Webware

Google's Gmail Labs has just rolled out a useful, but mildly creepy feature that gives you suggestions on who you should e-mail based on previous conversations. So, if you've had threads going with a group of people, it will recommend some of those folks once you've added at least two addresses in the recipients field. Best part is, they don't even have to be in a group of contacts you've created in Gmail's contacts manager.

I got this to work to an almost uncanny level. It handles things like family members without a hitch, but where it gets useful is in pulling up people you've recently been corresponding with and giving you a one-click link that adds them to the message. Short of organizing these people to a group (in which you may need to add or delete someone on a per-message basis), or doing a reply-all on an old message, this is a really fast way to get another thread started with the same group.

Now all Gmail needs is a quick way to take these contacts and add them to a sending group without leaving the message, something that can only be done in Gmail's contact manager.

Gmail can now figure out if you're trying to start up another conversation thread with people you've recently been talking to, saving you some time on starting a new message thread, or simply using a reply all on an old one.

(Credit: Google)

There aren't many new companies launching at this year's Web 2.0 Expo in San Francisco, which runs Tuesday through Friday. One of the few that are is Lunch.com, which strives to get a little more juice out of user-generated publishing.

Here's the premise of Lunch: You can review anything you want, from a TV show to a restaurant to a food product to a household appliance. I guess it aims to be, sort of, a Wikipedia for opinions. Founder J.R. Johnson, who started building the site after he sold previous creations VirtualTourist.com and OneTime.com to Expedia, said that Lunch started filling up its private beta by reaching out to frequent Amazon reviewers and received a very positive response.

You're also encouraged to network with other members and filter reviews through its "Similarity Network" function, an algorithm for finding like-minded users and matching them to one another. To ramp up Lunch's assessment of your preferences, you can play "speed-rating" games called Exhilarate, which are structured much like Netflix's recommendation feature.

Quite honestly, I have a hard time seeing people turn to a general reviews site when there are already well-established sites for reviews of businesses, books, movies, and the lot--not to mention a plethora of "social shopping" sites for consumer products. I feel like Lunch could've gained a lot more traction if it had made its debut two or three years ago, when user-generated content was a lot more noteworthy. But maybe that's just me.

I like b-sides is a new service from the maker of the now-defunct DiggSuggest. The idea is simple: you go to the Web site and upload the metadata from your iTunes library (in an XML file), and the site tells you songs that you might like, based on the bands it knows you do.

I tried this out with a small library of about 4GB and it gave me a good sampling of tracks. All of them come from Amazon's MP3 store, and can be previewed in a small player that sits on the side. There are also links that will take you to each song's download page, giving i like b-sides affiliate credit in the process.

The tool's secret sauce is in a small feature off to the side that calls out "cool tracks you have and are probably ignoring." These are tracks deemed popular that simply have low play counts. This same section also tells you how much of your library has independent artists, and how many songs you've bothered to rate. I'd love to see both of these go a little deeper.

While there's no built-in social element with friends or following, you can share the tool's recommendations with others. Each library gets a permalink to do this, and you or your friends can come back and "shuffle" to bring in new results (you can try this here). The bigger library you have, the more diverse of a selection you'll get, so revisiting the site as you acquire new music will get you even more recommendations.

Related: Mufin lets you discover new music with science

Once you've uploaded your iTunes library information, the tool will sort through it and give you recommendations of less popular tracks it thinks you might dig.

(Credit: CNET Networks)

I stumbled upon a useful site earlier today that's worth sharing. Called SimilarWeb, this small Firefox (and soon Internet Explorer) add-on sits on the side of your browser and pulls up sites that are similar to the one you're currently on.

It works remarkably well--at least with major sites. For example, visiting YouTube brings up a long list of other video hosts. The same went for social news sites like Digg, Reddit, and Delicious. You can scroll through these and open them up in new tabs, or pick from one of the tags SimilarWeb believes to be related to that page. This will pull up an entirely new list of places it thinks you should visit.

What makes the service shine is that users can re-arrange the lists and submit new sites that are not yet in SimilarWeb's index. There are thumbs up and down buttons which can raise or lower a site's standing on the list. Down-voting any site will actually remove it from the list. As a result, if users continue to vote the list gets more accurate.

SimilarWeb tells you sites that are like the one you're on.

(Credit: CNET Networks)

There is a notable downside to using this extension: your browsing performance will take a hit. For some reason it needs to load its own results before it loads the actual page, which in my case meant waiting an extra few seconds when visiting a new site. That might be a deal killer for some, although it can easily be avoided by learning the keyboard shortcut that dismisses it from running in the sidebar. You can also pull up the results from a drop down menu next to your browser's address bar.

