spyysalo / annalign Goto Github PK

Add arg to reject alignments with large (relative) Levenshtein distance

The diff algorithm can map non-space to space characters, leading to aligned annotations having initial or terminal space even if the original annotation did not. The mapping should be fine-tuned to avoid introducing extra space.

See the example in README.md for one case.

When a textbound annotation refers to text that gets entirely deleted in alignment, the annotation gets deleted. This can cause dangling references from other annotations (e.g. attributes) that point to the deleted annotation. These should be also be deleted.