|
|
|
|
|
|
|
|
|
  
|
AISR: Automated Indexing, Search and Retrieval We have developed a content-based Automated Indexing, Search and Retrieval (AISR) engine tailored for meeting the challenges of unstructured application environments, such as on-line chat scenarios and multi-channel databases with sparse information contents. Our system captures the contextual information of the unstructured (as well as any structured or semi-structured) data through the use of a new and innovative meta-data concept representation technique that utilizes Attributed Relational Graphs (ARG). The system does not rely on pre-defined concepts; it automatically defines and extracts sparse information contents and multi-feature concepts from the user query (for information discovery and retrieval applications) or from the chat data itself (for search, archival and retrieval in multi-channel chat scenarios.) These specific features of our new and innovative approach render our system design to be the most suited to deal with the challenges of finding hidden or concealed concepts in difficult scenarios. Our system utilizes inexact graph matching algorithms for robust and efficient indexing, search and retrieval. Our system has been developed as a prototype demonstration and successfully tested for information discovery in free-text databases and chat scenarios. Our demonstration delivers very high recall and precision in real-time even when presented with queries that contain very few, sparse, vague, abbreviated, or miss-spelled information. Our algorithms expand the ARG technology into a multi-level hierarchical scheme for processing sparse pieces of information and capturing the relationships among them to formulate global concepts. This basic principal of our ARG-based approach render our system much more effective in connecting the dots in chat and other unstructured, challenging application environments with sparse information contents. |
|
© 2007 ARGTEC Incorporated. All Rights Reserved. Legal Notice |