Hi, IMO, XWiki's search feature is severely affected by the fact that, as a structured wiki, it does *not* take advantage at all of the fact that the domain is known (and already mapped in lucene) and, instead, it asks users for a plain keyword.
If you look at Google, for example, keyword search is perfectly valid, since the domain is unknown (web pages, free text, etc), but for XWiki, it is an absolute *must* to provide an advanced search (faceted search) feature: - result type - language - author - date - wiki - object class - attachment - etc. Only then, IMO, we can really discuss about how we can improve on the results and tweak the scoring. Just take a look at the search suggest feature. I, for one, always use search suggest because it is the closest thing to a "focused" search that reflects my search intent. Until then, we can only hack solutions that will never be good enough for a user that wants to find something in XWiki. A magic solution that guesses what the user is looking for just by entering a single keyword does not seem to be achievable in the near future (at least not without a brain wave reader :) ). On the implementation side, it's actually just a matter of building a proper UI that generates the lucene query. Not sure if this helped very much for the topic at hand, but maybe it will inspire someone reading it :) Thanks, Eduard On Fri, Jun 29, 2012 at 3:50 AM, Sergiu Dumitriu <[email protected]> wrote: > On 06/28/2012 04:01 AM, savitha sundaramurthy wrote: > >> Hi all, >> >> While trying to retrieve the search results had the following doubts. >> >> *Problem: * >> * >> * >> * *Say XWiki has three languages English, Spanish, French. I >> give a query in English(some *proper noun*) , >> should it return the documents pertaining only to english or it can >> return documents pertaining to other languages too? >> >> If the scenario is such that it retrieves the documents irrespective of >> languages, I have few ideas to deal with it. >> >> 1) We can get the documents, merge them, add the scores and give it a high >> rating. This would help to avoid super >> results such as a display of a different match in each language to some >> extent. >> 2) Make it a part of facet search , where search results could be >> differentiated base don language. >> >> Would be really helpful to gain your suggestions. >> > > XWiki is pretty unique in the way it handles multilingualism, so I can't > think of an example to follow. > > Also, how a multilingual XWiki is going to be used depends a lot on the > particular organization using it, so one generic solution might not make > everybody happy, so multiple solutions to chose from (in the > administration) might be the proper way to go. > > Here's how I would like things: > > When searching for something, let's say "scorpions", and my current > language is English, I see first documents that are written in English: > > " > Search results for "Scorpions": > > [100%] Scorpion > [ 95%] The Scorpions > [ 50%] Scorpio > " > > After that, we also search for a few top hits in all the other languages > except English, and if we have strong hits (let's say score above 75%), we > display something like: > > " > You might be interested in these results in other languages: > > [ 98%] [de] The Scoripions > [ 90%] [fr] Scorpiones > [ 89%] [ro] Scorpion > > [[Search for "Scorpions" in every language]] > " > > Now, I'm not sure when exactly to display this: > - every time when there are hits with a score above a threshold > - only when there are hits with scores higher than the best scoring result > in the current language > - only when there are few results in the current language (less than 5) > -- > Sergiu Dumitriu > http://purl.org/net/sergiu/ > > > > ______________________________**_________________ > devs mailing list > [email protected] > http://lists.xwiki.org/**mailman/listinfo/devs<http://lists.xwiki.org/mailman/listinfo/devs> > _______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs
