Browse Prior Art Database

System and method to rank search results in a priority based on relation distance

IP.com Disclosure Number: IPCOM000166807D
Original Publication Date: 2008-Jan-24
Included in the Prior Art Database: 2008-Jan-24
Document File: 3 page(s) / 15K

Publishing Venue

IBM

Abstract

Our invention provides a solution to efficiently finding and ranking associations within the scope of an enterprise. This method builds up the social network within the enterprise by using colleague relationship information which integrates both the static tree structure information in the directory server and the dynamic behavior information that the user send and receive the Email. This social network is stored as a graph. The node of the graph represents an employee; the edge between two nodes is computed to represent their relationship. The value of an edge is computed in a way to combine both static enterprise structure information and dynamic contact information. Static one is set initially when the graph is created; dynamic one is updated real time. In this way, we can formalize the problem of association search as that of solving the near-shortest paths of the graph. The rough process of our method is like this: The requester input a target’s partial name, then the path length between the sender and the possible receiver is computed. This path length is computed on the graph formed under the rule of Manager-Employee, Sender-Receiver and Community relationship. At last our method recommends the list in an ascendant order according to path length.The advantage of our method is that, by building up a social network of the enterprise and storing this network as a graph, the problem can be simply formalized as solving the near-shortest paths of the graph. This method integrates more semantic information into search which will reflect the relationship among the requester and the search result more accurately, thus makes the finding and ranking more efficiently

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 3

System and method to rank search results in a priority based on relation distance

In our invention, we define three social relationships:


D1. Manger-Employee relationship. This relationship reflects that there is connection if a person is the direct manager of another person. We describe this attribute as attribute A. A is an integer which represents the extent of intimate of this relationship. We look on this relationship as a very close one, so we put it a relatively large number, for example, we could set A to be 3. Its inverse 1/A represents the weight of this relationship.

D2. Sender-Receiver relationship. This relationship reflects that there is a connection if a person sends to or receives from another person an email. We describe this attribute with attribute B. We look on this relationship to be less close than Manger-Employee relationship. So we may put it a smaller integer, for example, set B to be 2. Its inverse 1/B represents the weight of this relationship.

D3. Technical community colleague. There is a connection if the two persons belong to the same technical community in an Enterprise scope. It is quite often that many people belong to quite a few technical communities of the enterprise for common interest and research field. For example, somebody participates in testing methodology community, somebody participates SOA best practice community, etc. We describe this attribute with an attribute C. We put C an integer smaller than A and B. For example, set C to 1. Its inverse 1/C represents the weight of this relationship.

   D1 is relatively static information which can be obtained in corporate directory. D2 is dynamic information which can be obtained by monitoring the Email server constantly and updating our graph in real time. D3 is also relatively static information and can be obtained in the same way as D1.

We use an adjacency list to store the graph.

   Every vertex in the graph represents a person in the company. Every edge in the graph represents the relationship between two person.

    If any two person's relationship is D1, then the weight of the edge between these two person is 1/A=1/3.

    If any two person's relationship is D2, then the weight of the edge between these two node is 1/B=1/2.

    If any two person's relationship is D3, then the weight of the edge between these two node is 1/C=1/1.

  If any two person's relationship is any combination of D1,D2,D3,then th...