Factery Labs Makes Other Search Engines Look Incomplete
Most text excerpts that appear on search results pages aren't very useful. Imagine if instead your search engine showed a list of clear sentences summarizing the contents of each link on that search result page. That's what a new service called Factery Labs aims to provide for any service that utilizes the API it's launching today.
You give Factery a list of links and a keyword and it will build an index of all the facts asserted in those links about your topic of interest, delivered in XML or JSON format. The service can run on top of a search engine but could also be used in any number of other ways. I've been feeling unsatisfied with other search engines all day since seeing a Factery demo Monday morning.
After building that "fact index," Factery ranks the links submitted by the quality and density of facts related to query on the page. Compare the search results page on Google News for "Paul Allen" to the information that Factery extracts from links being shared on Twitter about Paul Allen. The Google News page tells you nothing, except that Paul Allen has cancer - over and over again.
Compare that with the Factery results page - I don't even need to click through if I don't want to, I feel like I got a great overview of the story just from my search page. Perhaps that's a problem - for a publishing industry that already says it's scared of search engines - but as a reader it sure isn't my problem, it's great. Why would I want Google News to tell me where I can go to find information if someone else will just give me the information?
The company's test demo searches Twitter and Yahoo Boss - neither search is as exciting as I'd hoped 100% of the time, but it's often remarkably good. Factery is also testing an interesting integration with Silverlight stream reader Sobees, in which linked pages from Twitter or Facebook are annotated with automatically extracted highlights via Factery.
I expect a whole lot of companies are going to at least try this API out and I'm excited to see the results.
How This is Unlike Other Real-Time Search Services
Factery is talking a lot about its ability to analyze links shared over Twitter, but that's probably just because Twitter is easy for people to understand. The fact is, the service can perform on-demand analysis of text behind any set of links. That's what differentiates it from other real-time search engines like OneRiot, which also analyzes the text of pages linked to on networks like Twitter and offers an API to display real-time search results on other sites. Competitor Collecta analyzes Twitter streams in real time and offers an XMPP API to push new search results live to any page.
Factery is a different kind of animal, though. It's more like a smart search inside any other search. It doesn't even have to be search, though. The company talks a lot about how they make mobile reading more efficient by pulling the salient information up to the surface of a page, instead of requiring mobile readers to load multiple pages.
I thought of five or six different ways I'd like to use it just while talking to the company on the phone. (I'm not going to share those here, either. I think some could offer an important competitive advantage.)
I'd Love to See This Work Everywhere
Yesterday I was testing a new Android app from the Sunlight Foundation that lets you track members of congress. One tab in the app is a search for your congressperson in the news. Unfortunately, the page excerpts give no indication why the politician you searched for appeared in that news story - just that their name did, somewhere. That search is powered by a Yahoo API, probably BOSS, but it's not any fun to use at all. How unsatisfying, I thought, when I could have a list of key facts concerning my search query in the list of links that the search brought back. But that was yesterday, and Factery is just launching today.
The possibilities are truly endless. That's probably why Ron Conway, one of the leading investors in the real-time economy, joined others in investing in the company. With $1.2 million in the bank, Factery is a modest developer play with a whole lot of potential.
Give Factery's API a try and let us know what you think. It's free to use; the company says it may start inserting "sponsored facts" (isn't that an interesting phrase) into results later but things like business model and to a lesser degree de-duplication are still works in progress. I sure do love this idea.
Share
Facebook
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
.....as long as content on the web is tangled in a display mark up language "html", we will forever need rocket Science algorithms to find the "facts" and to give "relevancy" to search results.....We are so used to the awful results that are given from search engines that have to make sense of content that to the computer has very little meaning; that we think that the results from Factery Labs are great....when in fact the results are only a slight improvement over existing systems owned by companies that despite their multi million dollar R and D budgets can today only at best give us a list of possible relevant ressults to our queries....a list of possible relevant results....that is where we are.....Will we continue to chase the myth of computer understanding for another 20 years and in the end still only have a list of possible relevant results.....Maybe it is time to re examine our notions of search as well as the creation of content.....Garbage in.....Garbage out
"...I don't even need to click through if I don't want to, I feel like I got a great overview of the story just from my search page..."
?!?!?!?!?
You think the running joke(s) about Twitter's lack of a business model are bad? Wait until the VCs get wind of the above.
Atoall.com's new search engine which is know as best search engine is now available for easy web search. Write in the address bar three button twice which are together on the computer keyboard (angle) e.g. ppooii.com. This Internet search engine has unique technology. You can contact us for also search engine promotion.
I'm not sure I get the stream-of-consciousness blurt from "William", but I guess I’ll take “slightly better”.. :-) Actually I hope it’s a lot more than that. Internal analysis here shows that the web contains a very large pool of usable facts. “Facts” is this case means “simple declarative sentences”. The web is a huge structured data set consisting of grammatical English sentences.
Duck Duck Go does this for regular search results with our new "Duck It" view, so you can find out information about a topic with zero clicks. For example: http://duckduckgo.com/?q=readwriteweb&v=d
Startling in the way it makes me say to myself, duh... Why didn't anyone think of this sooner?
I love the idea of a search-engine being smart enough to scan / aggregate all text on a search term into an intelligent presentation that is better than anything an individual blog, newspaper, government agency, or corporation could produce.
And imagine what could be done when it comes to writing policy or drafting legislation?
It makes perfect sense that this is the future of search! And it'd be charming if somehow Google totally blew it and somehow forgot to prepare for this new era...
Looking further beyond imagine what would happen if these algorithms could one day be easily tweaked directly by the user via a powerful GUI, as well as tracking user preference / interests?
If a system like this really worked well it's likely we'd no longer have to ever be writers again! Instead the great writers of the future will merely be the ones who have the most creative tweaks built into their personal search presentation apps.