Why you should think Positively
Lots of people want to know more about project Positively, so I thought I’d follow up on Friday’s post with a few more details about what Positively’s all about.
For the uninitiated, Positively is the codename for a new sentiment tracking app that Swingly has built for the 80Legs web spidering platform. We’re still in a private beta, but we’re showing what our brand of sentiment tracking can do for interested 80Legs customers.
The concept is simple: we use the power of 80Legs to simultaneously get tons of fresh content and to run Swingly’s entire natural language processing pipeline. Swingly searches through the docs spidered by 80Legs as they’re downloaded — if we find the droids content you’re looking for, it gets extracted, saved as easy-to-use XML, and streamed to your app.
How easy is it to set up? Here’s a fancy graphic that spells out all the steps:
Step 1. Select a Target. Users start working with Positively by telling the app what they’re most interested in tracking. Want to track a specific product or brand name? Just give us its name (or what you think its name is) — and Positively will figure out all the different ways it can be referred to. Want to track a product category (something like shoes or digital cameras or luxury cars)? Positively currently knows about 50 different product categories. Want to learn everything about a specific type of thing? Positively’s up to the challenge: right now, we know about more than 2200 different types of things, from priests to coffee shops to computer magazines to relief pitchers.
Step 2. Pick some docs. Since Positively uses 80Legs to track down docs for you, it needs to know what kinds of docs it should focus on spidering. Want only what’s been said in the mainstream media? Tell Positively to stick to the news. Want to know what’s in the blogosphere? Tell Positively to check out only what bloggers are saying. You’re the boss.
Step 3. Pick some starting links. Since 80Legs is a spidering service, it needs to know where to start looking. Just a few links will do. (Don’t have a clue? Positively can help you out here, too.)
Step 4. What do you want to know? As I mentioned on Friday, one of the coolest things about Positively is that it intelligently figures out what are the most interesting attributes for each name, product, or service you’re tracking. It also accepts tips, too. If you’re interested in a certain set of features (say, the sound quality of high-end headphones), we’ll make sure we find all the ways that attribute’s expressed out there on the Web — and bring all that data back to you. Here’s where the power of Swingly really comes in: instead of just searching for your keywords, Positively tries to find all of the expressions that mean the same thing as your seed link. It’s semantic search, and its time has come.
Step 5. How much data is enough? How much data is too much data? Positively lets you control the depth and breadth 80Legs crawl — and to set an upper bound on how many docs (or sentiments) it should get.
Step 6. Output? Last but not least, Positively lets you output data in a variety of formats. We like XML, but we can envision you wanting other formats, too.
So, that’s it. Consider yourselves initiated. If you’re interested in a sneak peek, contact me at andy@swingly.com, and we’ll show you what sentiment tracking-plus-web crawling is all about.
Yes, there are lots of great companies who are heavily invested in sentiment analysis. But we’re thinking positively — and we think you should, too.
We’re still in a private beta, but we’re showing what our brand of sentiment tracking can do for interested 80Legs customers.
Kanye takes over Swingly
Yeah, I know that this meme’s a bit played out, but imagine my surprise when Swingly started doing it’s best Kanye impersonation this morning!
More than a Feeling
Really tickled with all of the much-deserved positive press that 80Legs has attracted after yesterday’s successful public launch at the DEMO09 conference. (See here, here, and here. And here. And here.) We couldn’t be happier for Shion and Brad and the rest of the 80Legs team. They’ve got a great product, and they’re well-positioned to really dominate the crawling market. (Oh, and they’re from Texas Rice, which is a good thing in my book.) Congrats, guys!
Over the next couple of weeks, I’ll be talking a lot about the different kinds of semantic apps that Swingly (and its parent company, Language Computer) have built to run on the 80Legs platform. We’re psyched about combining Swingly’s broad-coverage semantic apps with the massive amounts of data that 80Legs provides. It’s a pretty unbeatable combination: 80Legs helps you cast a broad net, while Swingly lets you know exactly what you caught.
While I don’t want to steal any of the 80Legs spotlight, I couldn’t resist telling you a little about the Swingly sentiment analysis app (code name: Positively) that Shion used during his DEMO pitch yesterday.
Like a lot of other sentiment analysis services (such as those provided by ScoutLabs, Jodange, Evri, NStein, or Crimson Hexagon — just to name a few), Positively was designed to help users discover what people think about pretty much any person, product, organization, or service imaginable.
Want to know what people think about the Neill Blomkamp flick, District 9? Lots of sentiment analysis apps can boil down an Internet’s worth of noise to a summary score like this:
and a list of comments (usually tagged as positive or negative ) like this:
- The movie looks great to begin with and this trailer re-enforces we’ll likely get a solid, if not great film out of it. [1]
- “District 9″ seems an oddly misguided sci-fi movie. [2]
- It definitely has the goods: an interesting concept, Blomkamp’s clever filmmaking (the movie begins as a faux-documentary and gradually shifts into a survival tale) and ambitions that far exceed the Hollywood norm.[3]
Positively is different from most sentiment analysis apps in two ways.
