Amplitude

Measuring ROI on Semantics

mikepetit — Mon, 01 Mar 2010 18:18:00 GMT

We know, implicitly, that understanding the meaning of web content must make it easier to classify, to monetize, and to accomplish whatever mission we're on. I mean, how can you use it if you don't understand it? That's obvious.

...(read more)

Your Sick Isn't My Sick

mikepetit — Mon, 01 Mar 2010 18:02:00 GMT

Even a word like "sick" can have both positive and negative associations, depending upon context. So how can an automated service like OpenAmplify hope to address the fact that one man's bad sick is another man's good sick? That's one of the big challenges on the table for upcoming versions.

...(read more)

Why this stuff is hard :-)

mikepetit — Thu, 19 Nov 2009 21:38:00 GMT

We just convened a team in New York City for three days of software design sessions. Way fun. On Monday night, we went down to NuBlu, a cool club on Avenue C, to hear some jazz. Some of you may know that I've been a jazz musician for most of my life;...(read more)

The Evolution of Web Search: Part III

mikepetit — Tue, 17 Nov 2009 18:55:00 GMT

In Part II, I opined about the limits of topical search, and said we need web search that truly informs us, rather than just overwhelms us. To support our decisions, we need search that brings in attitudes, emotions, and all the other aspects of the decision...(read more)

The Evolution of Web Search: Part II

mikepetit — Thu, 29 Oct 2009 17:50:00 GMT

Really, what we want is for search to provide us not just with content, but with some up-front understanding of the content. We want to make decisions with the full meaning of the content: the attitudes, the quality, the emotions, all of it. Just as we do with any important decision, we want to make it by becoming informed, not overwhelmed.

...(read more)

The Evolution of Web Search: Part I

mikepetit — Mon, 19 Oct 2009 17:45:00 GMT

Life used to be so simple. You had only one choice for a telephone company; you picked up the phone and it just worked. Toothpaste was just toothpaste, and once you picked one, that was it, pretty much for life. It wasn't that long ago when we only...(read more)

Online Advertising Can Make Us All Cool

mikepetit — Fri, 16 Oct 2009 18:01:00 GMT

We all deal with online ads. Few of us would say that we want them; almost none of us would say that we like them. But, what if ads could make us cool ? I've mentioned my old Saab once or twice. It's actually a pretty special Saab, a 1991 limited...(read more)

Eight Years Later

mikepetit — Mon, 14 Sep 2009 15:33:00 GMT

I was in Stockholm on 9/11, eight years ago. Hapax was a new company, with a brand new product called FindEngine, an NLP-based search engine. I was sitting in a board meeting when Emil, our long-time Senior SysAdmin, stuck his head in the door, and said, "Mike, I thought I should tell you: it looks like a plane hit one of the World Trade Center Towers."

Bummer, I thought. Hope it's nothing too bad. "Thanks, Emil, appreciate the heads-up." After all, I was in a board meeting. Let's get our priorities straight, I was thinking.

A little while later, Emil interrupted again. His face said that something was wrong, very wrong. "Mike, another jet hit the other tower. They're both burning."

Meeting adjourned. We ran downstairs to the TV room, and sat in utter disbelief as it all unfolded. My friends and colleagues, among them current colleagues Sandra, Emil, Kalle and Mark, could only sit helplessly, shocked themselves, but also aware of what I was going through. They knew I was from New York. They knew that my brother was in New York, and most of my best friends, too. They knew something bigger than just two burning buildings was taking place. Nobody said much, but we all knew. And we were all so helpless.

The three days we all spent right after 9/11, working all-nighters at the request of the Swedish Secret Police to analyze a huge body of web intelligence, made us feel a little less powerless. My brother escaped the carnage, as did my friends, but every one of us knows somebody who lost somebody.

Eight years later, Emil, Kalle, Sandra and many ex-Hapaxians are still my friends and colleagues. We inhabit a different world, but the basic ties that bind us all together are intact and, if anything, more important than ever.

Oracle/Calais a Good Example of What We Must Do

mikepetit — Fri, 04 Sep 2009 16:22:00 GMT

Oracle's announcement that it can access Calais data and tag rows in its RDF-oriented database system is a great example of something I've believed for some time: for the Semantic Web to deliver benefits, it must become part of the workflow. As more and more tools of the trade (in this case, a prominent DBMS) integrate semantic data into their fabric, the value of better classification and deeper machine understanding of content becomes more apparent. Bravo, Oracle, and bravo, Calais. This is the way forward.

Dorthy.com serves up the dreamscape

mikepetit — Tue, 01 Sep 2009 13:20:00 GMT

A new site, Dorthy.com, is getting ready to deliver your dreams, along with some pertinent advertising :-)

They claim to be using, among other things, NLP to relate content conceptually, delivering that content to "dreampages" that you set up. Always wanted to sail around the globe? They'll feed the dream, along with brand-oriented advertising to help you gear up for the trip.

According to their press release, which you can read at http://www.reuters.com/article/pressRelease/idUS49708%2B01-Sep-2009%2BBW20090901, more than 90% of us have dreams we'd like to realize, and a goodly percentage of us use the Internet to figure out how to pursue them. Dorothy has a few ideas for us...

Cloud Computing Works!

mikepetit — Thu, 20 Aug 2009 15:50:00 GMT

We got hit pretty hard last night. Somebody (or somebodies) really had some stuff to analyze! That's great, of course, but it does pose a challenge to the infrastructure that underpins OpenAmplify, a challenge that requires management.

