Browse Prior Art Database

Office smooth comunication via active user behavior monitoring Disclosure Number: IPCOM000238315D
Publication Date: 2014-Aug-18
Document File: 1 page(s) / 25K

Publishing Venue

The Prior Art Database


This disclosure provides a machine learning based approach to detecting the status of the user in the instant communication environment. It collects and extracts features from the user's profile and dynamic activity and contextual features, together with their status mark information which is paritally available when some of the users 'flag' their status such as 'available/busy/away'. These features and 'flag' stutus enable supervised machine learning techniques to predict the unknown 'flag' status given the profile and contextual inforamtion such as location, time and current running application.

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

Page 01 of 1

Office smooth comunication via active user behavior monitoring

Instant messaging software is a necessary software utility in office collaboration. It provides an efficient way of communication. IM (Instant Messaging) softwares provides user status functionalities (Available/Busy/Away) to mark the user status. Such status mark will not block user away from incoming message, but in office environment, that will cause unexpected disturbance or over-waiting to users who does not change status on time.

Existing auto-status marking functionality only infer the "Away" status from no user operation for a long time. This is insufficient for filtering disturbance / waiting.

Correlation could be found between user's behavior / positioning and user's working status. e.g. sitting in a meeting room indicates a meeting status. Additionally, co-positioning among users also indicates co-behavior among these users. e.g. discussing in a same meeting room. With these two observations, we can induce the actual working status of an IM user and mark an appropriate status without manual operation on software.

We treat the problem as a multi-class classification problem. We regard the actual status (busy/meeting/available etc) as the classes to predict, use keyboard striking frequency, location, co-location information as the features, take users who often change their status on time as the learning sample producer. To train the IM data whose users do not update their status self-consci...