propagation of feature name+debug from factory, return correct features array for fsa ffs

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@486 ec762483-ff6d-05da-a07a-a48fb63a330f

1 files changed, 9 insertions, 0 deletions

diff --git a/graehl/NOTES.beam b/graehl/NOTES.beam
index 59314439..a48d1ab7 100755
--- a/graehl/NOTES.beam
+++ b/graehl/NOTES.beam
@@ -18,3 +18,12 @@ randlm worth using?  guess not.
 actually get all 0-state models in 1st pass parse and prune passive edges per span.
 
 allocate cube pruning budget per prev pass
+
+(has been tried in ISI decoder) models with nonstandard state comparison,
+typically (partially) greedy forest scoring, some part of the state is excluded
+from equality/hashing.  Within virtual ff interface, would just add equals, hash
+to vtable (rather than the faster raw state compare).  If this is done often,
+then add a nonvirtual flag to interface instead, saying whether to use the
+virt. methods or not.  or: simple flag by user of ApplyModels (per feature?)
+saying whether to be 100% greedy or 0% - no halfway e.g. item name uses bigram
+context, but score using 5gram state.