|
|
Mission: Pioneering ARG algorithms
Many of the indexing and retrieval techniques used in current systems are based
mainly on limited sets of user-provided keywords/concepts and simple algorithms or
purely statistical techniques that take into consideration few independent features.
Most existing search and retrieval engines lack the ability to capture and process
multi-feature concepts and relationships between the information contents in a robust
and efficient manner. Most of those techniques also lack the ability to efficiently
incorporate new concepts and to adapt to changes in the operational environments and
requirements.
- We provide our customers with the best algorithm designs and prototypes.
- We provide the end-user with definitive "state-of-the-art" performance at cost-effective prices.
- We team with system integrators to secure timely infusion of the best algorithms and technology for mission critical tasks.
|
Our system utilizes a hierarchical ARG representation as a unified and novel
representation scheme capable of capturing multi-feature concepts from multi-source
data. It captures both local and global concepts as a multi-layer graph composed of
attributed nodes and branches. At the bottom layer, the set of nodes in the ARG
represents the set of symbolic and numeric localized features in the messages
(e.g., n-grams, abbreviations, aliases, etc.) The set of branches, i.e., the links
between nodes, represents relationships among those features. For more information
about Attributed Relational Graphs (ARGs) please go to:
http://citeseer.ist.psu.edu/context/32271/0
|
