Chat Relevance and Targeting (CHART)

ABSTRACT: Technology to aid in the understanding and exploitation of text has not kept pace with the technology used to generate, record, transmit and store the data. The mismatch between the fire hose of text production and the soda straw of text consumption is particularly severe in systems that rely on chat. The proposed system, named CHAt Relevance and Targeting (CHART), ensures that chat-based data is delivered to the users who need it by passively capturing chat content and metadata and using a dynamically updated relevance model to pass it to the relevant users. CHAT"s relevance modeling is based on a probabilistic model of text content and user interests that is developed through a fully unsupervised process no labeled data is required, and no assumptions about language use or data relevance are needed. CHART"s relevance model is also used to identify external, non-chat text data that is relevant to the material discussed within the chat domain. CHART can retrieve relevant data in realtime in support of current operations or in an after-action or forensic mode for training and event reconstruction. BENEFIT: The proposed system, named CHART (CHAt Relevance and Targeting) dynamically builds and maintains a relevance model that maps topics under discussion in the chat domain with specific user"s interests and job function. CHART uses this relevance model to automate the delivery of relevant chat content to the users that need it. By targeting the right information to the right users, CHART provides Warfighters with a more accurate and up-to-date situational picture with less manual discover of content required. CHART is also able to identify external, non-chat data that is relevant to the emerging topics of discussion in the chat domain, and can supply additional amplifying and clarifying to Warfighters in realtime.