L2-Prosody-Analysis

**Miroo Lee (mil136@pitt.edu) 12-13-2021 **

1.Overview

This is Miroo Lee's project repo for Data Science (LING 2340). The goal of this project is to investigate how L2 learners' speech develop rhythmic properties of prosody by examining temporal modifications of phonetic segments as a function of lexical stress and domain-initial boundary lengthening.

2.Dataset

The data set I started my project with comes from the PELIC speech corpus from the University of Pittsburgh. The PELIC speech corpus is a large learner corpus, and the current project examined 2-minute semi-spontaneous monologues by Korean students. You can find more information about the corpus from here.

3.Contents

In addition to this README file, there are four folders and eleven other files.

In the root folder:

• final_report.md describes the results of the data analysis.
• README.md is the current document you are reading.
• LICENSE describes the licensing term for the project.
• .gitignore has git ignored file entries.
• project_plan.md describes the initial plan for the project.
• project_progress.md shows three progress reports throughout the semester.
• presentation.pdf is the slides of the presentation I gave at the end of the semester. This presentation only included the preliminary data analysis. More detailed results are documented in final_report.md
• search_wav.Rmd contains codes for identifying wav file names by filtering L1, level, and task type.
• search_wav.md same as the above but in md file.
• KOR_mono.csv is an output of search_wav.md. It is a list of two-minute monologue speech files of Korean speakers who were enrolled for three semesters.
• KOR_mono_scripts.csv is another output of search_wav.md. It is a list of transcripts for the corresponding speech files.
• export_from_three_tires.praat is a Praat script that compiles annotated information from multiple praat textgrids to a single txt file.
• wordList.csv contains a list of words found in wav_SAMPLES. The list also contains syllable structure and lexical stress information of each word.
• new_wordList.csv contains a list of words found in three wav files from the speaker ea4.
• data_analysis.Rmd contains codes for data cleaning & analysis.
• data_analysis.md same as the abobe but in md file.
• plots has plots from data_analysis.Rmd.
• scratchpad has codes I tried and documented for my project.
• wav has 129 wav files identified on KOR_mono.csv.
• wav_SAMPLES has subset of wav files which are annotated in textgrid files from wav.