Multi-constraint Search: Emerging
Lately, I'm bumping into more and more projects that aim to automate and handle complex consumer web searches and transactions, like booking travel for a multi-leg overseas trip with lots of family members who have varying needs. (You could think of these services as multi-constraint search.) To wit:
* At the Web 2.0 Summit last fall, I had lunch with a woman from Sun R&D whose team is working on processing systems that will have the power and intelligence to quickly handle queries with a lot of parameters.
* Rearden Commerce, one of the Web 2.0 Expo exhibitors, has a service that "helps employees purchase and schedule services." It does so in part by using a restricted network of vendors, which isn't terribly revolutionary (see: managed health care), but on the user side, it uses personal info and parameters you give it in order to find appropriate transactions.
* I met the other day with Anupriya Ankolekar, whose PhD work is on the potential intersection of the semantic web and web 2.0 (pleasant surprise: the discussion was nearly jargon-free). She noted that in semantic web communities, it's increasingly thought that one of the best ways to help people understand the idea and power of the semantic web will be to have services that handle complicated search-and-transaction processes for you--and that a handful of such projects are under way.
All of the examples I've seen so far have focused on consumer applications, but presumably, these services are in development--and probably further along--for B2B apps.
Nat notes that in computer science, "constraint satisfaction" is the the term used for the underlying process in a multi-parameter search. That's useful to know, but neither "constraint satisfaction" nor "complex search and transaction" seem like the right name for this emerging class of services. Does anyone know of a term in use or have a great suggestion? And do you have additional examples of projects underway?
tags:
| comments: 11
| Sphere It
submit: 
0 TrackBacks
TrackBack URL for this entry: http://orm3.managed.sonic.net/mt/mt-tb.cgi/2155
Comments: 11
[06.17.07 01:04 PM]
Search 2.0
[06.17.07 01:50 PM]
While much of the underlying techniques and technologies may be search based, much of what the person using those technologies experiences will, I believe, not look like or feel like what we think of as search, structured or otherwise.
Right now we have a bad case of tunnel vision - everything looks like search and, more to the point, like explicit search where someone manually asks for information and then takes the results and does something else with them. While sometimes this is what we really need, much of out activities are focused are not explicitly requests for specific information, but are expressions of needs or desires - "I need to buy a new printer" or "I want to take a vacation" or "we need to meet with clients and analysts on the east coast next month."
Let's use the common example of planning a vacation. While it can be expressed as a series of searches for flights, hotel reservations, restaurant reservations etc., what people really want to do when they're taking a vacation is to go somewhere and do something. There are certain constraints (how to get there, needing a place to stay) and other items that are optional (seeing art, scuba diving) but the current state of the art forces the vacationer to explicitly state all of this and to make sure that the various parts fit together. Yeah, the system returns information based on my search, but it's making ME do all of the real work.
Now imagine a system that could enable someone to create a trip based by stating what they really want to do - "I want to go to Florence and Rome for Sept 1-10 2007. I like art, fine authentic food and and wine. My budget is $200/day for a hotel. Oh and I want to make sure I see the David and the Colosseum." My request certainly implies a series of searches, but I don't think of it that way - I want to take a trip to Italy. I think of this as implicit search, where the underlying systems combine lower level information (flight, hotel, budget issues, timing) into a higher level product (an itinerary).
Much of our personal and professional lives is spent in this way - our activities might involve gathering information, combining it with previous knowledge and achieving something, but we don't think of that as search... we think of it in terms of the end goal. Implicit search isn't going to displace Google - we'll always need somewhere to turn to directly ask for information - but it could certainly underpin a lot of services that would enrich our online experience immeasurably.
[06.17.07 07:22 PM]
Social Networked Vertical Results are the future(word of mouth combined with premium vertical channels). The founder of MyLocator.com was able to go out capture a cluster of premium strategic vertical properties that are keyphrase, toplevel. He is the pioneer of Universal Vertical Locator Cluster Technology. Today he controls a porfolio of over 1100+ vertical locator properties. Creating a whole new genre of location he calls "Universal Vertical Locator Cluster Technology." Claims that what we need are Locator Engines. These early visionaires have to potential to for ever dominate this arena. I will say it for him "Search is Dead; Strategic, Social networked vertical results are Everything. Game Over. Search Solved.
