Automated electronic mail assistant

Multiple text parts of a body of an incoming e-mail are determined. For each text part, based on natural language processing, a topic category and a feature vector representation (FVR) are determined. For two text parts comprising a common topic category, a similarity value based on their FVR is calculated, and in case the similarity value fulfills a condition, the two text parts are aggregated into one aggregated text part. For each text part: a reference document is selected based on a FVR of the text part; a recipient is obtained based on the corresponding topic category; and an e-mail comprising the text part, and each selected reference document or a pointer thereto, is sent to the obtained recipient. Via aggregation of text parts based on FVR, computational resources for document retrieval based on FVR is reduced, as well as the number of outgoing e-mails.