Code for Räsänen & Kocharov (2024). "Age-Dependent Analysis and Stochastic Generation of Child-Directed Speech", Proc. CogSci-2024, Rotterdam, Netherlands.

Program code for training a GPT-2 architecture language model (LM) from CHILDES transcripts and generating new child-directed speech transcripts using the model.

Contents:

• GILES_main.py: The main script for training the model and generating text using it.
• measure_unique_utterances.py: Script for analyzing the proportion of unique utterances in the generated texts that never occur in CHILDES (or counting unique utterances in CHILDES).
• analyze_results.m: Plotting the results based on the corpus analyses (MATLAB).
• get_childes_naenglish.R: The R script to download NA English corpora.
• childes_to_ao_dataset.py: Extraction of CHILDES transcript data from downloaded CSV-files into age-specific bins.

Main dependencies

Python

For LM training:

• tensorflow==2.12.1
• tensorflow-text==2.12.0

For evaluation of the generated data:

• stanza (https://stanfordnlp.github.io/stanza/)
• evaluate (https://huggingface.co/docs/evaluate/en/installation)
• matplotlib
• pandas
• scipy

R

• childesr

Instructions:

• Download CHILDES transcripts from NA English corpora using get_childes_naenglish.R as CSV-files (see the list of corpora in the script). The script uses childesr library (https://github.com/langcog/childesr). Setup a path to directory, where you want to store CSV-files.
• Extract CHILDES transcript data from downloaded CSV-files into age-specific bins using childes_to_ao_dataset.py

childes_to_ao_dataset.py --input-dir INPUT_DIR --output-dir OUTPUT_DIR

• Run GILES_main.py to train the model and generate transcripts with it (after setting data paths inside the file).
• Run compare_datasets.py to extract linguistic features for the given texts (or collections of texts):

compare_datasets.py --input-dir INPUT_DIR --output-dir OUTPUT_DIR --sample-data --analyze-data

The same code give a possibility to check statistical significance of feature changes along with (1) an age of a target child and (2) a number of words in a collection of texts corresponding to an age bin.

compare_datasets.py --input-dir INPUT_DIR --output-dir OUTPUT_DIR --sample-data --analyze-data --statistical_tests --plot_figures

• Run analyze_results.m to plot the feature comparison of the generated transcripts against original transcripts.