Browse Prior Art Database

Method and System for using Commercial Intent and Relevance in Online Advertising

IP.com Disclosure Number: IPCOM000211502D
Publication Date: 2011-Oct-07
Document File: 5 page(s) / 142K

Publishing Venue

The IP.com Prior Art Database

Related People

Lei Wang: INVENTOR [+5]

Abstract

A method and system for using commercial intent and relevance in online advertising is disclosed. Commercial intent is captured by modeling users' intents based on search queries. The search queries are evaluated from the perspective of informational motive and commercial motive.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 41% of the total text.

Method and System for using Commercial Intent and Relevance in Online Advertising

Abstract

A method and system for using commercial intent and relevance in online advertising is disclosed.  Commercial intent is captured by modeling users' intents based on search queries.  The search queries are evaluated from the perspective of informational motive and commercial motive.

Description

Disclosed is a method and system for using commercial intent and relevance in online advertising.  Commercial intent is captured by modeling users' intents based on search queries.  Search queries are assumed to be generated from a user's language model.  The search queries may be provided by a user for obtaining information such as, an informational query which is general in nature from the perspective of the user.  Alternatively, the search queries may be for commercial purposes such as a commercial query.  For example, a user may raise an informational query like “ABC multiplier” to enrich knowledge; or raise a commercial query like “Place name Philharmonic tickets” to book a ticket.  In most cases both informational intent and commercial intent coexist in a query and affect each other.  For example, a user has an interest in a tablet device.  The user may raise a query like “tablet device with 3G” to gain informational knowledge about features such as 3G, as well as to learn some commercial information like price, how to buy.  Leveraging on this information the user may make a decision about whether to perform further commercial activities regarding the “tablet device”.  Accordingly, from a supply perspective, an advertiser is expected to organize advertisements that satisfy both the two kinds of intents concurrently to maximize the attraction of the advertisements.  Thus, the user’s language model is defined as a mixture model, consisting of two portions an informational model, and a commercial model.  The informational model is a traditional language model with consideration of entropy of each word for informational relevance, while the commercial model characterizes language bias of a commercial intent leveraging on users' clicks on advertisements.

In the informational model, informational intent is analyzed.  The informational model computes by using a relative frequency of the information entropy of a word as follows:

ic( w) denotes the information content of w, which equals to the term frequency of w in D. (that is tf(w;D)) multiplies the self information of w (that is − log P( w)).  First of all P( w) is estimated using Maximum likelihood estimation (MLE) on a corpus of 100 million user queries sampled from an industrial search engine’s query logs.  Then the method uses bayesian smoothing to smooth it with one estimated on a larger archive of web pages of the search engine.  The informational model, is computed in the same way (MLE on) of computing the ordinary, or using a snapshot of ordinary search results exist...