Thanks for the offer of help, I'll keep it in mind. I mainly ran into problems with the tests... ----8<---- baron@daedalus:~/tmp/link-4.1b-ruby/tests$ ruby tc_linkparser.rb Loaded suite tc_linkparser Started Opening ./4.0.dict F Opening ./4.0.dict E Finished in 0.010642 seconds. 1) Failure: test_basic(TestLinkParser) [tc_linkparser.rb:30]: Exception raised: Class: <StandardError> Message: <"Could not find dictionary."> ---Backtrace--- tc_linkparser.rb:16:in `initialize' tc_linkparser.rb:16:in `new' tc_linkparser.rb:16:in `initialize' tc_linkparser.rb:10:in `new' tc_linkparser.rb:10:in `new' /usr/lib/ruby/1.8/singleton.rb:95:in `instance' /usr/lib/ruby/1.8/singleton.rb:84:in `instance' tc_linkparser.rb:20:in `dict' tc_linkparser.rb:31:in `test_basic' tc_linkparser.rb:30:in `assert_nothing_raised' tc_linkparser.rb:30:in `test_basic' --------------- 2) Error: test_link_each(TestLinkParser): StandardError: Could not find dictionary. tc_linkparser.rb:16:in `initialize' tc_linkparser.rb:16:in `new' tc_linkparser.rb:16:in `initialize' tc_linkparser.rb:10:in `new' tc_linkparser.rb:10:in `new' /usr/lib/ruby/1.8/singleton.rb:95:in `instance' /usr/lib/ruby/1.8/singleton.rb:84:in `instance' tc_linkparser.rb:20:in `dict' tc_linkparser.rb:65:in `test_link_each' 2 tests, 1 assertions, 1 failures, 1 errors baron@daedalus:~/tmp/link-4.1b-ruby/tests$ ls . 4.0.affix 4.0.constituent-knowledge 4.0.knowledge tiny.dict .. 4.0.batch 4.0.dict tc_linkparser.rb words baron@daedalus:~/tmp/link-4.1b-ruby/tests$ ----8<---- I have two guesses as to what the problem is: - I copied liblink.a to the wrong directory(ies), or - 4.0.dict needs to be in the same directory as the library. I haven't investigated further on it due to unrelated personal reasons, though. The performance issues themselves are not with Lingua::LinkParser, but with Link Grammar itself. When running it on the sentences in http://notpublic.wrong.button.com/sent.txt . Those have been extracted by a friend from a NYT article. Some time out, others just fail, and on one Link Grammar just refuses to work because the sentence is longer than 70 words! Fortunately for me, the first phase of my work does not require _all_ sentences to be parsed - just enough to extract a sufficient amount of VS/VO pairs (~40000). Lastly: I've been in contact with Martin Chase, who worked on the Linguistics and LinkParser ports (http://raa.ruby-lang.org/list.rhtml?name=linkparser and http://www.deveiate.org/code/linguistics-overview.html - a Ruby gem is also available for the latter). I plan to use Linguistics because I also need to make use of WordNet. So far LinkParser is unoptimized (to the point that pruning (and probably memoization too) are a necessity), and needs the words in the sentence to be tagged if used with 4.0.dict . The plan I have so far is to use your Link Grammar wrapper instead of LinkParser for the Linguistics module, since it is considerably faster. Cheers! -CWS On Tue, 1 Feb 2005 20:17:26 +0900, Kaspar Schiess <eule / space.ch> wrote: > (In response to news:bb13341905013110287d1ef43c / mail.gmail.com by Claus > Spitzer) > > > I'll give it > > another try once I get access to a clean machine. > > Hey, just tell me if you need help on this. You are the only user as far as > I can tell (that is: until I figure out something usefule for this tool). > > Performance issues ? Isn't the thing rather well optimized ? I wonder what > you are running trough Lingua::LinkParser to get these issues... > > kaspar > > hand manufactured code - www.tua.ch/ruby > >