US Patent 10963784 Identifying portions of electronic communication documents using machine vision

In a computer-implemented method, an artificial neural network is trained to identify conversation segments, and/or segment portions, within electronic communication documents (e.g., emails). An input layer of the neural network includes input parameters corresponding to different characteristics of text-based content. A first electronic communication document is received, and its text-based content is processed using the trained neural network to generate one or more position indicators for the document. The position indicators include one or more segment indicators denoting positions of one or more conversation segments within the document, and/or one or more segment portion indicators denoting positions of one or more portions of one or more conversation segments within the document. An ordered relationship between the first electronic communication document and one or more other electronic communication documents is determined using the position indicators.