Raise-your-Voice

Repository to Svenja Guhr's Dissertation

See also the corpus publication of "theme-d-prose 1848-1920" on Zenodo, DOI: 10.5281/zenodo.12666499.

This repository contains:

Corpus and annotated corpus texts:

• Plain text corpus as TXT files (UTF-8)
• Plain text corpus as XML files
• Manually annotated training data as XML files with sound word annotations
• Manually annotated training data as XML files with sound event annotations
• Manually annotated test set with 10 XML files with sound event annotations
• Manually sound word-annotated XML file of M. v. Ebner-Eschenbach's Die Resel
• Manually sound event-annotated XML file of M. v. Ebner-Eschenbach's Die Resel
• XML files with automatically sound-annotated corpus texts with sound event annotation
• XML files automatically sound event-annotated and semi-automatically enriched with loudness levels

Models:

• Trained NEISS NTEE Models for automated sound word and sound event annotation: The models can be regenerated using the indicated training data [https://github.com/SvenjaGuhr/ link]. The models will be uploaded on Zenodo after the publication of my dissertation. For more information see NEISS NTEE's wiki (TEI Entity Enricher) [https://github.com/NEISSproject/tei_entity_enricher].
• Model trained on 20 manually sound word-annotated corpus texts
• Model trained on 55 manually sound event-annotated corpus texts

Jupyter Notebooks:

Preprocessing of the texts

• Preprocessing of plain text files as preparation for the manual sound word annotation in CSV Link
• Preprocessing of plain text files as preparation for the automated sound annotation in XML Link

Preprocessing of the corpus metadata

• Preprocessing of the corpus for generating metadata (i.a. token and word count) Link
• Metadata generation: Corpus Keyword Search and Extraction Link

Evaluation of inter-annotator agreement

• Evaluation of inter-annotator agreement with Krippendorff's alpha Krippendorff 2018 and gamma by Mathet et al. (2015) using the python package pygamma-agreement Link

Postprocessing of the annotated corpus texts

• One script providing the entire process from automated annotation revision, annotation extraction to automated loudness level labeling via the dictionary approach Link
• Postprocessing the sound annotations for extracting the annotations from XML to CSV Link
• Postprocessing the sound annotations through automated revision of the XML annotations Link
• Loudness annotation with dictionary approach Link

Building the Word2Vec Model

• Word2Vec_model.bin
• Jupyter Notebook for generating and using the Word2Vec model Link

Dataframes:

• Corpus metadata table as CSV

Annotation Guidelines:

• Guidelines for sound word and sound event annotation Link

License:

• Information on the GPL license