Skip to main content



Endorsements, Referrals, and Web Search

Imagine the scenario where I have two friends who really want to help me out this recruiting season. One of my friends is a current employee of the company I aspire to work for. Being a caring friend, they refer me to their recruiter. My other friend also sends a message on my behalf to a recruiter on LinkedIn for the same company (that they have no connection to). Clearly, the referral from my friend who is already an employee of that company is going to hold a lot more influence than the referral by more other friend. This shows that it is not simply enough to receive a referral. The person who is giving the referral needs to have influence for it to be meaningful. This concept also comes up in presidential elections. If someone such as President Obama is going to endorse a candidate, that endorsement will carry a lot of influence in the election. The credibility of that candidate will be boosted by an endorsement from President Obama.

The two algorithms we learned in class, HITS and PageRank, are built on this intuition: a website’s importance is determined by the importance of the websites that link to it. For example, if an extremely relevant/credible site contains a hyperlink to another site, the credibility of the other site should be boosted by this. In the hub-authority model we learned in class, each node keeps track of two values, a hub score and an authority score. Recursively, a hub score is calculated by summing the authority scores of its neighboring nodes and an authority score is calculated by summing the hub scores it its neighboring nodes. Each hub gives its “endorsement” to authorities it links to, therefore, nodes with neighbors that have large hub scores will have large authority scores.

Comments

Leave a Reply

Blogging Calendar

October 2019
M T W T F S S
 123456
78910111213
14151617181920
21222324252627
28293031  

Archives