|
|
|
|
|
Insightful Corporation—based in Seattle, Washington—has 120 employees and annual revenues of $22.3 million, representing an 18 percent increase over revenues for the previous year. Insightful also has offices in New York City, North Carolina, France, Switzerland, and the United Kingdom, with distributors around the world.
Insightful develops and delivers software and solutions for predictive analytics that have enabled thousands of companies to discern the patterns, trends, and relationships hidden in the data they collect. Insightful solutions are used by companies and organizations where analytics are critical to success, including financial services, pharmaceuticals, biotechnology, telecommunications, energy, and manufacturing, as well as research institutions and military and nonmilitary government organizations.
The technology developed by Insightful Corporation (Insightful) under DARPA’s SBIR helps analysts work through documentation more quickly and efficiently than traditional technologies, such as keyword search or simple entity extraction software. The program comprises two distinct parts: (1) a new algorithm for latent semantic analysis, and (2) deep parsing technology for relationship extraction from unstructured information.
Latent semantic analysis is a promising technique with applications to enhanced keyword retrieval and cross-lingual retrieval. A number of government laboratories have built prototypes based on the ’90s-era Bell Labs Latent Semantic Indexing (LSI) algorithm. Insightful’s development of a new algorithm, labeled Latent Semantic Regression (LSR), is scalable to large data sets and forms the basis of the company’s InFact® product line. InFact® is a production quality, end-toend deployable software system that performs relationship extraction, text mining, and search.
According to Giovanni Marchisio, Vice President of Engineering-Text Analysis, LSR represents a major step forward for the company. “The creative leap involved a deep understanding of computational techniques. We realized that the LSI algorithm, which everyone else uses to perform latent semantic analysis, is only a special solution of a more general computational problem. Our prototype replaced the core algorithm for LSI with the best of many computational alternatives that are available to solve the more general problem, resulting in a substantial increase in scalability and in an improvement in speed of several orders of magnitude.”
The deep parsing effort initiated under DARPA’s SBIR program produced the first prototype of a search engine that can understand and retrieve relationships facts and trends. Intelligence analysts and pharmaceutical researchers have requirements for search functionality that goes well beyond simple keyword search and involves an understanding of relationships and facts.
InFact® understands far more than keyword searches—it understands facts and actions. In a search for ‘blackhawk,’ a type of military helicopter, InFact® generates an aggregated list of all combat situations where the Blackhawk helicopter was deployed in the recent past, including all locations and military units that employ it, along with information about acquisition and maintenance costs and much more.
Insightful’s technology is finding its way into a number of military applications, including use as an advanced search engine and as an information extraction tool. In information extraction, InFact® can process millions of heterogeneous document sources and automatically populate databases with information it has understood and extracted across all documents it has read.
The company currently has a contract for information extraction with the U.S. Army Corps of Engineers. The effort focuses on automating the labor-intensive process of tagging geographic information in text files, interactively editing the tagged files, and displaying the tagged text and associated map. The U.S. Army estimates that 80 percent of its data contains spatial reference information in the form of addresses, place names, or coordinates. In order to support spatial analysis of text sources, our approach requires structured text files with place name tags containing or referencing coordinate information. While tools exist for automatically identifying place names in text, our efforts are focused on significantly improving current capabilities within an end-to-end system. By reliably automating these processes, InFact® has the potential to save the Corps of Engineers and other government agencies millions of dollars per year.
The company’s technology is also being used by: • The Air Force Research Labs, where researchers are developing a system for rapid diagnosis and improved screening of personnel that have been exposed to toxic agents in the field. • Major pharmaceutical companies to generate search and retrieve relationship-based summary reports of millions of biomedical journal articles and reports. • A major food manufacturer to extract facts from consumer reports.
The funding received by Insightful under this DARPA SBIR has had a significant and positive impact on the company’s research efforts. Besides being Insightful’s initial key contract in this research area, the DARPA SBIR offset approximately 25 percent of the company’s development costs during the period 2000-2002. The InFact® product line has contributed to the company’s revenues—approximately $900,000 in 2004, and $1.5 million in 2005.
After the successful debut of InFact®, Insightful created a separate business unit for text analysis and search, and hired additional business development and sales personnel. The DARPA SBIR also led to three U.S. Patents: 6,510, 406; 6,757,646; and 6,862,710.
Jeannette Israelson
Insightful Corporation
1700 Westlake Ave N
Suite 500
Seattle, WA 98109
Phone: (206) 283-8802 ext: 262
Email: jai@insightful.com