Browse Prior Art Database

A System for Updating Agents Based on Offline Processing

IP.com Disclosure Number: IPCOM000244483D
Publication Date: 2015-Dec-16
Document File: 1 page(s) / 28K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a system for updating agents based on offline processing.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 100% of the total text.

Page 01 of 1

A System for Updating Agents Based on Offline Processing

In a meeting transcription system with automated speech recognition , offline processing yields better transcripts than online transcription, which has constraints on speed and latency. However, as the transcripts are processed in online mode, other machine- learning agents process the transcripts downstream and take action (e.g., booking calendars and extracting summary/action items). As the transcripts are reprocessed during offline processing, the accuracy improves, and then the agent actions might no longer be correct.

The novel solution is a system for updating agents based on offline processing . Each agent can mark the parts of interest within the transcripts using a combination of user/time/keywords. Changes are propagated in a makefile-like fashion to upstream processing. The system then push-notifies the agent if a relevant part changes during offline processing, so that the machine learning algorithm/agent can be rerun. In addition, by marking the transcripts, the agents provide regions of interest on which the offline processing can focus. The user is notified of all changes made by the agents along with the changes that happened in the transcripts .

1