

### CGN related files:
<b>CGN_Fig2.py:</b>\
This script extracts and plot the basic temporal variation in the syllables and words of the CGN related to Figure 2 of the main manuscript.

<b>CGN_Tab1_Fig3_Fig7.py:</b>\
The ordinary least square and related figures.

<b>RNN_Model.py:</b>\
The RNN model

<b>RNN_subFun.py:</b>\
Subfunctions to use the RNN_Model

### STiMCON related files:
<b>STiMCON_Fig4_Fig5_Fig8A.py</b>\
Shows the basic behavior of STiMCON (Figure 4), the threshold/timing of activation (Figure 5) and ambiguous daga overall simulations (Figure 8A)

<b>STiMCON_Fig6.py</b>\
Shows how acoustic time and model time is not the same in STiMCON (Figure 6)

<b>STiMCON_Fig8C.py</b>\
Fitting of the da/ga data using the first active node as output (Figure 8C)

<b>STiMCON_Fig8D.py</b>\
Fitting of the da/ga data using the relative node activation as output (Figure 8D)

<b>STiMCON_core.py</b>\
Core script for the STiMCON model which has all the low-level code

<b>STiMCON_plot.py</b>\
Plotting output of the STiMCON

<b>STiMCON_sen.py</b>\
Creating sensory input going into the STiMCON
back to top