For most of the history of computing, that management required significant planning: for average load, peaks, failover, etc. It also required that one purchase, lease or otherwise make available the hardware and software needed to handle the peaks, etc. Since the hardware was obsolete the moment it was procured, one was always compromising on cost/performance. "TIme sharing" was an alternative, but didn't really scale very well. You just had to bite the bullet and write the checks. Big checks.

Ah, the good old days: I would spend weeks, or even months, preparing plans, RFP's, checklists...I could count on a bunch of really good, free lunches as I pitted the various vendors against each other. Eventually, I'd have an infrastructure that would suffice...for awhile.

Those days are gone, at least for me. OpenAmplify is built on the Amazon EC2 cloud. No more free lunches. But, no more headaches, either (or, at least, far less of them). Our infrastructure makes itself bigger when the load increases, makes itself smaller when the load decreases, senses when a server is unhappy and simply replaces it with a new one. Presto! All automatic, no muss, no fuss. I get a monthly bill from Amazon, and they manage all the servers, networks, power supplies, and all the rest. I don't even know what hardware they're running.

So when we got slammed last night, I got the best possible news: none. I didn't even know about it until Ram mentioned it to me in today's Managers Meeting.

Cloud computing rules.

Triangulating Partnerships (Welcome, Glen!)

mikepetit — Tue, 11 Aug 2009 19:55:00 GMT

We've recently had the good fortune to convince Glen Kushner to join the OpenAmplify team as VP for Business Development. In that role, Glen will build our business relationships, both bilaterally (that is, B2B, case by case) and more broadly as part of a partnership process.

That partnership process is deeply interwoven with our Community. We've said, many times, that one of the primary objectives of the OpenAmplify community is to encourage and support the development of real-world applications that extend the benefits of the Semantic Web to all. This implies cooperation: someone needs to define what is needed, someone needs to describe how to meet that need, and someone needs to build it to that specification. Sometimes, one person can cover all three bases; more frequently, it's a combination of talents that get the job done. We call it "triangulation": OpenAmplify at one point of the triangle, a customer or user at another, and someone (or some ones) at the third point to make it happen.

We try to encourage such collaboration at community.openamplify.com. The ability to create Web 3.0 projects, attract people to them, and take them over the finish line requires a certain kind of entrepreneur, and we want folks like that to make our Community their home. We underwrite that by providing a venue, a set of tools and, importantly, a meeting place for talent. We hope you'll create Projects from your best ideas, or add your abilities to other peoples' Projects.

Glen, in the meantime, will be "triangulating" all over the country, and, concurrently, keeping an eye on the Community to see what promising Projects we should put extra horsepower behind. You never know when or where Glen might pop up...

Welcome, Glen, VP of Triangles.

What do Twitter Users Actually DO?

mikepetit — Wed, 15 Jul 2009 20:38:00 GMT

I'm really gratified by the number of apps being developed using OpenAmplify to mine sentiment from Twitter. Clearly, there's a perceived need, and, as I recall, the folks at Twitter themselves have encouraged it.

I'm curious, though, as to what Twitter reveals about actions. Can it profile the Twitter community by career? Can we figure out how much intent there is to change things, or fix things, or...?

The new TopicIntentions analysis could give us some answers. Since it groups the actions of each major topic by ActionType (buy, sell, read, build...) it should provide a really easy way to see what Twitter users intend to do about things. I'll probably take a shot at mashing something up, but I'm betting the Community can do more, better than I can, and certainly faster.

So, tell me, what do Twitterers actually do?

Now Available: TopicIntentions

mikepetit — Thu, 09 Jul 2009 13:36:00 GMT

Version 1.1 of OpenAmplify is now in production! It offers lots of good, new things and improved performance. One thing, however, merits particular attention in this blog, as it meaningfully advances the state of the Semantic Web.

TopicIntentions is a new analysis, selectable via the standard API. It's designed to make Actions much easier to consume at scale, as it directly relates each topic to its associated actions, via ActionType. It will tell you:

The most significant types of actions related to the topic (buy, sell, build, vote, whatever...)
Whether the topic in question is the subject or object of that action (is the bear doing the chasing, or is the bear being chased?)
The associated decisiveness and guidance scores for that ActionType as they relate to the topic in question.

All this in a single XML structure that includes all the other topic-related signals, such as Polarity.

Why is this important? Because it permits users to focus, directly and efficiently, on what the author of the content is doing, did, or intends to do, and whether that decision can be influenced. Imagine being able to target online advertising directly to users whose posts or comments indicate an intent to take a desired action (buy or vote...) and who are seeking advice on it. Or imagine law enforcement being able to determine that certain conversations strongly advocate taking violent action against certain groups or persons. TopicIntentions makes these and limitless other uses of Actions possible, practical and scaleable.

Only from OpenAmplify!

Wicked Difficult to Handle Dialect

mikepetit — Mon, 29 Jun 2009 13:10:00 GMT

I just returned from a few days on Cape Cod. Notwithstanding the weather, it was a nice time. There were, however, some comprehension issues.

In a chocolate store, from the shopkeeper to a local customer: "Gift box is ready. Didja find the cod you needed?" I'm thinking, cod in a chocolate store? Of course, the reference was to a "card", specifically a Happy Anniversary card, and I kind of knew that, but it still caught me off balance. It was the first day, after all: my ears hadn't adjusted.

On a t-shirt: "Welcome to the Lawbsta Bah". Cute, but tough to parse, at least for a computer.

My point? We can, and we will, always try to close the gap between human understanding and machine understanding, but the limitless variety of spoken and written dialect pretty much guarantees that humans will always stay ahead of the software.

That means we'll never run out of challenges. And that's the good paht. Ain't life a howlah?