Message correlation in automated communication surveillance through singular value decomposition and word frequency association
Ryan Layfied, Latifur R. Khan, Bhavani M. Thuraisingham · 2005
Automated surveillance methods frequently rely on algorithms which detect the presence of suspicious keywords and topics within messages to properly flag suspicious content for review. However, subsequent messages based on the original may not carry the same characteristics that were initially detected. A correlation algorithm is necessary to find such messages and 'thread' a conversation together. Within this document, we propose such an algorithm and an experiment with which to test it.