Replication Data for: Comparing Speech-to-Text Algorithms for Transcribing Voice Data from Surveys

This Dataverse contains files required to replicate the results, figures and plots displayed in the research paper. Figures: The jupyter file "infographs-paper.ipynb" can load any wav-format file to create figures like the first three in the paper (Figure 1-3). The first visualizes the waveform of an audiofile. Figure 2 displays the waveform and time ranges (frames). The third figure creates a Mel Spectrogram. Plot: The barplot (Figure 4) displaying the Word-Error-Rates by ASR System as well as the average Word-Error-Rates in section "Results" is created by "code-main.R" file. The corresponding dataset in csv-format used here is called "transcript-data.tab". Sample Description: The sample is described in section "Data" of the research paper with additional information (age, educational level and difficulty of survey). The corresponding mean and standard deviation values are calculated in "sample-description.R" by loading the csv-formatted dataset "sample-description-data.tab". Code for Automated Speech Recognition: The original Python (and R) code used for applying various ASR systems to our data can be found as python notebooks.