The Spoken Wikipedia Corpora

The Spoken Wikipedia project unites volunteer readers of Wikipedia articles.
Hundreds of spoken articles in multiple languages are available to users who are – for
one reason or another – unable or unwilling to consume the written version of the article.
We turn this speech resource into a time-aligned corpus, making it accessible
for research and to foster new ways of interacting with the material.

• hundreds of hours of aligned audio
• from a diverse set of readers
• about a diverse set of topics
• in a well-researched textual genre
• licensed under a free license (CC BY-SA 3.0)

NEWS

• Please participate in our rating experiment on speech quality (if you are a German speaker)!
• Have a look at the amazing Wikipedia-Reader which allows to hyperlisten to the Spoken Wikipedia (and is based on the SWC); also have a look at the paper which proves the amazingness :-).
• The SWC has its own ISLRN.

Publications

• Our original publication at LREC 2016.
• Publication about our WikipediaReader that is based on the corpus published at SLPAT 2016.
• Interspeech paper that ranks speakers in the SWC by likability based on crowdsourced preference ratings!

Current Statistics

Download

Original Release (Spring 2016):

• Aligned text (XML format):
  • German (17MB)
  • English (19MB)
• corresponding audio:
  • German (11GB)
  • English (12GB)
• LICENSE: CC BY-SA-3.0. Each folder in the audio contains a info.json with the original URL of both audio and text, and the speaker's name.
• If you use this data, please cite our paper: bibtex

@InProceedings{KHN16.518,
  author = {Arne K{\"o}hn and Florian Stegen and Timo Baumann},
  title = {Mining the Spoken Wikipedia for Speech Data and Beyond},
  booktitle = {Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016)},
  year = {2016},
  month = {may},
  date = {23-28},
  location = {Portorož, Slovenia},
  editor = {Nicoletta Calzolari (Conference Chair) and Khalid Choukri and Thierry Declerck and Marko Grobelnik and Bente Maegaard and Joseph Mariani and Asuncion Moreno and Jan Odijk and Stelios Piperidis},
  publisher = {European Language Resources Association (ELRA)},
  address = {Paris, France},
  isbn = {978-2-9517408-9-1},
  islrn = {684-927-624-257-3/},
  language = {english}
 }

Sample Data

• English: Martin Luther Alignment, text, audio (alignment as JSON, meta-information as JSON)
• German:
  • automatic alignment: Super Mario Bros Alignment, WikiText, audio
  • manually corrected alignment: Photodiode, Isar

Software

• Our master script will download all required software (including the ones listed below) and do all the alignment work for you
• Wikipedia download and handling: https://bitbucket.org/natsuhh/swc
• alignment: https://bitbucket.org/natsuhh/prosub/ (uses MaryTTS for normalizing and Sphinx-4 for alignment)
• evaluation: to be published

WorkInProgress