Radar Networks has finally taken the wraps of its stealthy semantic Web platform. I met with company CEO Nova Spivack earlier this week to get the lowdown on Twine, the first application of the company’s technology.
“Twine is a new service for knowledge networking, sharing, organizing and in finding information from people you trust,” Spivack told me. “Unlike a social network that is about who you know, Twine is more about what you know. The social graph connects people and twine connects everything. It’s a semantic graph--connecting people, places, companies, products, Web pages, videos, photos and turning it into semantic Web content."
"Twine uses natural language processing and statistical, link and graph analysis, as well as Web crawling, data mining and machine learning to figure out what information users put into the system is about, what it means and what is should be related to. Then Twine connects it and organizes it for you automatically,” he said.
“Web 2.0, as Tim O’Reilly says, is all about collective intelligence. Twine does it in a smarter way—it’s Web 2.0 with a brain, which is sort of what web 3.0 is,” Spivack said. "From an industry and technical perspective Twine is the first mainstream example of how the semantic Web could manifest for end user, he claimed.
On the surface, Twine sound similar to many other products that collect things and organize them is some way, such as AdaptiveBlue's BlueOrganizer, which uses more of a top-down approach to creating structure, semantically rich content. BlueOrganizer knows about basic concepts like books, music, wine and travel destinations, but eschews using RDF, OWL and other semantic Web standards. In fact, many Web services describe themselves as semantic Web solutions, but they are mostly using semantic analysis on content to improve search relevancy that to create a massive scale data Web.
AdaptiveBlue CEO Alex Iskold isn't keen on bottom up semantic Web approaches, using RDF, OWL and other technologies. "I have a few measuring sticks," Iskold said. "In ten years, we haven’t gotten anywhere with it. We are very pragmatic. The utter complexity of RDF can confute even the brightest engineer. We deal with the nouns first and have simple metadata for books or movies," he said.?
(RDF breaks information down into subject, which identifies the thing the statement is about, such as a Web page; a predicate that identifies the property or characteristic of the subject; and an object, which identifies the value of the property.)
Twine, which is in invite-only beta now, is aimed at individuals or groups who see value in having smarter software that can provide deeper levels of organization and extend the notion of discovery.
“As Twine learns, it helps you to search better. Social search is based on the semantic graph. It ranks information in a couple of different ways—by relevance, time and relative to me, meaning the social distance from me in the semantic graph," Spivack said. "We also have a new way to rank information based probability analysis of the semantic graph and how you are connected to the information, showing you stuff that is likely to be from sources you trust or is things you would be interested in. Basically it combines social and semantic search—the more you put it, the more it learns about you.”
“As an individual Twine is useful, but it’s more valuable for teams. We are really targeting professionals and prosumers who are looking for knowledge management solutions across boundaries. People use Lotus Notes, SharePoint, wikis and primarily email in teams. Most legacy knowledge management is database driven and not designed to cross boundaries outside the corporate firewall. We think Twine will attract the change agents, and move inside companies just like instant messaging did," Spivack said. Twine will include access and permissions for protecting privacy and integrating with external systems.
Conceptually, Twine is similar to Facebook and other services that provide relevant feeds and notifications, in this case about all the content you have in the systems and what others have shared. Twine categorizes people, places, organizations and other “tags,” or concepts. Users can define new categories.
The interface also includes a tag cloud for quick reference. Radar Networks' platform has digested Wikipedia as part of its underlying structure, and has 300,000 concepts in the system. “Wikipedia has a certain amount of messiness but it keeps up with the world and culture nicely, Spivack said. The company is also looking at commercial data sets, such public and private company information, product catalogs and geographical data, he added.
Natural language search isn’t part of the Twine interface, but Spivack said his company might work with Powerset, which has developer a natural language search engine. Powerset is working along similar and complementary lines to Radar Networks. Its search engine semantically analyzes Web pages for meaning and concepts, and looks at other knowledge resources, such as Wikipedia and the New York Times, and then indexes the facts. Powerset has also formed Powerlabs, which is encouraging users to help build out the ontology for the search engine. Both Powerset is using Freebase, a kind of open public almanac that offers structured information on topics such as movies, music, people and locations, from the other well-funded, pedigree-rich semantic Web startup, Metaweb Technologies. Radar Networks may do the same.
Twine also provides visual cues. Text in orange is what Twine discovers on behalf of the user, and blue is used to show what the user input to the system.
In the future, Radar Networks may make an open source version of its software for lower-end usage, Spivack said. Twine will also add more depth to its algorithms, such as looking for related items that should be directly connected and connecting information to other information depending on who is looking. In addition, Twine could extend to the desktop with plug-ins for various applications, such as Facebook and salesforce.com, Spivack said.
The big vision is that we want all the information in one place and other applications and services can talk to Twine," Spivack said. "Twine is the cornerstone, like salesforce.com is to AppExchange. Project management or CRM applications and collaborative applications could be used in context of Twine in the future."
Radar Networks has raised about $9 million, including $5 million in a series A round, and has 27 people. Twine will be free and ad supported, with some limits on storage and the number of advanced features. A subscription-based premium content service is also in the works.
Radar Networks also has patents filed around semantic ads, which would provide targeting to individuals and groups in a more targeted way, Spivack said. "We will know a lot about you--what videos, travel destinations, hobbies you have. We can see these different things and how strongly they are connected together." Of course, he assured that everything on Twine, the user's data, is private and exportable with all the rich metadata. However, without Twine, the data won't be that useful.
Over the next several months, as real users test out the application, Radar Networks will learn whether Twine is a twisted, knotted ball lacking enough brain power or a true harbinger of the semantic Web.
For perspective on where Twine fits along the journey to the semantic Web as conceived by Tim Berners-Lee:
The Semantic Web will bring structure to the meaningful content of Web pages, creating an environment where software agents roaming from page to page can readily carry out sophisticated tasks for users. Such an agent coming to the clinic's Web page will know not just that the page has keywords such as "treatment, medicine, physical, therapy" (as might be encoded today) but also that Dr. Hartman works at this clinic on Mondays, Wednesdays and Fridays and that the script takes a date range in yyyy-mm-dd format and returns appointment times. And it will "know" all this without needing artificial intelligence on the scale of 2001's Hal or Star Wars's C-3PO. Instead these semantics were encoded into the Web page when the clinic's office manager (who never took Comp Sci 101) massaged it into shape using off-the-shelf software for writing Semantic Web pages along with resources listed on the Physical Therapy Association's site.
See also: Techmeme