I LINK :Search and Routing in Social Networks

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

ILINK:Search and Routing in Social Networks

Presenter : Chun-Ping Wu　Authors :Jeffrey Davcitz, Jiye Yu, Sugato Basu, David Gutelius, Alexandra Harris

KDD 2007

國立雲林科技大學National Yunlin University of Science and Technology

1


N.Y.U.S.T.

I. M.Outline

Motivation Objective ILINK FAQTORY Learning Framework Case Study Conclusion Comments

2


N.Y.U.S.T.

I. M.Motivation

This paper focuses on the problem of modeling how social networks accomplish tasks through peer production style collaboration.

3


N.Y.U.S.T.

I. M.Objective

To propose a general interaction model for the underlying social networks and then a specific model(ILINK) for social search and message routing.

4


N.Y.U.S.T.

I. M.ILINK

The ILINK model has various components. Node

Each node represents a user in the network, and has an associated profile. Supernode

A database D storing all the past message streams. Profile parameters E, R, F for all the nodes in the network. The set of all possible topics T.

Message A message m is routed between a node and the supernode.

5


N.Y.U.S.T.

I. M.FAQTORY

FAQtory system is implemented as a three-tiered client-server application. The client/front tier facilitates interactions between the nodes and the

FAQtory server via APIs. Web browser Email clients Instant messenger

6


N.Y.U.S.T.

I. M.FAQTORY-Routing Mechanism

7


N.Y.U.S.T.

I. M.Learning Framework

The learning framework in this case has to solve a set of interrelated learning problems. Heterogeneous topics

Cold-start problem

Privacy issues

Prior knowledge

Message matching

Scalability

Incremental learning

8


N.Y.U.S.T.

I. M.Case Study and Applications

Calo Test Pilot The FAQtory system of ILINK was developed as part of CALO, an

adaptive cognitive system funded under DARPA’s PAL(Perceptive Assistant that Learns) program.

PatoonLeader Provided with two features that leverage ILINK technology: “Suggested

Discussions” and the “Moderator’s Assistant”.

Other Message Applications Smart Rss Filter

Message Routing for Advertising

9


N.Y.U.S.T.

I. M.Conclusion

1010

The web has made large peer production frameworks possible, which are unprecedented in terms of both scale and intensity.

The importance of making such models work for real, ongoing web-based collaboration efforts is a growing issue and a wonderful challenge for machine learning and data mining techniques.


N.Y.U.S.T.

I. M.Comments

1111

Advantage The idea is good.

Drawback Many practical and theoretical issues remain.

Application Peer production

I LINK :Search and Routing in Social Networks

Documents

Transcript of I LINK :Search and Routing in Social Networks