First, unlike many other services which rely on large amounts of preprocessed data, Positively runs “live” as part of an 80Legs crawl. Instead of indexing data after it’s become stale, Positively analyzes the sentiments in pages as they’re downloaded. No indexing, no large-scale distributed processing. No headaches. Oh, and you can’t get fresher semantic content.
Second, Positively knows that sometimes you need more than a number. As it crawls, Positively automatically discovers attributes associated with each of the people, products, or services it’s investigating — and then figures out what people think about each of those attributes. Interested in District 9? You might be interested in its:
- plot
- actors
- humor
- cameos
- visual effects
Want to track down information on AT&T cell phones? You might want to know about their:
- battery life
- reception
- chargers
- apps
- features
- size
- display
No, these attributes don’t come from some big, pre-cooked list of things that might (or might not) be relevant for each product category. In order to discover why people feel the way they do, Positively hunts for each of the attributes associated with an item — and then discovers what people actually think about that attribute. Here are a couple of examples for District 9:
Despite its successful launch yesterday, Positively won’t be available to the general public until 80Legs goes live with its App Store later this Fall. We are, however, giving sneak peaks. Want one? Email me at andy@swingly.com.
Oh, and there’s Boston after the jump!
Tags: 80Legs, Sentiment, SwinglyThe not-so-itsy-bitsy-spider: 80Legs
Wanted to drop a quick post in honor of today’s public launch of 80Legs at the DEMO09 conference.
As followers of TechCrunch, Web2.0, SemanticWeb, GigaOM, and DEMO09 already know, 80Legs is a web crawling and online content analysis service which offers users access to more than 50,000 computers which can crawl as many as 2 billion web pages per day.
Need to take home a slice of the Web? 80Legs makes it easy. Point your browser to their portal, specify your seed list (and some crawl preferences), and away you go! It’s really that simple. (Here’s a screenshot of their dashboard for a job we ran earlier today.)
Sounds good? It gets better. 80Legs makes all of this computational power affordable as well. How affordable? How about $2 (yes, really $2!) per million pages crawled and $0.03 per CPU-hr used. Don’t be afraid to do the math: a 5 million page crawl costs just a little more than $12.
But wait — there’s more! 80Legs is so much more than a crawling platform. It’s also the ideal foundation for a new generation of semantically-aware content processing apps, as well. 80Legs makes it possible for users to upload small content apps (usually < 20 MB) that can be run on each downloaded page.
Here’s where Swingly comes in. Want to limit your crawls to pages in French? Or docs that mention relief pitchers? Or pages that discuss how people feel about a particular product or service? As Steve Jobs might say, there’s gonna be an app for that.
Swingly started building natural language processing apps that can run on top of an 80Legs crawl earlier this year using technology licensed from its parent company, Language Computer Corporation.
These include:
- Language Detection: Don’t read Dutch? Don’t worry. We won’t let you crawl those pages.
- Semantic Crawling: Tired of keyword searches? So are we. With Swingly’s semantic crawling service, you specify a seed list of concepts, not just words. We’ll run all the queries you need to get every last drop of relevant content, regardless of what keywords you tried out.
- Named Entity Recognition: Want to find pages that include only certain kinds of names? Swingly’s named entity recognition service makes it possible to track down names from more than 2500 different semantic categories, ranging from startup companies to bands to stock ticker symbols to financial institutions to lawn mowers.
- Sentiment Analysis: Want to know what people really think about people, products, or services? Swingly’s sentiment analysis apps analyze crawled pages for reviews, opinions, and other kinds of subjective attitudes related to a certain category. Unlike other sentiment apps, this app actually discovers the attributes associated with category — and tells you exactly what people liked (and didn’t like) about it!
These apps are now currently running on the 80Legs platform — in fact, the Swingly sentiment analysis app made its debut at today’s DEMO09 conference! All three sets of apps (plus a couple more that we’ve got under development now) will be made availableto the general public when 80Legs launches its App Store later this Fall.
Can’t wait? Want a sneak peak at one (or all) of the Swingly 80Apps? Email me at andy@swingly.com!
Tags: 80Legs, DEMO09, SwinglyJust a bit of Swingly Magic
Just a cool piece of Swingly magic that I thought I’d share right before the long weekend.
Want to know what might be semantically related to your keyword query (or question)? Swingly’s got an app for that.
Details after the jump…
Etymology: “Hitting the Reset Button”
my first reset button
While it’s still only September, 2009 appears to be the year of “the reset button“. Someone left you a mess that you’ve now gotta clean up? Your initial approach didn’t fare so well? Made some mistakes that you don’t want to look at anymore? Well, it might just be time to hit, press, or push the reset button.
I spent some time looking at how this expression morphed from its humble industrial beginnings (e.g. hitting the reset button on a furnace, etc.) to a popular panacea prescribed by politicians and pundits alike (e.g. hitting the reset button on the healthcare debate).
More after the jump…