[developers] PET with inflectional and lexical rules

Yi Zhang yzhang at coli.uni-sb.de
Wed Apr 5 13:11:24 CEST 2006


Hi All,

I was playing with the latest version of PET(v0.99.11) and ERG (jan-06)
earlier today and noticed something not to be expected. Here is an
example session (notice that (4) failed to parse).

------------------------------------------------------------------------------

[yzhang at takeshi erg]$ cheap -packing=7 -verbose=0 /opt/delphin/erg/english.grm
reading `/opt/delphin/erg/pet/english.set'... including `/opt/delphin/erg/pet/common.set'... `/opt/delphin/erg/pet/global.set'...
loading `/opt/delphin/erg/english.grm' (LinGO (Jan-06)) reading ME model `/opt/delphin/erg/rondane.mem'... [11846 features]
58789 types in 4.9 s
Abraham likes the cats
(1) `abraham likes the cats' [0] --- 1 (0.01|0.01s) <26:45> (771.9K) [0.0s]
Abraham likes the cats.
(2) `abraham likes the cats.' [0] --- 1 (0.02|0.02s) <26:45> (773.2K) [0.0s]
Abraham likes the cat
(3) `abraham likes the cat' [0] --- 1 (0.02|0.02s) <25:45> (758.1K) [0.0s]
Abraham likes the cat.
(4) `abraham likes the cat.' [0] --- 0 (-0.00|0.01s) <25:21> (687.2K) [0.0s]

------------------------------------------------------------------------------ 
I set the `punctuation-characters' as following so that the `.' does not
get removed by the tokenizer: 
punctuation-characters := "<>{}+*`".

All these inputs worked fine with LKB. And the problem remains similar
whether I use the line-oriented input or the yy input mode.

I flopped the vanilla ERG.jan-06 from
http://lingo.stanford.edu/ftp/redwoods/6th/erg.jan-06.tgz 
and commented out the
`sm := "rondane.mem".' in pet/common.set to avoid some warnings.
Is there anything else I should set up for `cheap'?

Or is it likely to be a bug in PET?

Best,
Yi





More information about the developers mailing list