This repository contains the data, model code, and processing tools for the term paper. Paper online: https://bjarke.me/2019-cogsys-paper

After cloning this repository, install the dependencies:

pipenv install

Run the four pre-experiments:

pipenv run python tools/3-experiment/1_pre_experiment.py
pipenv run python tools/3-experiment/2_pre_experiment.py
pipenv run python tools/3-experiment/3_pre_experiment.py
pipenv run python tools/3-experiment/4_pre_experiment.py

Analyze the results:

pipenv run python tools/3-experiment/5_main_experiment.py

The output can be found in the data/3-experiment/ folder.

This repository is organized according to the following file structure:

• data/ — all data related to the research
  • 1-visual/ — visual stimuli data
    • generated/
      • graph/ — some statistics for the generated data
      • pattern/ — visualization of the generated data
    • prototypes.csv — the prototypes, not included in training set
    • stimuli.csv — the visual input stimuli
  • 2-acoustic/ — acoustic stimuli data
    • original/
      • 1-wordlist/ — complete and manually augmented word list
      • 2-audio/ — audio recordings of narrated words
    • processed/
      • audio/ — automatically cut audio samples after running tools
      • graph/ — some statistics for the cut audio samples after running tools
      • wordlist/ — selected word list with 100 word senses (categories)
    • stimuli.csv — the acoustic input stimuli
  • 3-experiment/
    • graph/ — analyzed results and statistics for the experiments
    • log/ — log protocols after running the experiments
• lib/ — implementation code in Python
  • data/ — code related to data processing and generation
  • model/ — model implementation
• paper/ — code for the paper (LaTeX)
  • img/ — images used in the paper
• tools/ — tools to automatically process data and run experiments
  • 1-visual/ — tools related to visual stimuli
    • generate_patterns.py — re-generate the visual dot patterns
  • 2-acoustic/ — tools related to acoustic stimuli
    • 1-wordlist/ — tools related to word lists
      • process_wordlist.py — re-generate the selected word list
    • 2-audio/ — tools related to audio recordings
      • process_full_recordings.py — (currently defunct) cut full recordings
      • process_audio.py — process audio files to generate stimuli data
  • 3-experiment/ — tools related to running experiments
    • 1_pre_experiment.py — run 1st pre-experiment
    • 2_pre_experiment.py — run 2nd pre-experiment
    • 3_pre_experiment.py — run 3rd pre-experiment
    • 4_pre_experiment.py — run 4th pre-experiment
    • 5_main_experiment.py — (currently defunct) run main experiment
    • 6_analyze.py — analyze the results of all experiments
    • 7_prepare_images.sh — copy images used in the paper
• Pipfile — track dependencies to install
• testall.py — run all unit tests