Transcribing audio data: overview and transcripts of several automatic transcription tools

Throughout institutions, audio recordings are being made regularly. To be able to further process these recordings, the audio often needs to be transcribed. In order to avoid having to transcribe the audio manually, there is a wealth of tools available for doing so automatically. In this record, we present an overview of several often-used tools to automatically transcribe pre-recorded audio data, including their features, costs, and security. To check the quality of the tool, we also recorded an audio fragment in Dutch that we ran through all tools in this overview in March of 2022. This original audio fragment (Test_interview_20220203.mp3), the cleaned-up transcription (Test_interview_cleaned_transcript.odt) and each tool’s raw transcript of the audio fragment (Test_interview_[name-tool]_raw_[date-run]) are included in this record as well. The raw transcripts were downloaded as .docx or .txt files and the .docx files saved as .odt. No edits to the transcripts were made before saving them, except an incidental removal of a personal email address or hyperlink. The overview contains information and transcripts of following transcription tools: Amberscript HappyScribe Kaldi NVIVO transcription Sonix SpokenOnline Transcribe Trint Microsoft Word 365 Online <strong>About</strong> This overview was created through a collaboration between Utrecht University’s Research Data Management (RDM) Support and the DataHub SSH programme situated at the faculty of Humanities. The details in the overview have last been updated April 19, 2022. Please note that at the time you are downloading these files, the quality of the (Dutch) speech-to-text conversion may have been improved by the respective supplier.当前各机构均定期开展音频录制工作。为满足后续处理需求，通常需对音频内容进行转写。为规避人工转写的繁重工作，目前已有大量自动化音频转写工具可供选用。 本数据集旨在梳理多款主流预录制音频自动化转写工具，涵盖其功能特性、收费标准与安全机制。为评估各工具的转写质量，我们于2022年3月录制了一段荷兰语音频片段，并使用本次梳理的全部工具开展转写测试。 本数据集同时收录该原始音频片段（Test_interview_20220203.mp3）、经过规范化整理的标准转写文本（Test_interview_cleaned_transcript.odt），以及各工具针对该片段生成的原始转写结果（命名格式为Test_interview_[工具名称]_raw_[测试日期]）。 原始转写结果均以.docx或.txt格式下载，其中.docx文件已转换为.odt格式。保存前未对转写内容进行额外编辑，仅零星移除了个别个人邮箱地址或超链接。 本次梳理涵盖以下转写工具的相关信息与转写结果：Amberscript、HappyScribe、Kaldi、NVIVO transcription、Sonix、SpokenOnline Transcribe、Trint、Microsoft Word 365 Online <strong>关于本数据集</strong> 本梳理工作由乌得勒支大学研究数据管理支持团队（Research Data Management, RDM）与人文学院数据中心SSH项目（DataHub SSH）合作完成。本次梳理的最新更新时间为2022年4月19日。请注意：当您下载本数据集文件时，各供应商对荷兰语语音转文本技术的优化或已完成，转写质量或有所提升。