John Battelle has written about Transparensee, first giving an overview of their discovery engine and later using their fuzzy matching on a real estate database.
Steve Lavine, the CEO of Transparnsee, is in town this week and I had a chance to talk with him for an hour or so this afternoon. During that chat, he gave me the green light to show off something I'd seen a few weeks ago when it was still in development.
Transpaensee has put together a technology demonstration using digital camera product data from a popular technology publisher. The goal was to build an interactive shopping aid that lets you explore the universe of a cameras using whatever attributes you like to slice the data: price, resolution, editor's ratings, size, etc.
When he talked me thru the demo on the phone, I was very impressed. The site worked the way I wanted it do. The last few times I've shopped for a digital camera, I knew the key featues I wanted and had to jump thru many hoops to figure out what the landscape looks like.
The really slick part is that the fuzzy matching means it's harder to back yourself into a corner when exploring the products. If there aren't many exact matches, no worries. The software provides the next best matches so you can see what's nearby. Example: finding all the 5.0-5.9 megapixel cameras that have firewire (screen shot). There are two exact matches but several close matches are included below them. By ajusting the "megapixels" or "interface" sliders, I can control which feature is more important in my shopping and refine them on the fly. And, for complete control, I can click the "show form" link to unvil all the controls.
That makes it an interesting mix of searching and browsing. It's like having an intelligent guide by your side at a large electronics store. You still get the ability to drill down quickly but that doesn't completely shut you out of looking at what's nearby on the shelves.
Anyway, I can't do the demo justice just by writing about it. (Yeah, I know I said something like that last time I wrote about Transparensee.) Go play with it. Try the sliders. Drill into the various dimensions. See what you think. I really like the quick refinements I can perform.
Keen observers will note that we already provide some functionality like this in Yahoo! Shopping. If you look there for digital cameras, you can drill into various buckets like "over $400" and then further dive into "canon", for example. And then there's SmartSort, which takes a similar but different approach.
Anyway, it's cool to see better interfaces becoming available across the board. Given the amount that on-line commerce and listings will grow in the next few years, there's a lot of incentive to get this stuff right.
[Disclaimer: I have a very small financial interest in Transparensee. Of course, I have a much larger one in Yahoo! and never felt compelled to say so. Funny, that.]
Posted by jzawodn at September 06, 2005 10:38 PM
Looks is more of what one could use than the existing ones.
But still not natural enough. One thing which always bugs me - it seeems as if I am the only person on the planet who wants the "not" thing. "Not label X" instead of Label a b or c in my Gmail.
I have no feeling what I want, but the moment I see something, I can tell you what I don't want. This is usually a stronger selector than "choose option a".
Or in this case it is not about important for me, but the range of things I find important. Yes, make it high end, but max xx dollars and I don't need feature a that much. But show me everything else.
One thing with shopping sites is - as I just bought my tablet through it - that they all list things differently and not "pure" enough. I don't care about cases. I want a tablet. And I want this thinkpad, but the configuration is only important connected to the price.
"If this tablet in its high end setup costs only as much as the ones I am looking at, I don't care that it then has the bells and whistles I just told you I don't want because they make it expensive - I take it."
So, every step is a good step. ;)
Looks like another (good) implemetation of faceted browsing. I will have to add this to my list of interesting examples - http://user-experience.org/uefiles/facetedbrowse/
Still kinda funny that everyone uses digital cameras for the demo (I think I saw my first digital camera faceted browsing demo 5 years ago, and they keep on coming).
I guess I'm having difficulty understanding exactly what is different about Transparensee, over and above the other facet applications out there. Keith mentions both Endeca and Sidrean. Can someone explain what Transparensee does better/ different?
Eric, the major difference appears to be "fuzzy" results. In other words, showing closest matches - even if there are no exact matches. Also, being able to order those results based on how closely they match your preferences based on ALL the search criteria (as opposed to just sorting on one field) and then use the sliders to change the degree of importance.
It reminds me of a slightly more advanced version of the parametric search that a company called Cadabra had in the late 1990s. That company became Goto.com's shopping division, but was eventually shut down after the company decided to focus solely on paid search (and become Overture...). I wonder if Yahoo! still owns the technology/IP for it.
The problem with this particular camera search is not the technology, but with the parameters for the search-the only parameter truly relevant to picture quality is the focal length. The main parameters a user should be using when looking for a camera should be:
- Price
- Size/weight
- Picture quality
Andrew S. : Here's something that has similar functionality and allows you to search by the criteria you mentioned:
http://www.dontbuyjunk.com/index.php
The site doesn't delineate between fuzzy and hard matches in the way Transparensee does - it's implied that things match on a gradiated scale.
Kevin: Very interesting link, but it appears Dontbuyjunk has a lot more there than spec filtering. While Transparensee has some interesting functionality for specifications, it appears Dontbuyjunk allows users to customize their search based on more useful things than specs. For instance, I can tell it I care more about Portability when looking at digital cameras (one of many attributes), and it looks like they reorder the product listings based on review data for that attribute. Very powerful, and much more useful for the shopper!
It's interesting. Hopefully this will be adopted by all the major engines soon.
Somehow, in the race to improve search technology, the mass of structured data is all but forgotten. By focusing where few others are, Transparensee delivers real innovation by adding fuzzy search logic to SQL queries. This innovation will undoubtedly bring significant benefit to e-commerce and other transactionbased businesses that rely on structured databases to deliver service value.
It looks like Interface was very user-friendly. I have seen a demo and he has some valid points and the point about shopping to extract exact data not garbage one he hit jackpot. I have search over internet and comes up with lot of garbage data and difficult to decide what to do. It seems to me that this technology will help us and save our time to find exact shopping product.
