Lyrics

[MarkGotham]

Details to follow.

[MarkGotham]

The lyric files (lc*.txt) provide an initial resource for cross-referencing and text-based studies.

They should be robust for all simple cases with no ossia parts or repeats: i.e. if the song is 'through composed' with all the lyrics encoded below the vocal line, and no repeats for multiple verses with the same music (more details follow below).

Known issues and omissions with lyric files:

1. Some songs (following their IMSLP sources) encode subsequent verses (second, third ...) in separate text boxes. These are not included in the separate lyric files.
2. Multiple verses below the same vocal stave: only the first verse is suppled in the lyric file. Usually this is indicated by a 1. at the start of the lyric file (i.e. 1. means there is at least a 2. somewhere in the score). Again, this doesn't affect through-composed songs: where they begin with 1., they still have do have a 2. where the second verse begins: 1. at the start of the text file and no 2. indicates a missing second verse.
3. While most of the syllables are successfully joined back together (e.g. hel- -lo in the score returns to hello), there are still some occasional issues with this.
4. Related to 2. and 3. more complex repeats may produce fragments like the first word of the second verse only (in a first time repeat bar).
5. Two cases of lyrics on the piano part that have resisted extraction so far:

• Reichardt,_Louise/Sechs_Lieder_von_Novalis,_Op.4/3a_Geistliches_Lied/lc5092560,
• Reichardt,_Louise/Zwölf_Deutsche_und_Italiänische_Romantische_Gesänge/02_Wenn_ich_ihn_nur_habe/lc5100067

Other potential improvements also to consider

1. A more structured representation (e.g. more json/yaml files), perhaps including the part name as given ('voice', 'gesang', 'stimme', etc.)
2. link up each score entry to a lieder.net page for cross-referencing versions.

[MarkGotham]

Returning to this.

1. Wrt Lieder.net see Metadata and database #3 .
2. Given the existence of most song texts elsewhere, perhaps prefer here to include the syllabic division (which is the more unique information offered here) or both.
3. Consider manual edits. In which case, lc<num>.txt as the best we have, perhaps with lc<num>_raw.txt where different
4. If expanding to wider languages (e.g., Slavic) consider including a transliteration here, perhaps using this package

[MarkGotham]

Considerably improved in recent commits. More anon.