Discover’d!

I was going to wait until September 1st to share the news, but I’m too excited to keep quiet. I’m proud to report that Pluribo is featured in the September issue of Discover Magazine. (A low resolution scan is below, but you should go buy the issue to support this great publication.)

The article is humorously entitled “The Software That Will Make This Section Obsolete”; in case you don’t get the joke, it was written in the section of the magazine that generally has reviews of new products. The author, Bryn Nelson, did a great job of describing the potential of our technology while still keeping the tone light-hearted. He proclaimed that our “grammar often beats that of a human reviewer” and showed amusement at how our system can sometimes serve up some “terse put-downs” if it detects people thought a product sucked. Nelson reported, “For one heavily reviewed version of the iPod that was earning four stars overall, Pluribo yawed, ‘A ho-hum product’”. (Sorry, Steve J., we still love ya.)

On a personal note, it was a poignant honor to be featured in Discover. Back when I was in middle school, I used to be an avid reader of the magazine. From my little hometown in Oregon, Discover got me excited about the wonders of science and technology. In some small way, it inspired me to go down the career path I’m currently on. I can only hope that there may be some future AI, machine learning, or natural language processing guru out there who felt a hint of inspiration from reading this article this month.

More Than A Gizmo

One of the nice things about our launch was all the incredibly valuable feedback we got from you (our users). From the myriad comments we received, there was one major theme that was echoed time and again: you want Pluribo to work on more than just electronics.

Well, we totally agree. We’ve been hard at work tuning our systems to extend Pluribo into new areas. Rather than be hasty, we’ve tried to be cautious because we want the quality of our summaries to remain high; we think you deserve the most helpful and accurate summaries possible!

Towards this goal, this week we are testing a very exciting new product area for Pluribo to cover. We hope to release it as soon as next week, so stay tuned! In the meantime, we’d love to hear your thoughts on which new product areas you’d most like to see next?

A Resounding “Yes!”

It’s been about six weeks since we launched and the first thing we want to tell all of you is “Thanks for your warm reception!”.

For the last 16 months, we’d been in stealth mode working on our idea. It had been a lonely sort of experience, working in isolation and unable to get feedback from the world at large. We have to admit that we had a small sense of trepidation when we launched: Would other people find Pluribo to be as cool and useful as we thought it was?

From all the e-mails, twitters, blogs, and mainstream press coverage we’ve received since then, it seems that the resounding answer is “Yes!”. In fact, we were so overwhelmed that we thought we’d share a few of the highlights:

* On twitter, GreenSmith called us a “godsend”, and gubatron said he “thinks that watching Pluribo’s demo is a glimpse of what the semantic web will bring”.

* In the blogosphere, it was a thrill to be acknowledged by TechCrunch, who described us as the “CliffsNotes for Amazon reviews”, and by ReadWriteWeb who appreciated our “accurate summaries” and thought of us as “akin to a Zagat review”.

* The hyper-intelligent communities at MetaFilter, LifeHacker, and Slashdot also weighed in, giving us some excellent support and feedback.

* But perhaps our favorite of all these articles was the one written over at Ars Technica by David Chartier, who had a very forward-looking perspective on our technology. David wrote, “Pluribo is on to something with this summarization engine” and predicted that summarization techniques like ours might one day be invaluable in taming information overload on the web. He imagined, “I can see all kinds of other consumer applications for things like long Wikipedia entries and even articles like the one you’re reading.”

On that note, there’s no better way to end this blog entry. Thanks again for all your kind words and we can’t wait to show you what else we have up our sleeves!

Out of Many, One.

We are pleased to announce today the launch of our very first product.

Over the last year, we have built a technology that can automatically summarize user reviews on the web. It uses novel artificial intelligence to take a webpage filled with reviews and boil it down into just a couple of fluent sentences. You can think of it as an automated Zagat-style brain for user reviews.

Today we are releasing a Firefox extension that applies our summary engine to reviews on Amazon. With the extension installed, all you have to do is visit any Amazon webpage that has a lot of reviews. Our technology will go to work behind the scenes and summarize all those reviews down into a couple of lines.

(At the moment, the extension only works for electronics items, but we’re adding new product categories all the time.)

To see a quick demo, take a look at our screencast. Full details, of course, are available on our home page.

We think we’ve come up with an interesting new approach to dealing with information overload. By using summary instead of search, we hope we can help people make better sense of opinions on the internet. Rather than filtering information and throwing out potentially useful opinions, our technology synthesizes all the available information.

Our belief in this approach is so central to our identity that we’ve chosen the name “Pluribo”. It is inspired by the phrase “e pluribus unum”, which means “out of many, one.”

So please take Pluribo for a spin. We look forward to hearing your feedback!

Be SAASy. Say No to the “Beta” Label!

