Some corpora have multiple data types for a single document, such as the score and an audio recording. If these follow a naming convention like song_title.mscx and song_title.wav for scores and audio and lie in the same (sub)directory, they can easily be matched. In that case, the FileCorpus class should provide a way to iterate over tuples of files with the files function and potentially support reading in these files with different file_reader function, return a tuple of processed data by the data function.
Some corpora have multiple data types for a single document, such as the score and an audio recording. If these follow a naming convention like
song_title.mscxandsong_title.wavfor scores and audio and lie in the same (sub)directory, they can easily be matched. In that case, theFileCorpusclass should provide a way to iterate over tuples of files with thefilesfunction and potentially support reading in these files with differentfile_readerfunction, return a tuple of processed data by thedatafunction.