Skip to content

Latest commit

 

History

History
43 lines (32 loc) · 3.74 KB

README.md

File metadata and controls

43 lines (32 loc) · 3.74 KB
Raw

ABX Discrimination Task

This repository contains the following files, needed for the ABX discrimination task experiment on French and English vowels:

  • scripts folder containing the scripts, needed to generate the tables, to cut the sound files, resample them etc.;
  • outputs folder containing all the tables, .csvs or other text files, generated by the scripts;
  • stimuli folder, containing the recorded .wav files and the annotated corresponding .TextGrids. It includes the intervals folder, where all the segmented intervals are stored, and the triplets folder, where all the sound files used for the experiment are (concatenated three intervals: A, B and X sound);
  • lmeds_material folder, where all the necessary files for the online experiment are stored;
  • analysis folder containing the anonymized data, the scripts necessary for the analysis as well as the outputs of those scripts.

Steps to take:

  • Have sound .wav files with their corresponding TextGrids and the silence file ready in ./stimuli folder

  • make cut_intervals: cuts the .wav files into intervals (only the stimuli)

  • make create_experimental_list: creates the list that has balanced quantity of each possible vowel that will be tested, as well as balanced appearance of each possible speaker and the context C_C

  • make create_stim_filename_distance_lists: adds the interval filenames to the experimental lists, creates the Stimulus_list.txt needed to concatenate intervals and creates a distance_list.csv that has different column names (needed this format for the analysis)

  • make concatenate_intervals: concatenates the intervals with the silence .wav file into a concatenated triplet with a new name and creates also the .mp3 and .ogg versions needed for the LMEDS

  • make make_lmeds_sequence: creates the part of the LMEDS sequence where the triplet filenames are inserted as stimuli to be tested

  • copy .wav, .mp3, .ogg files into audio_and_video which is in lmeds_material/sounds

  • copy the contents of lmeds_material folder to the LMEDS folder on the server, fix the sequences (.txt files)

  • run the test

  • make calculate_distances: calculates the euclidean distances with the DTW algorithm and appends them to the distance_list_final.csv

  • make anonymize_lmeds_data: anonymizes the names of the participants and creates two new subfolders with a language name where all data is anonymized

  • make split_output_to_results_files: maps the results from the LMEDS to comprehensible tables (sequence, presurvey, postsurvey, postsurvey2)

  • make analysis_step1_preprocess_strut: the data from sequence, postsurveys and the presurvey is combined and ordered into a new csv, ready for the analysis.

Important

  • LMEDS needs to be downloaded separately
  • audio files need to be copied over into the sounds/audio_and_video
  • LMEDS sequence needs to be manually fixed
  • .cgi and folders lmeds/test_name/individual_sequences and lmeds/test_name/\outputs need to have permissions to be modified by the test takers (chmod 777 $filename$)
  • results need to be stored in separate subfolders named by language to allow the analysis processing

Supervised model part:

  • more info inside the folder ./model/supervised/
  • need kaldi (http://kaldi-asr.org/) and abkhazia (https://github.com/bootphon/abkhazia/blob/master/abkhazia/)
  • same stimuli were used as for the human experience: for technical purposes, they were tested as separate files (one interval per file + 500ms of silence before and after), as triplets (same as people heard them) and within the carrier sentences as they were originally recorded
  • the results of the supervised model contain only the stimuli, never the added silence or the carrier sentence.