Author: Amarpreet97
First, Wowd. When Steve Jurvetson pitched "federated search" at the Churchill Club Top Tech Trends event I covered in May, he was actually shilling for this outfit, which is one of his investments.
Other than economic, the big benefits of Wowd, CEO Mark Drummond and CTO Boris Agapiev told me, is that its index is fed by what people are actually looking at, not just what Web sites are linking to; and that it sees all the Web pages that users visit, most importantly those that can only be accessed by entering data into a form (the so-called "deep Web").
There are obvious privacy implications with this. So Wowd does not index sites where it sees personal data, like bank or healh-care pages, and users can easily exclude from the index sites they're visiting that are not automatically excluded.
When a user sends a query into Wowd, it's sent out to other Wowd users' comptuer that are close-by on the Net. They each contibute results back to the query where it displays. It's like the BitTorrent of search.
The service helps you discover what's hot on the Web, since it knows what people are linking to in close to real time. It also can segment search results by social proximity (giving people different results based on what their friends have clicked on), but that's not built into the launch product.
Challenges include the privacy issue, as well as the need to get users to install resident software on their computers for this to work. Drummond did say that the search engine will work well even if only a small proportion of users are running the resident software. And, obviously, it's hard math to make this work with good performance and reliability, and at scale.
The idea of replacing the search data center with a cloud of users' computers is intriguing, but I'm extremely skeptical that this will work: That it will provide noticable value over and above Google, not to mention work technically. On the other hand, there's so much money in search that you don't need to gain a double-digit percentage of the market to have a winner on your hand, especially if your cost of providing service is negligible, as it should be for Wowd.
The service in private beta now, but I should have access codes for it shortly. I'll post when I do.
The claims: Yebol being built using new neural network technology. Founder and CEO Honfeng Yin says his algorithms are related to the "associative memory model," and that his goal is to convert information to knowledge (I've heard that before). Compared to Powerset's semantic search, which Microsoft bought, Yin says that Yebol's technology is far more scalable, and that his index already has one billion pages in it. The service is currently running an Amazon's on-demand cloud services.
Enough talk. How is it? As a tech demo, great. Like Bing, Yebol displays associated keywords or search terms alongside search results. Unlike Bing, instead of just displaying categories (which is useful in itself), Yebol displays related words that I found surprisingly useful. For example, in a search for "Google," Yebol displayed, among other things, a list of Google employees that included not just the names in the news but other important people who are or once were affiliated with Google. A search for "Top Gear" displayed a list of sport cars that have been reviewed on the show. (There are other search engines that pick out information from results to categorize them; see Clusty, for example.)
Yebol does a great job extracting and categorizing key terms from search results (left column of screenshot).
(Credit: Screenshot by Rafe Needleman/CNET)
Yebol isn't ready for prime time yet, though. Long search queries (like my Bing test, "recipe for halibut and cilantro") gave unsatisfying results. And I wasn't happy with the way Yebol presented its top sites for each search result: It just showed the site names, with no context.
Yebol shows us that semantic search is not dead, and that it can make results richer than they are today. Yebol may not be the site that brings that to us. But I would not be surprised if Yin is one of the people who does.
Yebol should go public later in June or in early July.
Other than economic, the big benefits of Wowd, CEO Mark Drummond and CTO Boris Agapiev told me, is that its index is fed by what people are actually looking at, not just what Web sites are linking to; and that it sees all the Web pages that users visit, most importantly those that can only be accessed by entering data into a form (the so-called "deep Web").
There are obvious privacy implications with this. So Wowd does not index sites where it sees personal data, like bank or healh-care pages, and users can easily exclude from the index sites they're visiting that are not automatically excluded.
When a user sends a query into Wowd, it's sent out to other Wowd users' comptuer that are close-by on the Net. They each contibute results back to the query where it displays. It's like the BitTorrent of search.
The service helps you discover what's hot on the Web, since it knows what people are linking to in close to real time. It also can segment search results by social proximity (giving people different results based on what their friends have clicked on), but that's not built into the launch product.
Challenges include the privacy issue, as well as the need to get users to install resident software on their computers for this to work. Drummond did say that the search engine will work well even if only a small proportion of users are running the resident software. And, obviously, it's hard math to make this work with good performance and reliability, and at scale.
The idea of replacing the search data center with a cloud of users' computers is intriguing, but I'm extremely skeptical that this will work: That it will provide noticable value over and above Google, not to mention work technically. On the other hand, there's so much money in search that you don't need to gain a double-digit percentage of the market to have a winner on your hand, especially if your cost of providing service is negligible, as it should be for Wowd.
The service in private beta now, but I should have access codes for it shortly. I'll post when I do.
The claims: Yebol being built using new neural network technology. Founder and CEO Honfeng Yin says his algorithms are related to the "associative memory model," and that his goal is to convert information to knowledge (I've heard that before). Compared to Powerset's semantic search, which Microsoft bought, Yin says that Yebol's technology is far more scalable, and that his index already has one billion pages in it. The service is currently running an Amazon's on-demand cloud services.
Enough talk. How is it? As a tech demo, great. Like Bing, Yebol displays associated keywords or search terms alongside search results. Unlike Bing, instead of just displaying categories (which is useful in itself), Yebol displays related words that I found surprisingly useful. For example, in a search for "Google," Yebol displayed, among other things, a list of Google employees that included not just the names in the news but other important people who are or once were affiliated with Google. A search for "Top Gear" displayed a list of sport cars that have been reviewed on the show. (There are other search engines that pick out information from results to categorize them; see Clusty, for example.)
Yebol does a great job extracting and categorizing key terms from search results (left column of screenshot).
(Credit: Screenshot by Rafe Needleman/CNET)
Yebol isn't ready for prime time yet, though. Long search queries (like my Bing test, "recipe for halibut and cilantro") gave unsatisfying results. And I wasn't happy with the way Yebol presented its top sites for each search result: It just showed the site names, with no context.
Yebol shows us that semantic search is not dead, and that it can make results richer than they are today. Yebol may not be the site that brings that to us. But I would not be surprised if Yin is one of the people who does.
Yebol should go public later in June or in early July.
About Author
Content writer
Article Source: http://www.1888articles.com/author-amarpreet97-12539.html
No comments:
Post a Comment