US Patent 10346524 Position-dependent word salience estimation

Methods, systems, and computer-readable storage media for receiving two or more electronic documents, each electronic document including text data, a second electronic document including a link to a first electronic document, processing word representations of words of the first electronic document using a first encoder to provide first output and a context vector, processing text data of the second electronic document and the context vector using a first decoder to provide second output, determining, by an attention mechanism, a plurality of weights for each word in the text data of the first electronic document based on the first output, and the second output, and providing a word salience value for each word, a word salience value comprising a sum of weights of a respective word.