Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Click Aware Normalized Click Through Rate (CTR) Model for Ranking Documents

IP.com Disclosure Number: IPCOM000195595D
Publication Date: 2010-May-06
Document File: 3 page(s) / 35K

Publishing Venue

The IP.com Prior Art Database

Related People

Anand Murugappan: INVENTOR

Abstract

A click aware normalized Click Through Rate (CTR) model is provided for ranking documents. In an instance, an "Expected Click" (EC) score for documents in a presentation slate is obtained based on position of a document clicked by a user on the presentation slate. In another instance, EC scores are assigned to a position in a presentation slate based on average number of clicks corresponding to the position.

This text was extracted from a Microsoft Word document.
This is the abbreviated version, containing approximately 52% of the total text.

Click Aware Normalized Click Through Rate (CTR) Model for Ranking Documents

Abstract

A click aware normalized Click Through Rate (CTR) model is provided for ranking documents.  In an instance, an "Expected Click" (EC) score for documents in a presentation slate is obtained based on position of a document clicked by a user on the presentation slate.  In another instance, EC scores are assigned to a position in a presentation slate based on average number of clicks corresponding to the position.

Description

Disclosed is a click aware normalized Click Through Rate (CTR) model for ranking documents.

Typically, Click Through Rate (CTR) models are used to predict quality of documents, such as, for example, online advertisements or web search results on a results page of a web search engine.  However, the accuracy of CTR models is diluted due to biases in presentation of documents and relevance of other documents present on a presentation slate of documents displayed to a user.  The Click Over Expected Click (COEC) model normalizes CTR values based on rankings of documents.  The COEC model captures the possibility of user clicking a document based on the ranking of the document in the presentation slate.  Further, the COEC model also reduces penalty of a document as the ranking of the document increases.

The click aware normalized CTR model as disclosed herein predicts quality of documents by minimizing position bias and bias due to relevance of other documents present on a presentation slate of documents displayed to a user.  Similar to conventional COEC models, the click aware normalized CTR model uses a ratio of "Click" to an "Expected Click" (EC) to predict quality of documents.  "Click" is a positive vote received on a user click and EC is a penalty for the COEC ratio of a document being considered by the user.  However, the click aware normalized CTR model utilizes position information of a document clicked by a user to determine EC value for the document.

Thereafter, the click aware normalized CTR model computes SUM(Click_r)/SUM(P(Seen_r)), where "SUM(Click_r)" is the total number of clicks a document received and "SUM(P(Seen_r))" is the probability of a user viewing the document when displayed at a position (or rank), "r".  The value of "(P(Seen_r))" is determined by approximating a conventional CTR value for a document displayed at position "r".  The click aware normalized CTR model determines a probability value of a user viewing a document at position "r" based on quality of documents ranked above....