this is for holding javascript data
Dylan Freedman edited chordprogcomp2.tex
about 9 years ago
Commit id: 5cd95119bd17827ed4933c46219cdbb53e0b4929
deletions | additions
diff --git a/chordprogcomp2.tex b/chordprogcomp2.tex
index 9a917ef..a4ffda9 100644
--- a/chordprogcomp2.tex
+++ b/chordprogcomp2.tex
...
Smith-Waterman is useful in the context of comparing chord progressions as it has mechanisms to deal well with inexact data, using different gap costs and chord distance substitution functions that compensate for small errors. Both the Harte distance metric and TPS can be easily used as a substitution function.
There are downsides
in to the Smith-Waterman algorithm. It can only be used to extract one optimal local alignment, and the score returned reflects only that local alignment, whereas n-grams comparison does well with multiple like regions of local similarity. There are adjustments to the algorithm to return multiple good alignments, but they prove computationally expensive. This paper focuses on only returning one good alignment. Another difficulty is that of comparing scores from different Smith-Waterman results on different data. Normalization measures will be discussed later in the paper CITE.
% % C, F, C, *, G, F, C, G, Dm, C
% % | | ins | | | del | |