Tag: vertical search

Alt Search Engine Day in Review–Part 2

Charles Knight of Read/Write/Web launched the first Alt Search Engine Day at the Intercontinental Hotel on April 21, 2008. I discussed the first three presenting panels in a post yesterday. This is a continuation of that post and is a recap of the remaining three panel discussions. It was a captivating day because of the different approaches, philosophies and personalities of the participating companies. As a result. I tried to capture the Q&A from each participant as closely as possible rather than trying to summarize each panel to provide a sense of the event. The differing opinions also did not lead succinct summarization. The panels were:

Panel 4–Intent, Monetization and the Next Five Years
Panel 5–Visualization of Search
Panel 6–It is all about semantic search

Panel 4–Intent, Monetization And The Next Five Years

Laurieanne Lassek of Seeqpod (not in photo)
Peter Adams of Matchpoint
Tom Eng of Healia
Ravi N. Raj of Kosmix
Chase Norlin, Pixsy Corporation
Kasian Franks, Seeqpod

Q: What is your site’s defining characteristics?
Pixsy-we are more about prospect of discovery than relevance
Healia-because we are in the health industry, the quality and trustworthiness of the information are very important as well as personalization
Matchpoint-our users wish to make a transaction; they are interested in speed and relevance
Kosmix-we are for people who want to dig deeper, who want to go beyond a cursory five minute search Seeqpod-determining intent is critical for relevance and vertical search is better at pre-determining intent

Q: How can a site monetize beyond advertising?
Matchpoint-we enable verticals to monetize beyond display ads with a provider form
Pixsy-we are a service provider, we only want to generate search query volume on their partner sites, the sites already monetize, we create pages for them to monetize
Kosmix-user experience and monetization are closely tied together for us
Seeqpod-offer contextually relevant product search, and make the steps very close for finding related merchandise, after all, the product or service may be more relevant than the original search

Q: What do we see for vertical search in five years from now?
Seeqpod-we will be competing with Google on the vertical level. There are large portions of the web without links. The question is how does the technology and how does a algorithm reach the deep web and organize it? In addition, we may also want people to interact with their search results over time-share, carry, tag, etc.
Pixsy-the biggest trend is advertiser blend, content is starting to look more like ads, ads are looking more like content,this may influence search results..
Kosmix-vertical search will be offer much deeper content than a major search engines
Healia-it will be formed by context, predicting intent will be the Holy Grail
Matchpoint-the ability to ask about intent may make it more relevant, will improve ambiguity in the search stream

Q: What are you going to do when Google decides to roll into your market?
Matchpoint-they are already in our market, but we have differentiated ourselves and continue to innovate at a high velocity. The biggest barrier is to keep innovating. “Search” is giant market, it can handle more than one player

Kosmix-we are synergistic with Google
Pixsy-they are not in our business yet,if you can you build audience and monetize, you are in good shape
Seeqpod- it is raw competition algorithmically, other companies have models that will succeed where Google will not, much on the web can’t be indexed

I think Healia said it best, think of technology as solving problems, five to ten years it maybe a different company we are discussing, not Google…

Panel 5–Visualization Of Search

Laurent Baleydier of KartOO
Alex Zivkovic of Cluuz.com (Sprylogics)
Randy Smith-KoolTorch

The panel suggested two new opportunities are semantic search and visualization. They believe people will adapt visualized search because it captures much more information in a single screen and makes the search process more efficient. I have yet to try it out, but found the concept fascinating.

Kartoo–in quick review, there are three new strategies for visualization:
1. Get the full picture using semantic maps
2. Animate graphs to allow a user to jump from place to place
3. Facet mapping (KVisu.com)

Cluuz.com-everything they do involves networks or graphs. They use meta-search , extract, then disambiguation to create a graph showing the relationships. They have found users enjoy using the diagram.

KoolTorch-provides an aesthetically balanced view of extensive research results. The founder, wanted a better way to find results rather than scrolling past the top ten, etc. Only search engine that shows more than 100 results in one view, grouped into categories (Ebay becomes circles of books, collectibles, etc. WILD!)

It is possible to drill down through each group and keep going through every level of the database..

This panel did not follow a formal Q&A. As a result I tried to summarize the conversation.

Q: What is the added value of a map versus list of text?
It is more efficient manner of showing extensive information. The goal is too show a large amount of search results in one screen and help the user arrive where they want to go more quickly.

Q: What is the goal of semantics?
It is about showing the right stuff. We can complete semantic analysis to make the “maps” deliver better results. Semantics is a means to improve the search function and deliver better results. Allows us to start at a more accurate level. At some point, semantics with visualization will work together.

Q: How do you protect your interface from your competitors
Patent it!

Q: How are you going to gain users? What search queries have you seen users use to use a visual?
It must be really intuitive. If you have to educate, it is a problem. Users will use a map to eliminate the noise, when there are many meanings to a word. There will be an education phase. But if it is more efficient and more productive, it will be readily adapted..

Q: How do you answer a simple query?
We offer a hybrid which includes the visualization and a list of results.

Q; How many entities can you show?
The maximum is 50 otherwise it can be confusing. Color and shape can also be used to represent different concepts.

Overall, the panel suggested that the use of semantics and visualization together will create the best interface.

Panel 6. It’s All About Semantic Search
Eegi & Nagaraju

Nagaraju Bandaru of BooRah (on left)
Frank of Eeggi (on right)
Barney Pell of Powerset
William Tunstall-Pedoe of True Knowledge
Kathleen Dahlgren of Cognition

Q: What is your biggest challenge?
True Knowledge-key challenge is building the structured content to answer the questions properly.