SimilarWeb would make a good companion for the now toolbar-free StumbleUpon, which actually learns from your browsing habits to give you pages it thinks you'll like. Combined with this you'd get another avenue of exploration.

Here's an explanatory video from the site's about page. Worth noting is that you don't need to have your volume on.

Music search and recommendation tool Mufin is opening to everyone this morning. The service, which launched in private beta in early October, lets you find music that's similar to a track you know based on a scientific analysis of its composition.

New on Thursday is both a Facebook app and the previously mentioned iTunes plug-in that scans your library to give you recommendations. Unlike Apple's "Genius" analyzer system in iTunes, Mufin actually scans your tracks for relational relevance instead of giving you an aggregate hodge-podge of recommendations based on the playlists and purchases of iTunes users. The only catch here is that you're limited to Mufin's relatively small 4 million-song database, which is roughly half that of iTunes.

Users are also getting the option to save playlists and notes--the service's equivalent to a shopping list. Previously these would disappear between sessions, which kept it from doubling as an ad-hoc music streaming tool.

As for the Facebook application, it's little more than a widget that lets you search for tracks without leaving the social network. It does however give you a "discovery wall," which lets you share and view tracks bookmarked by friends--similar to the MySpace version that was available back at launch. If you're looking for something a little more anonymous, the Mufin team is now providing weekly recommendations for music to look out for, although you'll have to purchase them off-site.

Previously: Mufin lets you discover new music with science

Note: The iTunes plug-in is currently Windows-only. You can download it here (.EXE warning). Here's what it looks like:

The new iTunes plug-in scans your library and offers up song recommendations based on musical similarities. (click to enlarge)

(Credit: Mufin )

Recommendation tools are a dime a dozen these days. That's not necessarily a bad thing though--the more exposure you get to new content, the greater your chances of finding a new favorite. To aid in that search is newcomer Taste Kid, a search tool that provides simple recommendations for music, books, TV shows, and movies.

The tool uses previous user searches to figure out how terms are related, which means as time goes by the results are honed, and hopefully more accurate. While each of these results is missing links to purchasing pages and audio samples, each one has a YouTube video that you can view right from the results.

What makes the site particularly useful are the unrelated items that get stuck on the bottom of each exploration page. The site tacks on the most recently added and popular bands, books, TV shows, and movies, which turns each result into its own browsing experience. In just a few minutes you can peruse about a dozen videos to find something you like--and if not, a simple click on something you do like takes you to another recommendation page to start the process over again.

Where the tool loses some of its luster is the lack of a breadcrumb trail to get you back to your original search, or any sense that it's getting to know you better. Mufin, a music recommendation tool I took a look at a few weeks back does this, and it makes it a far more engaging experience. Ideally future iterations of the tool will keep track of this with a cookie.

[via TheNextWeb]

Start a deep dive into music, movies and TV show recommendations with Taste Kid, a simple recommendation tool.

(Credit: CNET Networks)

Launching in private beta on Wednesday is Mufin, a Berlin-based music discovery service that helps users find similar tracks based on identifiers within a song. It scans each track for its density, tempo, and rhythm, then draws comparisons with other songs that match. What you get is a system that lets you find similar tracks by sound, even if they cross multiple genres.

Last week I chatted with Petar Djekic, the service's marketing manager, about this process and its origins. What's interesting is that it's a spinoff from the technology that was created to identify songs from short clips--the same thing that's found in Midomi and Shazam. The difference here is that songs must first be worked through a cataloging process that takes just a few seconds to scan a 10 to 20-track album.

The information gets fed into a massive search engine which assigns a percentage of similarity from track to track. It can also do the same for albums, so if you're in love with the specific work from an artist you can see other entire albums you might like.

You might have wondered what other songs are similar to Rick Astley's 'Never Gonna Give You Up.' Mufin knows.

(Credit: CNET Networks)

You might be asking yourself how this could be at all useful in an age where two of the highest profile MP3 players (iPod and Zune) now have built-in recommendation tools both on the player hardware and computer software. Djekic's answer to this is that both of these solutions, as well as those found on Last.fm and Amazon.com, are all based on user behavior. You can get similar tracks, but not without some effort either on your end or that of other users. Djekic says anyone can simply come to Mufin with their existing library and get a list of related tracks without having to make friends or provide ratings.

In addition to its search tool, Mufin provides a playlist creator for tracks that have streaming rights. You can't share these playlists with anyone else (yet), but you can save them for later, as well as pop them out in their own window to play while you're off doing other things. If you're more keen to own the tracks the service is integrated with iTunes, and soon Amazon's MP3 store, so you can buy what you like.

