The San Francisco New Tech Meetup is on tonight at CNET headquarters. There are four companies presenting, two of which I have covered before: MerchantCircle and MyThings. That leaves two newbies, both interesting plays that lack only for business models.
Tellfriends is a nicely designed site for social recommendations. You can recommend a favorite local bar or restaurant to the crowd. Other people can vote your recommendation up or down, and you get standing in the community when you make good recommendations. Sounds like Digg, no? Or maybe Yelp? It's a little of both, resulting in a social-bookmarking site that's focused on local advice. Nice idea, I suppose, but maintaining the site's focus on local content is going to take some work. Also, while the site does express a somewhat different idea from other services, it's sandwiched into an overcrowded space of recommendation engines and social-bookmarking sites, and it's not yet different enough to stand out.
Readware is a 15-year-old company with a new CEO, Pierre Wolff, and a fascinating technology: a tool to perform contextual analysis of text. For example, it can tell if a story is about politics even if the word "politics" isn't in the story. And it can tell what kind of story it is: an explanation, a news story, a story about a conflict, a legal issue, and so on. The Readware engine can be used to search for stories based on concepts instead of keywords, or it can be used to assign keywords or tags to stories for automatic indexing. Both of these are powerful capabilities. Think of a search engine that's smarter than Google (a lot smarter), or a Wikipedia that automatically cross-links its articles based on concepts not just words. (See also Powerset, WordNet, Freebase, and the legendary Cyc project).
The tidbit I dig about Readware is that its structure of ideas and concepts is based on the linguistics of Classical Arabic. That's right, the ancient civilization that gave us algebra also developed a language complete enough to form the basis for a system of automatically parsing modern human thought.
Since everything that uses Readware runs through this conversion to Classical Arabic before being kicked back out in a modern language, concepts that you search for in one language can be found in another: If you search for "politics" using English keywords, you can also find French articles. It takes the company four to eight months to add languages to its system.
The problem with Readware? It lacks a sustainable business model. Though the company will soon announce a partnership with a major RSS company, it needs a product--either for enterprises or individuals--that it can sell off the shelf. The new CEO hopes to find inspiration or partners at tonight's New Tech Meetup.