Q: What can be done now that couldn’t be done before your site was created?
Eeggie-we can really search the internet in any language, we are building a real world wide engine, we are introducing the tail searches and can be very specific, you can talk to the machines now
BooRah-words mean different things to different people, we can now add in personal preferences, we can extract relevant results
Cognition-can find only relevant results based on a map of the English language, most search results now are irrelevant, we deliver high precision and high recall
True Knowledge-we can answer any question through inference and enable users to contribute to the site
Power Set–better reflection of the results of the search and synthesized results

Q: What percentage of search is semantic based?
Powerset-if you look at natural search engine questions-2 to 4%, linguistic query is 50% or so, 50% failure is due to linguistic mismatches

Q: What is one thing you believe that others may doubt?
Cognition-pure math is hopeless due to the symbology of language
Eeggie-we can handle the language questions very easily
BooRah-can this really be that smart?
True Knowledge-people will respond rapidly to change

Q: How many people today think search is perfectly sufficient?
Audience-no one.
BooRah-Intent for local search is clear. Solve relevance first. Information must be synthesized because there is so much more information available on everything now.

Q:How do you get in front of the user?
Compelling user features, viral marketing and quality over quantity.

With that, the event concluded. It was a day of learning about innovation, ideas and networking and lots of discussion about ontologies and the semantic web.

I wonder if five years from now, if we will be discussing one of the sites in the room ad nauseum or if will we still be talking about “the Google.” What do you think?

FriendFeed, Eurekster, UpTake, and Spock presenting at 4/21 AltSearchEngine Conference

AltSearchEngines.com logo

On Monday April 21, I’m moderating a panel on Reaggregating User Generated Content at the invite only Alternative Search Engine get-together hosted by Read Write Web. Charles Knight of AltSeachEngines.com.

What questions should I ask the panelists? Drop a comment here and I’ll try to get the questions answered!

Here’s some info on the panel and panelists:

Q: Who are the panelists?

Spock people search logo

Jay Bhatti, Co-Founder of Spock

http://www.spock.com/jay

UpTake Travel logo

Yen Lee, Co-Founder and CEO of UpTake

http://www.uptake.com/team

FriendFeed logo

Bret Taylor, Co-Founder of FriendFeed

http://friendfeed.com/about/team

http://friendfeed.com/bret

Eurekster logo

Steven Marder, Co-Founder and CEO of Eurekster

http://www.eurekster.com/about/management

Q: What is the panel about?

The Panel Topic is “Vertical Search Opportunity: Reaggregating User Generated Content”

Through blogs, social networks, and review communities, more and more user generated content is being created. Problem is, that it is increasingly getting fragmented at multiple sites. This content can be useful to people for discovery, research, and decision-making on what information to consume, what music to listen to, or what products to buy. What are the opportunities for vertical search to bring all this user generated content together?

Panelists: Your mission (if you choose to accept it) is to learn from each other and force each person in the audience to rethink their approach to user generated content because of what they have heard.

Q: What questions are you going to ask?

NOTE: I would love to get questions from you to ask. Drop a comment here to let me know what is on your mind.

Panelists: I’m only going to ask questions if the audience doesn’t. But just in case, here’s what I had in mind:

  1. How does your company aggregate user generated content?
  2. What are the most valuable sources of user generated content in your vertical space? What are the challenges of getting at those sources?
  3. How much do you rely on automated machine approaches vs. human approaches? Why did you choose one vs. the other?
  4. Is what you are doing complementary with the sources you are aggregating? Who is threatened by what you do?
  5. How do you use user generated content to (a) get traffic from Google, social networks, and the blogosphere?
  6. How do you use UGC to acquire customers? To keep them? To engage them?

Why FriendFeed belongs on this panel:

FriendFeed is a lifestream aggregator that pulls together feeds from Twitter, Flickr, Google Reader, your blog, etc. By connecting with other friends, it is organizing user generated content around people…their interests…and their friends. It also has a new search feature that is desperately needed to help people use FriendFeed for discovery and search on specific topics. This is critical especially because there is more and more comments (UGC) that are entered straight into FriendFeed itself.

Why Eurekster belongs on this panel:

Eurekster allows people to create specialized search applications focused on a specific community or area of interest. It is enabling subject matter experts to layer human intelligence on top of automated search. Swikis provide a system for injecting user generated content and user contribution into a search engine. Many of the same themes are here as for FriendFeed.

Why Spock belongs on this panel:

Spock is aggregating structured and unstructured information about people. But beyond just being a people search engine, Spock also enables people and their friends and fans to participating in creating high quality people information on Spock. How machine intelligence and human participation can work together to solve people search has a similar feel to what FriendFeed is doing and Eurekster as well.

Why UpTake belongs on this panel:

UpTake is capitalizing on the increasing fragmentation of user generated content in the form of reviews, opinions, travel blogs. There is an opportunity to reaggregate and use machine intelligence to help people find the right reviews and recommendations for travel planning. Just as FriendFeed aggregates digital lifestyle, UpTake aggregates digital reviews and reputation for travel products and services.

OK your turn: What questions do you have for any of these companies? What is most interesting about the issue of reaggregating user generated content and how vertical search plays into it?

Drop a comment here by Sun 4/20 midnight and I’ll try to ask the question!

Custom Search

The Vacation Bloggers

BlogCatalog Viewers

MyBlogLog Readers

Meta