Derived Text Formats

The complete corpus is downloadable in aggregated representations suitable for quantitative analyses. All data is made available for non-commercial scientific research only.

Bag-of-Words

Tab-separated frequency lists giving the yearly count of every word form. Available as token counts and as lemma/part-of-speech counts, each for the full corpus and for two prominent archives separately:

Token counts by year (whole corpus) ↓
Lemma/POS counts by year (whole corpus) ↓

Token counts by year (charts archive only) ↓
Lemma/POS counts by year (charts archive only) ↓

Token counts by year (hiphop archive only) ↓
Lemma/POS counts by year (hiphop archive only) ↓

Word N-Grams

Token bigrams and trigrams for the full corpus, with association scores and context measures; see Amin/Fankhauser/Kupietz/Schneider (2021). Suitable for lexicographical research, collocation analysis, and distributional semantics studies.

Token bi- and trigrams ↓

Word Vectors

GloVe (Global Vectors for Word Representation) computations delivered as a CSV dataframe. Trained on the full Songkorpus, suitable e.g. for word similarity tasks.

Token vectors ↓

Please cite the corpus when using it, see Citation and Chronicle.