Closed LinguList closed 6 years ago
The pattern should be stored for each alignment in the column "pattern" or "correspondences" in the first entry in the list (following the ID of the word). This will allow for sparse data, and in formats like CLDF, it would be placed in another table.
the pattern browser, which I just added, should be sufficient, we don't even need to edit patterns. It is enough to just display them for the time being.
Patterns, i.e., correspondence patterns, can either be inferred automatically from alignments, or assigned as such by the users. I imagine a simple straightforward system in which the user adds or corrects a given pattern attached to an alignment. At the same time, a pattern browser would allow users to search through all patterns which are assigned the same ID for a given alignment site.
so, two alignments (cogid 1 and 2) like the following ones:
will then be rendered ihn the patterns as:
This means, a new panel has to be added, namely the pattern-panel, which also allows users to edit patterns directly (re-assigning the cognate-ID to another pattern). The pattern is similar to an active reconstruction, but it maintains the awareness of the user that there may be conflicts, while the automatic pattern-detection algorithm could be used to make a preliminary parse.