Building a pragmatic action-item system

A system comprising:

• data processing hardware;
• and
• memory hardware in communication with the data processing hardware, the memory hardware storing instructions that when executed on the data processing hardware cause the data processing hardware to perform any of the operations comprising:
• obtaining a transcript comprising a plurality of dialog acts recorded during a multi-party communication session;
• extracting, using a first machine learning model, from the transcript, a plurality of extractive action items, each extractive action item associated with a task expected to be completed after the multi-party communication ends;
• generating an abstractive action item by:
• clustering a subset of the plurality of extractive action items into a coherent group associated with a same respective task;
• and
• generating, using a text generation model, a summary of the abstractive action item that concisely describes the same respective task based on lexical features of the clustered subset of the plurality of extractive action items;
• receiving task completion information comprising an individual responsible for completing the respective task associated with the clustered subset of the plurality of extractive action items and a due date for completing the respective task;
• presenting, in one or more graphical user interfaces, the summary of the abstractive action together with the received task completion information;
• receiving user feedback indicating whether or not the summary of the abstractive action item is accurate;
• and
• fine-tuning the text generation model based on the user feedback.