new search engine

[slug]

hi.
the new search engine is online at http://search.pbase.com/search

it is still very much in development, but it should be an improvement over the old search engine.

the most obvious and probably useful change will be that galleries are now indexed. not just images.
usernames, and fullnames are indexed to point to your root gallery.

unobvious but important changes to the internal workings.
the search engine is now running on its own separate database so there should be no performance issues from it affecting the rest of the site.

while it won't be updating for the next two weeks, after that, the new design should provide for fast updates to the index, and faster queries.

you won't see any changes to it for the next two weeks, but the next steps are to improve the user interface which is now very crude, and then there will be an ongoing project to improve the results.
the current ordering of images and galleries is very simplistic and does not factor in various data that could help. it's not too bad though.

as with most common areas of PBase, the images and galleries displayed are those in public, non-passworded galleries from paid accounts.

-slug

[ilanphoto]

Just had a quick look at this new addition to the Pbase family

Way to go - I am impressed

No more limit on 500 images and now we get to find galleries
Love it, now I will have what to do at work for the next few weeks

Oh, please don't tell my boss

Ilan

[lwh]

I like it. Thanks

[lord_of_the_badgers]

Now that's MUCH better. Thanks Slug!

[arjunrc]

thanks for the feature - much awaited. Some nits:

1. it does not do a proper substring search. For example, searching for 'beach' does not match word 'beaches'. This would be very useful to add (does match 'beautiful beach' if you search for 'beach' but thats because they are word delimited)

2. 3 keyword search only - ok, you will fix it as per your message

3. Looks like it assumes that all words must be present. In other words
"arjun roychowdhury panorama" assumes I want all of them present. This is a little odd from an interface perspective, where you would expect a "+" for explicit combination. A very small nit. I know you will be sprucing up the interface

4. gallery numbering/navigation is messed up for now. Very easy to bring it to show bogus stuff like "showing page 13 of 1" etc. if


5. Not sure I really understand the search logic. In my case, most of my images have a keyword 'arjunrc'.
Now, if I do a search on 'arjunrc' I get a lot of images. expected.
Now if I do a search on 'arjun roychowdhury arjunrc' - I get only the root page. Why ?
Similarly, if I search on 'arjunrc maliko' I get two pictures but if I search on 'arjun roychowdhury maliko' I get nothing. Not sure of the logic again. Would be great if you could explain it.

6. Not sure if I can ensure that the image I search for will be limited only if owned by me. For example, if I search for 'arjun roychowdhury xyz' I get the feeling that even if someone elses picture has the same name as mine, it will show up. In other words, is it possible for you to put in a search mechanism to ensure that all searches happen withing a 'username' or 'full name' if required ?

thanks
arjun

[den123]

Just had a quick look at this new search engine. I like it. Very good idea with separated pictures and galleries.

Thanks

[gmcconn535]

The new search engine is fantastic!! The only real glitch that I have noticed is that if there are more galleries than images found on the search, you cannot get to the galleries beyond the number of images (i.e, if there are 48 images and 65 galleries, you can't get to the galleries beyond 48.

[matiasasun]

The new search engine is fantastic!! The only real glitch that I have noticed is that if there are more galleries than images found on the search, you cannot get to the galleries beyond the number of images (i.e, if there are 48 images and 65 galleries, you can't get to the galleries beyond 48.

... Maybe that´s possible because a lot of images are named 1234_5678_9.jpg but most galleries have names.

[gtepke]

Thanks for the work on the new search engine. However, it does seem to have one of the same bugs as the old search -- if you include a non-alphanumeric character like an apostrophe or hyphen in your search, nothing is found, even though many photos and galleries use these characters. For example, search for "Ashy Storm-petrel" (without the quotation marks) and then for "Ashy Storm". Since the names of many bird species use these characters, the search tool is still almost useless for the bird photogs on Pbase (there are quite a few of us). Hopefully this is one of the problems that you will be ironing out. Thanks. Glen

[matiasasun]

Arjunc`s post and the last one deservs an answer. I think it is a very usefull way to have more access to a lot of new galleries. I`m Happy.

But I do notice that the most common words in internet searchs does not appear. Some time ago I find an image post by Erich M. named "nude". He did comment me that single image had 15.000 visits. That`s a lot. Even for a gif animated of a girl moving his head.

That`s why my question is:
Does the users of the PBase system does not search using those words? or is it that they were removed from that list?

Thanks very much
Matias

[gleemonex]

i really wish there were a way to sort the output by images added most recently.

example: searching for toronto brings up 4581 images, many of which, i've seen before. it'd be nice if there were a way to show the most recently added images as an option. as it is now, there's not even a way to skip to the end -- you have to sit there hitting "next 12".

[castledude]

i really wish there were a way to sort the output by images added most recently.

I agree, actually I wish the form for pictures was similar to the form for seaching the forum.

[snip]
as it is now, there's not even a way to skip to the end -- you have to sit there hitting "next 12".

Look at the Address after you do the seach and hit next 12 the first time.

it will be something like this

http://search.pbase.com/search?q=las%20vegas&begin=12

if you change the number after the word begin it skips directly to that item.

[olliemaitland]

Small suggestion: Could you put the post vars in the search box so it has what you just searched for in the search field?

[slug]

ok. the form shows the current query now.
will continue to work on the other issues.

been working on the indexing the last couple of days.
the index should now keep itself updated within a few minutes, but search results are cached until they are at least 24 hours old.
so any changes you make should show up on the search page in at most 24 hours.

-slug

[kstuebin]

I don't think this is how the ranking should be done. Titles often have nothing to do with the photo. For example I could post a rose and call it "Summer Beauty." If I want it to show up in the search rankings I'd have to call it "Red Rose" Then I'd have to have "Yellow Rose," "Pink Rose," etc. Pretty boring names.

For the heck of it, I did a search of my state, West Virginia. Every single gallery that appeared had West Virginia in the title. I have lots of galleries of West Virginia that do NOT have West Virginia in the title and they don't show up. I do have it in the keywords, caption and location. But still no sign of them. Some of my galleries do appear on the third search page but they have West Virginia in the title.

If you refine the search further, say "West Virgina Mountains," you end up with strange results. A bunch of flowers. Only two galleries of actual mountain scenes. Because mountains is in the caption or titles of these flower galleries. I then tried "West Virginia scenic" because that is a keyword I use. Now, all the results are of my galleries and photos.

Just some feedback on how it's working. If you want to get a high ranking them put whatever it is in the title. So look out for "West Virginia Misty Morning," and "West Virginia Shadowy Pond." Just kidding. :) I'm not going to do that.