Breakthrough R & D

TextWise has a long history of working with the government to design and develop sophisticated content-centric applications using Natural Language Processing (NLP) and Machine Learning (ML) techniques. Listed below are Active and Completed government-funded initiatives:

MEDLINK

MEDLINK is a medical information retrieval system that makes finding information faster and more efficient. It provides easy access to both alternative and traditional medical literature for health professionals. The MEDLINK system uses advanced NLP techniques to process contents and queries, and to enhance search results. It is comprised of 4 components: entity extraction, document classification, text structure, and search/query. The basic technology and methodology behind MEDLINK will be applied to the financial vertical in future releases. Currently, it is funded by a research grant from NIH (National Institute of Health).

CINDOR (Conceptual INterlingua DOcument Retrieval)

CINDOR allows a user to ask a query in one language and simultaneously retrieve documents in several languages. CINDOR's cross-language search solution uses our proprietary, language-neutral Conceptual Interlingua to move beyond keyword searching, with natural language questioning and automatic query term expansion to further improve your cross-language search results. Contract awarded through an intelligence agency within the Federal government.

DOCKET (Data Mining, Monitoring and Filtering Tools for Electronic Mail)

DOCKET is a natural-language based information extraction and monitoring system for email. DOCKET automatically extracts and visualizes entities such as people, places, and things and their relationship to each other. The system is also designed to automatically sort email by user-defined categories with the option to visualize the results. With the use of sophisticated information retrieval technology, DOCKET offers the ability to search a knowledge base of email using natural language queries. Contract awarded through DARPA and the National Science Foundation (NSF).

DR-LINK (Document Retrieval Using LINguistic Knowledge)

DR-LINK is an advanced natural language query system that allows a user to ask a question in "plain English" and get back a list of relevant documents. DR-LINK provides automatic alerts, data visualization, relevance feedback and document classification. Simple to use and powerful to search with, DR-LINK defines how search systems should work. Contract awarded through DARPA.

EMMA (Evolving and Messaging Decision-Making Agents)

EMMA is a unified framework for agent learning and agent collaboration. Within this framework, we have developed a decision support system based on collaborative, autonomous, intelligent agents capable of acting upon both automatically perceived needs and direct orders from the user. Individual agents learn from previous cases, users, and other agents. They may consult dynamic knowledge bases in the particular application domain and apply reasoning techniques. Contract awarded through Air Force Research Labs.

EVA (EVolving Intelligent Text-based Agents for Geospatial Information)

EVA is a multi-agent system which searches the World Wide Web for multimedia information and uses machine learning techniques to learn user profiles and information seeking behavior. Individual agents use neural networks for local searching and learning. Genetic algorithms are used to facilitate the evolution of agents on a global scale. NLP technology not only allows users to write sophisticated queries, but also allows the system to extract important information from the queries and the retrieved documents. Grant awarded through the National Imager & Mapping Agency (NIMA).

KNOW-IT (KNOWledge Base Information Tools)

KNOW-IT delivers a new way to explore information and discover knowledge using innovative tools based on in-depth semantic analysis of textual information. The system can take raw text, such as a newspaper article, as input and create a knowledge base of concepts and relations automatically. These concepts and relations can then be browsed to locate desired information. Contract awarded through DARPA.

TIPSTER SUMMARIZATION Phase III

The TextWise Tipster Summarization project developed indicative, query dependent summaries for both single and multiple documents. We used metadata, phrases, and sometimes representative paragraphs in our multiple document summaries. Our multiple document summaries included thumbnail sketches of documents returned in response to a query, and topic overviews, supplying more detailed multiple document summary. Contract awarded through DARPA.