Browse Prior Art Database

Method and System for Providing Unbiased Ad Quality Filtering

IP.com Disclosure Number: IPCOM000250371D
Publication Date: 2017-Jul-06
Document File: 3 page(s) / 48K

Publishing Venue

The IP.com Prior Art Database

Related People

Marc Bron: INVENTOR [+4]

Abstract

A method and system is disclosed for providing unbiased ad quality filtering or retargeting of advertisements (ads). The method and system uses a logistic regression model for identifying characteristics of users providing negative feedback, relating the negative feedback to a quality of the ads and quantifying a bias in ad feedback that come from an identification of sub-population of the users prone to providing the negative feedback.

This text was extracted from a Microsoft Word document.
This is the abbreviated version, containing approximately 38% of the total text.

Method and System for Providing Unbiased Ad Quality Filtering

Abstract

A method and system is disclosed for providing unbiased ad quality filtering or retargeting of advertisements (ads).  The method and system uses a logistic regression model for identifying characteristics of users providing negative feedback, relating the negative feedback to a quality of the ads and quantifying a bias in ad feedback that come from an identification of sub-population of the users prone to providing the negative feedback.

Description

Disclosed is a method and system for providing unbiased ad quality filtering or retargeting of advertisements (ads).  The method and system uses a logistic regression model for identifying characteristics of users providing negative feedback, relating the negative feedback to a quality of the ads and quantifying a bias in ad feedback that come from an identification of sub-population of the users prone to providing the negative feedback.

The method and system uses a logistic regression model to identify characteristics of users that are prone to be biased in negative feedback for ads.  The characteristics of users include information such as, but not limited to, a gender, an age and a location.  The method and system, then, quantifies a bias by incorporating characteristics of the ads and the characteristics of the users in terms of bias towards the negative feedback.  The manner in which the bias is quantified is through a probability of receiving a hide on the ads when controlling for a type of the users providing the negative feedback.

  

Also, the method and system uses a hide rate in a reliable way to make decisions about a quality of the ads.  In order to determine the hide rate, the method and system obtains characteristics of the users that do not want to see ads as wells as characteristics of the ads that the users do not want to see.  The characteristics of the users that do not want to see ads include variables such as, but not limited to, a user state, a user interest and an interaction between the user state and the user interest.

The characteristics of the ads that the users do not want to see include, but are not limited to, text based features, image based features and advertiser features such as brand.  The text based features can be, but need not be limited to, spam, readability level, and adult content.  On the other hand, the image based features can be, but need not be limited to, images containing text and flesh ads.  The method and system, then, considers the bias in suggesting the ads of higher quality based on the hide rate.  Thus, the ads of higher quality may have lower odds of receiving the negative feedback.

By tracking whether the negative feedback provided by the users is higher or lower than the mean hide rate, the method and system predicts that a specific user group (for example, users interested in technology) may be more likely to provide the negative feedback than others (for ex...