Mufin is a really cool and handy service. There are already 3.8 million tracks in the catalog, which is growing rapidly. It's currently in private beta, although we've got 100 invites to give out. To get yours, visit the service through this link.

Note: Last.fm is owned by CNET parent company CBS Interactive.

Note 2: Information regarding album processing times has been amended. It was originally stated that it took 2-3 days to process an album, whereas that's how long it takes to be fed into Mufin's search index.

On Tuesday night StumbleUpon is changing the way users interact with the service, ditching the need for a software-based browser toolbar in place of a small frame that loads on top of the Web site you're on. Users with the toolbar installed will still be getting the same experience, but the idea is that anyone can begin stumbling without having to install anything.

To get the Web toolbar to show up in the first place, users must now begin their stumbling experience from the StumbleUpon home page. The site is now broken up into categories. Once you've clicked on a link the experience begins, with the persistent toolbar following you from site to site and keeping track of your ratings to provide you with new stumbles.

Earlier this week, StumbleUpon founder Garrett Camp told me this was an idea that had been kicked around the office for years--six in fact, and the only reason it hadn't happened sooner is that Camp and others felt it would diminish the number of people who were populating the service with rated content. That number is still staggering, with more than 35,000 new URLs submitted every day by 6 million registered users. Camp hopes this new install and registration-free solution will make those numbers even larger, and improve some of the uptake as people get to try the service without that first hurdle.

In addition to its exploratory angle, StumbleUpon is introducing a new partner program. Sites that have StumbleUpon installed will be able to offer their users a new "Stumble This" button with a counter on it. When a user clicks this it adds to the number, which can help promote it for other StumbleUpon members. It's also got an option right underneath the counter that lets users jump to another piece of related content, something Camp says should drive traffic to other existing posts. It's worth noting this is different from the previously existing StumbleThru feature, which would do this randomly.

StumbleUpon's new home page will serve as a starting point to various bits of media, and exploring it no longer requires a software toolbar. (Click to enlarge.)

(Credit: StumbleUpon)

The partner program is launching on four sites Tuesday night, including political blogging network The Huffington Post, HowStuffWorks, Rolling Stone online, and National Geographic. Of the four, Rolling Stone and National Geographic are the most interesting, as users will be able to explore the photo archives with the service's recommendation engine. Like service Photoree, which we checked out back in August, this can be a fun and engaging experience.

Camp says there are 10 other partnerships in the works, including several for video and music content. Eventually the system will be open for anyone to place it on their blog, although Camp says the system needs to be fine-tuned before it's ready for that.

The future of StumbleUpon

When I asked Camp for comment on the rumored sale of StumbleUpon from parent company eBay, he said he "couldn't talk about any rumors." However, what's interesting is that this new system could be ported over to eBay, or any other product site, which is something many were expecting when the company was acquired last year. "This does open us up," he said. "We're a lot more media focused, and this would allow us to do product discovery."

Presumably with such a system in place you could jump around the site and discover new products while rating them at the same time--something the auction site does not currently provide. Camp says StumbleUpon might one day provide that, but for now he says that realm has already been covered pretty well by search. "(We're) more interested in doing media stuff. There's a greater need for discovery than products right now."

The new StumbleUpon.com should be available right now. Camp says user profiles, reviews, and friends lists will get updated to match the new style in the coming weeks.

Last week I got a glimpse at GoodGuide, a new product recommendation service that rates products as safe, healthy, and green based on their ingredients, manufacturing, and distribution. It's part of a new class of services that are bringing data transparency that can save money, lives, and the environment, not necessarily in that order.

BillShrink.com recently launched a recommendation service that focuses on unpacking the complexity and cost of cell phone plans. The free service analyzes wireless phone bills and rate plans from the major carriers, and then monitors usage and makes recommendations for saving money. Another company, Validas has a similar, though fee-based, service and claims to save customers more than 20 percent on average.

This week, BillShrink is adding credit cards to its service offering.

"No more smoke and mirrors. All fees, risk, and reward will be out in the open on BillShrink.com so people can find the best match for individual needs and feel secure that they're making financial decisions that will actually help build them credit," company CEO Peter Pham said in a statement.

Users enter data about their credit card habits, and BillShrink comes up with recommendations for more cost-effective cards.

BillShrink monitors plans, promotions, rewards programs, and other data points for people, making new recommendations that could save them money. Given the dismal state of the economy and the lack of transparency in many consumer services, BillShrink, Validas, and GoodGuide are tapping a real need.

Pham said that BillShrink could extend its personalized recommendation engine to other domains, such as car insurance and mortgages, where the choices are also confusing and complex, and results can change on a daily or monthly basis. Perhaps these new data transparency services will help prevent some of the obfuscation by service providers and reduce poor decision-making by consumers.