From: Andrew N. <and...@vi...> - 2008-04-15 13:09:29
|
Mark - I would love to include the author browse tool in vufind. Feel free to try to munge it in with the code in SVN. Andrew ________________________________________ From: Mark Triggs [mt...@nl...] Sent: Tuesday, April 15, 2008 8:40 AM To: Andrew Nagy Cc: James Farrugia; vuf...@li... Subject: Re: [VuFind-Tech] MarcImporter indexing update Hi Andrew, Andrew Nagy <and...@vi...> writes: > Hi Mark - and thanks for jumping in here - I know you guy s have played > around with this quite a bit. > > I just committed a revised query generator that uses both and'd and or'd > searches together and weights the AND higher than the OR. This way we > get both results. In my testing I think it works fairly well. Ah, that's a clever idea. We don't explicitly boost AND matches, but some of the other things we do (like boosting phrases with varying levels of slop) sort of have a similar effect. I wonder if we'd see improvements by doing the same thing... > Also as an aside - I have editted the synomyms.txt file that controls > synonyms and cleaned it up as well as added in some default synonyms > that I think will help. The new synonyms are for numbers and roman > numbers. I've created the following: > > II,ii,2,two,Two Cool. We haven't done roman numerals but we probably should. So far we've got a huge file of things like: 36th,thirty sixth 37th,thirty seventh 38th,thirty eighth 39th,thirty ninth and various acronyms for the Australian states and things like that. > So know if you do a search for "world war 2" or "world war ii" the > results will be the same. You could do this with your other examples > such as: stephen,steven Yeah, I know many of the examples I've given can be worked around with synonyms, but I don't think they'll catch everything. Searches like 'Where the wild things are' vs 'Where the wild things were' are the ones I'm really worried about--where the user has misremembered one of the words. > Keep your eye out for 0.8.1 today! Brilliant! Once it's ready to go I'll get started on merging our search code in and let you know when I'm got something to show. This is really well-timed as we've just released our next round of features and there's not too much other stuff on my VuFind plate. Our most recent notable thing is probably browse indexes (which our staff and certain users wouldn't let us neglect): http://protocat.nla.gov.au/Browse/Home?browse=author&from=knuth http://protocat.nla.gov.au/Browse/Home?browse=subject&from=art If anyone's interested in this sort of thing feel free to drop me a line. We've implemented a new Solr request handler to support this, but it's not too hard to get going. Cheers, Mark -- Mark Triggs <mt...@nl...> |