Tonight I’m at a huge PAWS meetup with well over a 100 people (standing room only) here at the Microsoft campus in Mountain View (known to insiders as "SVC"). These are my notes from the presentation, so apologies if they look like a Virginia Woolf novel.
Uptake (Gene McKenna) - Has a vocabulary of travel related terms, sentiment analysis based on types of classes (positive/negative terms for romantic), an ontology (e.g., what does it mean to be kid-friendly?). Then run reviews through their engine and extract interesting facts. Indexed over a million products, 20 million opinions, etc. For hotels, they show snippets of reviews that will be relevant, rather than have the user go through hundreds or thousands of reviews. For example, a search for family hotels in San Francisco gives you a set of family-friendly hotels with highlights from the reviews.
Expert Systems (Walter Pezzini) - FY08 $13.5M, profitable and 100+ employees(!!) and based in Italy. Demo of Cogito Answers, which is answers NL questions from unstructred data (Wikipedia), some structure (eBay), or structured data (American Football DB). This was developed for a customer and is not live on the Web. First question "who was OJ Simpson's wife?" got an answer from both Wikipedia. Second question "do you have a Jets keychain for under 105?" gets results from eBay. Third question: "for which teams did Farve play" went to the American Football DB. The demo seems to show that Expert Systems can figure out which source to get the data from, but I would have been even more impressed if he asked a question that could have been answered from multiple sources to see how it collated the data. Someone from the crowd: "Do you need to know the schema of the database?" Answer: some tools for customers to map tables to concepts.
Freebase (Jason Douglas) - an (awesome) open, editable, free database with about 5M topics in the graph with their relationships. Using the Metaweb Query Language (MQL), did a mashup with last.fm data + twitter follower data, to determine the most popular band member's name. He should totally build a Web 2.5 mashup site and start his own company.
zAgile (Andrew Lampitt) - Why semantic Web in the enterprise: heterogeneous, siloed apps + no shared vocabularies + no resuse across apps and processes. Demonstrates how Wikidsmart (which is the best product name I've seen this millenium) can help to integrate across many different software development apps (version control, bug tracking, project managment, etc.) in a nice Wiki interface. You can try out their sandbox once you log in.
Zemanta (Bostjan Spetic aka Bos aka Zemanta co-founder) - takes a few lines of text, recognizes what's important, pulls back pictures relevant to the text, suggests links from multiple sources, and other pieces of information. It's easy, just copy some text and try Zemanta here. Think "research assistant" or "content enhancement." Obviously sits on a lot of technology to do this, but tries to hide the complexity from the user: "so you and your mother can use it." All of the processing is done on Zemanta's servers. You can use Zemanta by installing it as a plug-in and it will work in your Gmail, Yahoo Mail, Wordpress blog, Typepad blog, etc. Also has an API that allows you to do Semantic Mashups.
Noovo (Andrej Babergoj) - Noovo is a recommendation engine in Beta that suggests content and tell you why it recommended it and who recommended it. You can give items a thumbs up or down. (This was a real demo: one of his friends shared an NC-17 picture with him. Friggin' awesome.) Allows you to join groups, connect with your social networks, etc. - things you'd expect in a Web 2.0 recommendation engine. Social graph (who are your friends/colleagues/family) + Activity graph (what you do) + Personal Interests (what you like). Someone asked a question (which was also on my mind) about how they compare to Twine. I didn't completely understand the answer =)
Microsoft Research (Bora Beran) - showing off some work done at MSR, specifically in the Technical Computing group. Showed off a few plug-ins for Microsoft Word. First, Chem4Word allows you to all sorts of features to add chemicals to Word Docs. For example, he typed in "water" and got a list of synonyms and was able to insert an dihydrogen monoxide molecule into the document. Zentity allows you to manage your scholarly work (not using OWL or RDF, but an identity data model) for things like citations and gives you an advanced way to search through the data. SciScope is (one of) Bora's project, which is an interatcive map with tons of environmental data. Takes data from multiple data sources, but totally transparent to the user.
One last special mention: I met Guha Jayachandran, who did Folding@Home as his PhD project at Stanford. He's got a new search engine called Cruxlux which analyzes perspective. Very excited to see that.
Let me know if there are any errors in my trascriptions. And follow me on Twitter!