URSAUnicode Retrieval System Architecture |
| Overview |
|
URSA is Computing Research Lab's Tipster Phase III research and development effort to make text processing and information retrieval transparent to languages. URSA combines unicode display technology developed at CRL with translingual information retrieval, multilingual collection visualization and document management, with special emphasis on design principles that have been validated by examining the analyst in real-world scenarios. |
| Tipster Phase III |
|
Tipster is DARPA's design and research effort to unify the detection and language processing capabilities of a diverse range of research entities into a single, plug-and-play architecture. The Tipster Document Manager is the central feature of the Tipster design. In Tipster, there are collections, documents, attributes and annotations. Collections can contain documents and collection attributes, while documents can have annotations and attributes. Other Tipster features include support for detection and information extraction technologies. Because the data in a Tipster document is simply a byte stream, a document can contain Unicode text, video, audio, or any other imaginable data type with equal facility. It is up to the application that makes use of the documents to interpret the data correctly. |
| Unicode and URSA |
|
URSA combines the latest advances in information retrieval with the coherent Unicode text model to make language-transparent IR a reality. Ongoing development is focusing on integrating Unicode detection technology with the Tipster architecture. The model we are developing utilizes annotations on the documents to describe the text for indexing. External document annotators can produce segmentations for Oriental languages, or stemmed word markup for Western languages, which are then interpreted by the URSA engine and indexed for later retrieval. The URSA engine will be fully conversant in Tipster detection needs, including complex query expressions and natural language queries. |
| Papers and Presentations |
|
Check-out some of our papers or download presentation slides on Cross-language Text Retrieval, URSA and related efforts: Download the FREE Adobe Acrobat Reader for PDF files here
New |
| Demo |
| The MUNDIAL system was an
early proof of concept for Translingual or Cross-Language Text
Retrieval. The new MUNDIAL system is substantially more robust and
now provides an interface to more search engines. MUNDIAL allows
searches of documents on the World Wide Web using English language
queries that are translated into ten other languages. Click here to try the MUNDIAL demo. Click here to try J24 The URSA retrieval engine is now being used in the J24 retrieval system for evaluating document visualization in interactive retrieval. J24 presents thumbnail views of documents to show keyterm distributions. Click here to try Arctos The URSA retrieval engine is now being used in conjunction with URSA's cross-language technology for the interactive creation of queries, multilingual document retrieval and translation. The queries can be sent to WWW search engines or to J24 for interactive document visualization and translation. Click here to try Keizai This is the next iteration of a WEB based interface for Cross-language text retreival. It helps you construct text queries in French, Italian, German, Spanish, Chinese, Japanese and Korean. Named entitiy extraction in these languages and translation capabilities are used to present the text results |
| Contacts |
| For more information on the URSA project, Unicode detection and translingual information retrieval or user-centered detection system design, please contact principle investigators Bill Ogden or Mark Davis. |