Hi folks, One of the areas we continue ping search vendors on is how they allow a user to tell the engine what subject domain they are talking about. Some...
Many of the major search engine portals have endorsed the new XML sitemaps standard. It’s great to have them agree on something. But for those of you used...
Setting aside the issue of whether or not it's a good idea to get all results for a query (and the fact that there are other ways of getting these such as the...
This is all from memory, but maybe it will help. 1. seek.allhits is not a supported API call. If it works differently in the next release and breaks your code,...
Walter: Thanks for the reply. I'm aware that getting all hits is not recommended for the reasons you mention (lack of relevance order, performance, etc.). ...
Ask support for the source to pyqueryall.spy and promise to rename it if you change it. Since the pyqueryall script is an API, Ultraseek needs to control how...
Are there any industry benchmarks, frameworks or methodologies for comparing quality of search engines? Does any reference data on relevancy of leading...
Hi, As far as I know there is no industry standard comparison of the search appliances. However there are big differences between the enterprise search...
... SharePoint ... FYI: The Mysteries of SharePoint Search There's a myth in town. SharePoint searches documents. While the search feature of the SharePoint...
TREC is an annual search competition, but they restrict what the winners can say about it, and most established vendors don't bother with it anymore. Most...
There are some good practices in this area, but the applications are too diverse to have common benchmarks. For relevance, my favorite metric is Mean...
... I doubt that anyone does this because it is very hard and not very useful. So, you will need to let go of that and try an approach that actually works. ...
Well isn't that the whole idea behind clustering? Northern Light did this years ago -- in fact I think their folks were the first I ever heard utter the term...
Walter offers, as we'd expect, words of wisdom. Let me add that if you're benchmarking competitors, be sure your management understands all of the parts of...
... If one or both of the contexts are specific domains, then that is jargon and not ambiguous. Truly ambiguous phrases are rare, and just are not a big...
I agree that click-throughs can certainly add context, a good validation of that at NF. We're not directly shopping for a solution, but I'm thinking about it...
Jean: I'm currently evaluating new vendors and "tuning" our existing search implementation relevancy ranking. In addition to Walter's MRR measure, I'm using...
Ed, Thanks for your response. I got quite a few responses to this question. Jean ... From: ed.dale@... To: search_dev@yahoogroups.com Sent: Tuesday, May 22,...
I would tune ranking and precision before recall. If you have all relevant results but the first one is on the second page, you will have VERY unhappy users. ...
This should be on topic. I'm hiring for positions in Denver. I would consider remote working if you are qualified. Travel may be involved. Full posting is...
Occasional job postings related to search engineering positions, cralwers, search consulting, etc. are "on topic" and are OK in moderation. However, general...
Thanks for setting those guidelines, Mark. They make all the difference in ensuring high signal-to-noise. _____ From: Mark [mailto:mbennett@...] Sent:...
chris.biow@...
May 24, 2007 3:26 pm
318
On a Linux disto, say RedHat, what is the offocial K2 startup sequence in the init.d chain? This would be to start the various K2 processes and Dashboard, etc....
I always just make a simple one which calls <verity install dir>/.../bin/k2adminstart for startup, and .../k2adminstop for shutdown. Randall ... From:...
Walter: Thanks for the detail. I am fairly new to evaluating search, and may have gone off course. I did not get into the detail of our measures, since Jean ...
Thanks Peter, I'm curious, where to you typically put the sym links (what kernel phase, etc) for Solaris and Redhat? The numbered rc#.d directories, etc. Mark...
Hi Mark, Generally, the run-level will be either 2 or 3. The startup can be fairly late in the piece, so something like /etc/rc2.d/S99verityk2 should be OK....
Good stuff Peter. I think I was headed towards "what is considered the 'best practice'", which I think has been covered to some extent. I saw some warnings...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
