Entries Tagged as ''

We’ve Found the Last Mile of Filtering

… and it is human. Try as we might in computer science we have not created a piece of software able to pass the Turing Test.

An interrogator is connected to one person and one machine via a terminal, therefore can’t see her counterparts. Her task is to find out which of the two candidates is the machine, and which is the human only by asking them questions. If the machine can “fool” the interrogator, it is intelligent.

The Turing Test has long been used as a litmus test for intelligence. What the test really assesses is the ability to understand the dynamic state space of context produced by human language. And here we’re only talking about written language, not the really hard stuff like vocal inflection and body language that get added to the mix in face 2 face conversation. The state space is dynamic because languages change over time, words change meaning or are used in different ways, or new words invented. There is a great post that talks about the use of jargon as conversational shorthand here.

Filtering unstructured data possesses the same qualities as the Turing Test. The interrogator would like to find intelligent commentary on their subject of interest. There are many ways to whittle down the problem and some are chainsaws that can eliminate broad swaths of non-relevant content (think spam and porn - unless that’s what you’re searching for of course) in a single stroke. And, there are some especially useful tricks you can use when constraining the problem to a limited space (like publicly traded companies). But at the end of the road of filtering it is you and I that make the call on the quality of data because it is subject to our tastes, our interests, our context.

This is what Sergey and Larry cottoned on to when developing Google page rank and the guys at Digg tried to get at. User voting, either implicit via links or explicit Digging gets part of the way there, but ends up short of the goal. In order to get as far down that last mile of filtering as possible we’ve employed human technology (or as Bezos puts it in his Mechanical Turk, “artificial artificial intelligence”). This greatly simplifies the problem that must be solved, which becomes, how do you pre-filter the data enough so that the human filtering problem is scalable.

More on that later…

What Traders & Portfolio Managers Actually Need

We spent over a year working exclusively with Wall Street firms, and now a couple of the largest sell-side investment banks and over 60 hedge funds use our service. This group of customers is EXTREMELY demanding, so it has been a great place to refine our technology.

Here is what we’ve learned about traders. They need their info to be:

- absolutely comprehensive (“don’t miss anything”),
- timely (“I need it in real-time”),
- appropriately filtered (“just give me the important stuff”)
- and granular by topic (“those guys care about fuel-cells, I only care about Xbox”).

We offer traders a choice of interfaces: email alerts, datafeeds, GUI’s or little widgets that don’t take up much room on your screen.

Or New Media sources are the most comprehensive in the business: Blogs, boards, forums, social networking sites, websites, local news sources, international news sources, trade rags, and we even throw in mainstream news as well.

The technology platform we use is an extension of similar business intelligence tools many of us created together in a previous company. Most of us have been working on the problem of filtering unstructured information for 8 years. Our CTO has a patent related to it.

We take a bottoms up approach. “Scraping the blogosphere” simply doesn’t work – we find many more obscure nuggets using our approach. If people are telling you different, they actually don’t understand the problem. We build “TopicNets” that continuously track activity around the things you actually care about.

Ranking the results in some sort of informed way has been the main problem we’ve been working recently. We can use a variety of approaches, from “relevance” to “credibility of the author” to “topicality” to “level of influence”. Again, we give the traders the ability to rank results differently depending on what is important.

Filtering New Media information is a tough problem. But finding one good nugget that no one else has can generate one killer trade…..

Yet Another Company I can’t Explain to My Mother

It was bad enough in the days of Web 1.0 (has enough time in Web 2.0 passed for us to refer to 1998 as Web 1.0?). Back then, Most of us Collective Intellect staffers were working in some dot.com or telco software company or “ on the web” company. We were writing code, provisioning bandwidth or drawing up strategic powerpoint presentations with bullet points like “demand for IP is doubling every 6 months” and “build traffic first, then worry about how to make money”. Despite repeated attempts at explanation, my mother still described to her friends what I did as “something with computers”.

A lot has changed since then, and some things have stayed the same.

THE SAME: Venture Capitalists love disruptive technologies, experienced management teams and big marketss
DIFFERENT: Venture Capitalists actually expect businesses to make money.

THE SAME: The amount of information out there is growing exponentially.
DIFFERENT: Blogs and MySpace have replaced “homepages”. In between Chinese dudes’ lip-synch videos and teen-age diatribes are great posts about pharma, consumer technology and healthcare.

THE SAME: Small companies are innovating faster than Big companies.
DIFFERENT: The small companies are filled with entrepreneurs who have been around the block and are wary of dot bomb ideas. Plus our girlfriends/boyfriends are now our spouses and and they won’t let us work at companies with stupid ideas anymore.

And while my mother still describes my job as “something to do with bluggs on the Internet”, and least she faithfully reads my blog through her RSS reader now.

Welcome to Collective Intellect. We are the company that monitors and filters the crazy world of New Media for the largest institutional investors and Fortune 50 companies in the world. If you have budget for it, we’d love to help you.

Now excuse me while I go try to explain to my wife why I have a profile on MySpace……

-DK