[itsdb] Q: Analysis limit and result storage limit
Stephan Oepen
oe at csli.Stanford.EDU
Wed Mar 14 19:30:53 CET 2007
hi berthold,
> I was wondering what is the difference between these two tsdb++ options?
> How do they affect parsing with Pet? Does setting of analysis limit
> activate selective unpacking, if cheap is called with packing enabled?
generally, both parameters are passed to the processing client: `Result
Storage Limit' just controls how many results are to be sent back to
[incr tsdb()] and should not impact the parsing procedure itself. but,
as you suspect, `Analysis Limit' is meant to turn on best-first search
(until that many analyses are found or the search is exhausted), though
only if `Exhaustive Search' is turned off at the same time.
in PET, when run with packing, best-first search amounts to selective
unpacking. there used to be a non-packing best-first mode, where the
agenda was ordered by MaxEnt scores of partial constituents. i cannot
quite say whether this mode is still functional.
most likely one will benefit most from settings like the following:
(setf *tsdb-exhaustive-p* nil)
(setf *tsdb-maximal-number-of-analyses* 1000)
(setf *tsdb-maximal-number-of-results*
*tsdb-maximal-number-of-analyses*)
this is, naturally, assuming the latest PET from SVN and a MaxEnt model
provided as part of the grammar.
all best - oe
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+++ Universitetet i Oslo (IFI); Boks 1080 Blindern; 0316 Oslo; (+47) 2284 0125
+++ CSLI Stanford; Ventura Hall; Stanford, CA 94305; (+1 650) 723 0515
+++ --- oe at csli.stanford.edu; oe at ifi.uio.no; oepen at idi.ntnu.no ---
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
More information about the itsdb
mailing list