MyLocator.com
Think Outside the Box
[06.18.07 04:52 AM]
Sarah. Could you clarify this.
When you say:
"* Rearden Commerce, one of the Web 2.0 Expo exhibitors, has a service that "helps employees purchase and schedule services." It does so in part by using a restricted network of vendors, which isn't terribly revolutionary (see: managed health care)"
What does see managed health care mean?
Is it something that Rearden does?, is there a managed health care site? or are you just making a comparison?
[06.18.07 05:41 AM]
To your question - "a term in use": Qualitative computing."
Moving beyond the current quantitative tag-based systems, we view qualitative meaning-based indexing and autonomous intelligent software agent technology as strategic to the emerging “semantic generation” of search, persistent awareness, information extraction and content management. This is a competitive - must have - for early adopters to gain competitive advantage. Ontology-based next generation architecture of rich-media content indexing enables maximum personalized information awareness. These are disruptive domain specific software systems and methods that enable the capture (abstraction) of the full contextual meaning expressed within the content of information objects of any type - text, images, sound, shape, voice, etc. The reusable “teaching” within each is understood, abstracted as graph fragments and independently indexed for in-context rich media retrieval. Vertical language (community jargon) queries are your on-demand personalization - the more contexts articulated, the higher is the precision of the re-ordered returns.
[06.18.07 05:43 AM]
Does anyone know of a term in use or have a great suggestion?
A couple of suggestions would be:
PERSONALIZED VERTICAL SEARCH or
MULTI-SEARCH TRANSACTIONS
... depending on the goal emphasis.
[06.18.07 07:14 AM]
Another related term in AI is "planning systems".
[06.18.07 10:24 AM]
In one very broad vertical market, even the semantic web won't be sufficient -- scholarly communications. Universities, scholarly societies, scholarly publishers, even scholars themselves have a vested interest in not just finding something, but knowing the "scholarly authority" of an article -- not just its popularity, but also its provenance, context, chronology, the scholar's other works, and myriad other lines of investigation.
There's an article that just came out in the Chronicle of Higher Education's Review that I wrote outlining some of these issues (written for administrators and academics, thus somewhat limited in scope), available in fulltext for free at http://chronicle.com/free/v53/i41/41b00601.htm
The implications are significant, in terms of where scholarly communications will go, and its implications on universities, libraries, and scholarship, in the next five to ten years.
[06.18.07 10:40 AM]
Great comments, folks.
Rick, I tend to agree that these services won't necessarily look like search, per se. We could say the same thing about a growing number of search-enabled Web apps; it'll just be part of the underlying structure.
John, in using the term "managed health care," I was referring to the general system in which an insurer gives access to a predefined network of providers rather than to the whole universe of health care providers.
[11.15.07 06:04 AM]
Some ideas:
TARGETED SEARCH - not original, but it may make a nice contrast with "general search" or "open search" as a name for what Google does, and it should make sense to the general public.
MULTI-CRITERIA SEARCH - less elegant that "targeted search", but also clear to the general public, I would think.
SPECIFIED SEARCH - I don't like this one as much, but it could accrue most of the meanings that "targeted search" implies.
BABY-BEAR SEARCH - Not too hot (your search returns 6 quadrillion pages), not too cold (terse and unhelpful results), but juuuuuuuuuuusst right (customized to your criteria).
HOMING IN - as an alternative to the "search" concept altogether. Maybe it's still "search" under the hood, but perhaps a strong distinction from "search" would make the services easier to position.
PINPOINT SEARCHING - playing again on the root concept of "specificity". I like this one a lot. It sounds both technically credible and publicly informative. It also shortens to PINPOINTING and all the other gerund forms that characterize a lot of good technical names.
Cheers all!
Post A Comment:
Radar RSS feed
Tony Trupp [06.17.07 12:23 PM]
You hit the nail on the head here Sarah. Companies realize that there is no way they can compete with google in terms of breath of simple keyword searching. Instead, the are opting for highly structured search that run on top of nicely segmented relational databases (vs the blob type storage used in keyword searches), and focus on one niche or industry. This gives users a higher degree of control over the results, as they can filter down the matches to meet a set of criteria. Our site VocalNation.net is experimenting with this for political discussion, allowing results to be filtered by media type, political leaning, etc. or browsed by region (continent, country, city, or state).