Search Changes

Posted on Friday, February 26th, 2010 by lukecrouch
Category: General, Status Updates

Until this week, our search system has been powered by Syracuse, a custom Lucene search server we wrote four years ago. Although the code is technically open source, we never developed a community for it, so the project has idled quietly within our software stack. We’ve received many user requests to “fix” the SourceForge.net search system, from stale index data to general relevance scoring, so today we’re starting to address the issues.

We’re moving to Apache Solr, which has a robust feature set and active community. Solr will give us the ability to add new features like spell-checking, highlighting, and a clean faceted searching experience. We’ll also fix stale index data, incorporate new data like ratings, tags, and icons, and drop some of our unreliable data like the activity rating.

We’re beginning the update by migrating our Project search feature to Solr. We’re combining our Software Map and Search areas into a single Search interface that incorporates faceted browsing by category to recreate the Software Map experience. In addition to browsing by topic, we can now browse projects by development status, license, and programming language.

We want to hear what you think of the new search features - send us support tickets, leave comments here, or just catch us on IRC. Let us know what we’ve overlooked or what we should add or change to improve the search experience. We will enhance the initial release based on your feedback. When we’re reasonably happy with Project search, we will update Tracker, Forum, and Mailing List searching as well.



Reader Comments

armanfe on February 26th, 2010

Dear executive,

I am not sure how related my request is, but I would very much appreciate your insight.
We are planning to use Gnaural to test Auditory beat frequency in Parkinson (PD) patients, something that was mentioned in the original
Oster Sci Am paper. We will start by confirming the original
observation by checking the threshold for beat preception in PD
patients and controls. After that we will repeat the study together
with EEG to see if the brain wave measures follow the perceptions.

We were wondering if you know if anyone else has used Gnaural for studies like
this? Would you have any suggestions for how to automote a series of
different frequency combinations. Ideally it would be nice to also
automate the response collection — eg space bars if subject hears the
beats and any other key if not.

Thanks

Arman Fesharaki, MD
SUNY Downstate hospital
Brooklyn, NY

faustie on February 26th, 2010

“When we’re reasonably happy with Project search, we will update Tracker, Forum, and Mailing List searching as well.”

Forum? I thought you guys killed the forum.

qubodup on February 26th, 2010

Hello, I am a foss blogger (specifically games) and sf.net’s browse function has been very resourceful to me to provide interesting and fresh news in the past.

I am glad that the “sort by last release” functionality is restored!

1. I would like it if I could enable displaying screenshots in the browse view. They on one hand often indicate a usable state of the software and on the other hand give it a visual ‘recognition’ effect.

2. What I *really* would love to see is a “browser by last repository-commit/push” sorting functionality. Many projects are not following the RERO philosophy and don’t get noticed by the project browser.

Thanks for your efforts!

qubodup on February 26th, 2010

PS: I added my suggestions to the ideatorrent.

leeschlesinger on February 27th, 2010

Armanfe, please leave your comments in the Gnaural project’s own forum (https://sourceforge.net/projects/gnaural/forums/forum/498481). It’s highly unlikely those folks will find them here, as a comment to an unrelated issue in the community blog.

dickanderson on February 27th, 2010

The search seems to be full of spam now!

It maybe unrelated to the new engine, but the project I’m working on used to appear at the top of a search for “insurance”. Now it appears at ~50 after a load of spam messages about cheap insurance.

Can’t say I’m impressed.

lukecrouch on February 27th, 2010

Hmm … the new search relevance is likely just demonstrating an existing unmanaged area of spam. For now you can sort results by downloads, release date, or rating to push spam projects to the bottom (they rarely have any).

We’ll consider incorporating some more data into the general relevance score as well.

ddietz on March 1st, 2010

Search produces a lot of results which result in “Error 404″ due to projects which have been deleted…

ddietz on March 1st, 2010

p.s. e.g. try searching for “Java Content Management System”… Maybe these search results can be filtered…?

andreamedeghini on March 1st, 2010

I like the new search system, but I think that relevance is not a meaningful metric, if it does not take into account other factors such as downloads, activities, etc.

peterdietz on March 1st, 2010

Very good news, thank you!
When faceted browse expands to viewing mailing-lists, this could almost be a way of self-documenting. Sometimes RTFM questions get asked over and over again, and being able to see the entire list-archive browseable by Feature_A -> Subfeature_SA, just by extracting keywords from old posts is great news.

This might turn mailing lists with a decent amount of traffic into a self documenting archive. Perhaps allowing project admins the ability to make a bit of a faceted browse site-tree would be additionally helpful.

petas874 on March 3rd, 2010

The search seems to be a bit more narrow now, which is good. When I searched for my own project (which I guess is the first thing that everybody does, hehe) by name I used to get hundreds of other answers before it, even though they didn’t match the name at all.

Now I’m only getting the right thing. Good work!

But I do think that the layout/design of the new search is a bit “less nice looking” than the old one. It reminds me of these spam/placeholder search sites you sometimes get to when mispelling the address of a site. It could benefit from some small UI tweaking!

Keep up the good work! Cheers

gushie- on March 5th, 2010

Lots of abandoned projects now appearing near the top of the default search. If a project has nothing to download and has had no activity for years could it be made less “relevant”?

pizza2004 on March 10th, 2010

I personally don’t enjoy this new search feature, and wish I could still use the old one. I particularly like how every project had a ranking before, out of every project on the website, it really gave me something to aim for as far as improving goes.



Both comments and pings are currently closed.



Status Terms Privacy Advertise About HTML5 Center SourceForge.JP


You are viewing a mobilized version of this site...
View original page here

Mobilized by Mowser Mowser