Since its launch 18 months ago, Attivio, Inc. (www.attivio.com) has touted its fresh approach to unified information access. The company's Active Intelligence Engine (AIE) combines unstructured content-such as emails, documents, images, and webpages-with structured business data and makes it all searchable. The just-released AIE v1.5 introduces optional modules for Classification, Sentiment, and OCR, as well as some significant new features including high-availability ingestion and querying, guaranteed indexing, and next-generation security. The company has been scoring some significant customer wins for AIE of late, and the enhancements in this version represent what CTO Sid Probstein calls "customer driven innovation."
The enhanced security in AIE 1.5 might be the most significant addition. Probstein says, "It's common to talk about ‘next-generation' in this industry, but now I'm more convinced this [enhanced security] is it." He says AIE's new model brings the database approach of using separate tables into the search world. AIE 1.5 eliminates the need for an enterprise search engine plus a relational database to manage complex document-level permissioning by storing content, groups, and users as separate tables in its universal index. Probstein says this allows significant cost savings and near-real-time access control updates because there is no need to reindex as changes occur.
"In information management, there is a delicate and difficult balance between providing users with access to the information they need while still protecting the integrity of the data. Organizations are often faced with a dual challenge: they need to empower users to have immediate access to the most relevant information, but also need to ensure only those users with certain privileges or permissions are allowed to access it," says Mark Beyer, research VP at Gartner, Inc.
Brian Babineau, senior analyst for Enterprise Strategy Group (www.enterprisestrategygroup.com), says, "The ‘administrative' as opposed to the ‘feature' enhancements are probably the most noteworthy additions to the solution. The multi-node configuration support, the backup and restore utility, and the security model are not the sexiest announcements, but these make it much easier for large enterprises to run and rely on this particular solution. And, our research shows that the easiest way for any technology buyer to justify an investment is to prove operational cost savings-these capabilities do that. Any application that users need to do their jobs-of which enterprise is usually one of-must be available at all times, and with privacy regulations in abundance, security is not a luxury. If I were to single one of these out, it would be the security model, which enables customers to easily update access permissions in real-time without requiring a re-index operation-a process that can take a very long time."
The new modules are designed to provide users with more relevant results. AIE previously used rules-based classification. Probstein says the new Classification module in AIE 1.5 uses machine learning. It scored better than 87% accuracy in a recent test using the 20 Newsgroups data set (http://people.csail.mit.edu/jrennie/20Newsgroups). Users create new models by providing a list of category labels and a sample set of data for each category (at least 100 average-sized documents for each category). This takes only minutes, according to Probstein.
The Sentiment module includes a set of domain-specific sentiment models (general purpose, ecommerce, entertainment) along with a Model Builder utility for training and creating new sentiment models. Users can use the Sentiment module in conjunction with the Classification module-classifying documents to different domains and then applying the domain-specific sentiment model.
"The sheer volume of structured data and unstructured content that exists in the average enterprise today presents a significant challenge for organizations to index, organize, and route the information they have in house, let alone address the influx of new information on a daily basis," says Probstein. "With AIE's Classification and Sentiment Modules, content is automatically classified by category and measured by sentiment, enabling our customers to identify and process the relevant information they need, when they need it."
Babineau says, "The Classification Module will make it easy for people to find information that they may not have known existed, and the Sentiment Module-probably the ‘coolest' new module in my opinion-will help customers immediately see how internal and external constituents feel about a particular topic. This is vital given today's social media world where new products, movies, services, etc. are reviewed in real time."
The optional OCR module uses OEM technology licensed by Attivio. It ingests text from scanned documents and images, including hand-written text. Supported formats include PDF, TIFF, and JPEG. It can also be used to extract tabular data from supported file formats (e.g., tables found in PDFs). Probstein stresses that this enables searchable data, not just searchable documents.
IntraLinks, Inc. (www.intralinks.com) currently has one of the most sophisticated implementations of AIE-the system handles some 20,000 searches a day. "Our customers rely on our technology to help them locate, organize, and share some of their most confidential and important intellectual property," says Fahim Siddiqui, EVP of product development and operations at IntraLinks. "Not only is it important to them to know that the information exchange is secure, but they need to be able to analyze all the documents at the most granular level. Attivio AIE v1.5 allows our users to easily locate information and alerts users when relevant documents have been added."
Nick Patience, research director for information management at The 451 Group (www.the451group.com), says the improvements in AIE 1.5, especially the Sentiment module and the enhanced security, are "really exciting propositions." He says the company's recent customer wins are impressive. He adds, "I think these guys are not only survivors [in a tough industry], but they are on the forefront of a shift away from reliance on relational databases and towards search-based applications."
Attivio is also working on a new module that extends what it had been doing with the SQL JOIN command (see the NewsBreak about the launch of AIE 1.2). The new AIE SQL module will let customers write and execute SQL queries against AIE directly or from existing enterprise BI tools and applications. The company says it is now accepting applications for participation in the beta program for its AIE SQL engine, which launches in September. Probstein expects general release sometime in early 2010.
