summaryrefslogtreecommitdiff
path: root/graehl
diff options
context:
space:
mode:
Diffstat (limited to 'graehl')
-rwxr-xr-xgraehl/NOTES.earley4
1 files changed, 4 insertions, 0 deletions
diff --git a/graehl/NOTES.earley b/graehl/NOTES.earley
index 4156063a..0953708c 100755
--- a/graehl/NOTES.earley
+++ b/graehl/NOTES.earley
@@ -105,3 +105,7 @@ vs.
3) * -> x . *
with 3, we predict all sorts of useless items - that won't give us our goal A and may not partcipate in any parse. this is not a good option at all.
+
+======
+
+-LM forest may have many in-edges per V. (many rules per NT lhs). so instead of generating all successors for scan/predict, i wanted to have them in sorted (admissable) -LM cost order and postpone once the prefix+rule part is more expensive than something else in the agenda. question: how many such postponed successor things