Tomorrow when we release Pluribo, it will not carry that famous “beta” label that seems so in fashion among web-based applications. Why? Because people who use “beta” labels don’t really understand what software-as-a-service is all about.

In the old days of packaged software, back when you shipped distinct releases every year, beta labels made sense. People needed a way of differentiating what was a work in progress versus what was “finished” and deserved to go into the shipping box for customers.

Well in the world of software-as-a-service (SAAS), we pledge to our users that our software will never be finished. We will always be working to enhance it. Of course, in these early stages, it won’t be totally perfect and bug-free. But we aspire to release improvements early and often. That is the beauty of SAAS.

To put it another way, what we are really saying is that we promise that our software will *always* be in beta. We are restless people here at Pluribo. We want our software to stay on the bleeding edge– in that fertile ground where there is a continuous feedback loop between what people do with our software and how we improve it.

Given that we never want our software to leave beta, we see no point in using the label at all.

We hope other developers out there will join us in saying goodbye to beta labels. (Hello, Gmail? Get with the program! It’s been 4 years!) Only be relegating beta labels to the past can we truly embrace the promise of software-as-a-service.

The Tyranny of Five Stars

In our last post, we talked about how “summary” is a much better way than “search” for people to make sense of opinions on the internet.

There is, of course, one form of summary that already exists and is ubiquitous across the web: The Five Star system. There’s only one problem with it. The Five Star system sucks.

The major problem with the Five Star system is that it can only tell you whether people liked a certain item. It can’t say anything about why they felt that way.

For example, if you see that a movie got 4.5 stars, what does that really mean? Did it get that rating because it has good acting? A good plot? You can’t help but wonder, why didn’t it get 5 stars? Was the cinematography lacking? Or maybe it wasn’t funny enough? The star system can’t help you quickly discern that kind of information. You still don’t know the “why”.

Even in its stated task of helping you see how much people liked an item, the Five Star system is pretty useless. The vast majority of items on the web have a rating between 3.5 and 4.5 stars. Is there really any substantive difference between 4.2 stars and 4.4 stars? Even if one exists, it is hard to discern. The system doesn’t inspire much confidence.

Taken together, it’s easy to see that the Five Star system– as it stands today– has almost no communicative value at all. It’s time to liberate ourselves from this tyranny and call the system what it is: a weak numerical hack.

That is why at Pluribo we’ve been building a technology that can augment the traditional Five Star system with something much more powerful than a mere numerical summary. It will actually be capable of succinctly summarizing both whether and why people liked an item.

We will be launching it (in limited form) very soon. And we hope that once we do, it can help make that old Five Star system look positively paleolithic.

Search is for Facts. Summary is for Opinions.

Let’s get the first mystery out of the way: why we started Pluribo.

We started Pluribo because we wanted to make it easy for people to understand other people’s opinions. As it stands now, the internet is complete chaos when it comes to learning about opinions.

Don’t get us wrong. In general, this chaos is a good thing. There are millions of sites where people express themselves, whether it be on blogs or on traditional review websites (like Yelp, Amazon, IMDB, etc.). But although this creates a vibrant ecosystem for personal expression, this also creates a problem. With so many opinions in so many places, it can be awfully hard for people to make sense of it all.

You’d think there should be a painless way to answer a question like, “What do people think of the new iPhone?” But there isn’t.

Unfortunately, Search– which is the internet’s swiss army knife for organizing information– is not up to this task. It’s great for facts, but it is pretty useless when it comes to organizing opinions.

When you are looking for a fact, all you really need to know is the top search result. That’s usually enough to be satisfied. There’s no need to read the same fact a dozen times on a dozen different pages. Once is enough. As soon as your search has found the needle in the haystack, you’re done.

When it comes to opinions, however, you seldom want to know what just one person thought. You want to know what a lot of people thought. Consequently, using search can be disappointing. It doesn’t save you from the effort of having to read a bunch of different reviews, which could be spread over a lot of sites.

Since search is ineffective at organizing opinions, what is the right approach?

At Pluribo, we believe the right approach is to summarize rather than search. Instead of returning an ordered list of results (as search engines do), a better approach would be to actually just summarize all of the relevant opinions down into a short and easy-to-understand form.

In essence, what the world needs is a summary engine rather than a search engine.

That is exactly what we’ve been building for the past year. But more on that later….

Activating Uncloaking Device

After a year of hard work, we’re finally on the cusp of our first public announcement. As such, we thought this would be a good time to dust off the company blog and give the world a proper “hello”.

As we’ve been working in secret all these months, it’s been a little frustrating to not be able to fully share with the world exactly what we’ve been building. Now, at last, we’re thrilled to leave stealth mode behind!

So stay tuned to this blog as we begin to reveal Pluribo to the world.