Browse Prior Art Database

Dynamically Generating Collaborative Recommendations based on Action Graphs

IP.com Disclosure Number: IPCOM000236276D
Publication Date: 2014-Apr-16
Document File: 3 page(s) / 137K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a system that generates email recipient recommendations through the analysis of text as the user types the message. The system analyzes a message for indicators of an action or set of actions that a user needs to take, and then matches appropriate recipients based on those action indicators.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 3

Dynamically Generating Collaborative Recommendations based on Action Graphs

Mail clients and online social networks are a universal mechanism for connecting people and information in logical and organized ways that enable sharing and processing of information between the users. The most common mechanisms for sharing and processing information are the inbox, wall, activity stream, timeline, or profile. These mechanisms enable a user to rapidly both share and gather information among others in the networks.

When constructing a message, it can be difficult for a user to recall exactly to whom to address a message. For example, User A needs to address an issue with the plumbing. User A searches email files for Contact B, the plumber who fixed a problem in 2010. User A sends the mail to Contact B.

There is a clear need to better enable User A to recommend Contact B as a candidate to resolve the issue.

The novel contribution is a system that generates email recipient recommendations by:

• Detecting a user's authoring of a message
• Analyzing the message for candidate actions

• Mapping the candidate actions to the action graph
• Suggesting to the user candidates for the message

The system can be used in the inverse to suggest actions based on selected users or a given distribution list, as well as used iteratively as the document is authored.

The system uses a unique action graph mechanism. This action graph is derived from messages (i.e. from only internal messages, or all messages) in a collaborative system. The graph can be generated from the entire system or from a subset of an organization.

The system can perform analysis at different levels of granularity (e.g., sentence, paragraph, page, message, or thread). Messages may include collaboration artifacts such as blogs, wikis, activities, text messages, collections, files, and folders. The system analyzes, for each message, the User (i.e. sender), the recipient lists
(i.e. To, CC, BCC), the body, and associated metadata. A message with linked or

embedded content may be treated as a single message for analysis.

The system analyzes a message for indicators of an action or set of actions that a user needs to take. These action indicators can be gleaned from the data model, activity streams, verb(s), or a cursory analysis of the message content using Natural Language Processing (NLP) (i.e. gisting). The system can also implement a feedback loop to optimize the action discovery.

The invention uses graph libraries to map nodes and edges. The nodes are Actions and Users. Users are not connec...