U.S. patent application number 10/840667 was filed with the patent office on 2005-03-24 for search engine for selecting targeted messages.
This patent application is currently assigned to Overture Services, Inc.. Invention is credited to Anick, Peter, Chavet, Laurent, Maghoul, Farzin, Marais, Johannes L..
Application Number | 20050065917 10/840667 |
Document ID | / |
Family ID | 32851352 |
Filed Date | 2005-03-24 |
United States Patent
Application |
20050065917 |
Kind Code |
A1 |
Anick, Peter ; et
al. |
March 24, 2005 |
Search engine for selecting targeted messages
Abstract
A search engine receives query terms from a client. In response,
the search engine executes a search on a web directory to identify
zero or more documents that match the query terms. The identified
documents are associated with one or more categories. The search
engine probabilistically selects one of the categories associated
with the identified documents. Each message in a message database
is also associated with one or more of the categories. The search
engine accesses the message database and selects at least one
message associated with the selected category. The search engine
returns a web page containing references to the documents matching
the query terms and the one or more messages selected from the
message database to the client.
Inventors: |
Anick, Peter; (Marlboro,
MA) ; Maghoul, Farzin; (Hayward, CA) ; Marais,
Johannes L.; (Mountain View, CA) ; Chavet,
Laurent; (San Mateo, CA) |
Correspondence
Address: |
BROWN, RAYSMAN, MILLSTEIN, FELDER & STEINER LLP
900 THIRD AVENUE
NEW YORK
NY
10022
US
|
Assignee: |
Overture Services, Inc.
|
Family ID: |
32851352 |
Appl. No.: |
10/840667 |
Filed: |
May 6, 2004 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
10840667 |
May 6, 2004 |
|
|
|
09799863 |
Mar 5, 2001 |
|
|
|
6778975 |
|
|
|
|
Current U.S.
Class: |
1/1 ;
707/999.003; 707/E17.108 |
Current CPC
Class: |
Y10S 707/956 20130101;
Y10S 707/99931 20130101; Y10S 707/99943 20130101; Y10S 707/99932
20130101; G06F 16/951 20190101 |
Class at
Publication: |
707/003 |
International
Class: |
G06F 007/00 |
Claims
1. A method of selecting a targeted message, comprising: receiving
one or more query terms; probabilistically selecting a category
from among a set of categories responsive to how well the category
matches the one or more query terms; and selecting the targeted
message responsive to the selected category.
2. The method of claim 1, wherein the probabilistically selecting
step comprises: identifying a set of documents responsive to the
one or more query terms, wherein documents in the set are
associated with one or more categories in the set of categories;
and selecting the category responsive to the one or more categories
associated with the documents in the set.
3. The method of claim 2, further comprising: defining a hierarchy
for the set of categories.
4. The method of claim 3, wherein the category hierarchy is
substantially flat.
5. The method of claim 3, wherein the categories in the hierarchy
are defined responsive to a plurality of message types.
6. The method of claim 3, wherein the step of selecting the
category comprises: selecting a hierarchical ancestor of a category
associated with a document in the set.
7. The method of claim 2 wherein the step of selected a category
comprises: randomly selecting the category from among the one or
more categories associated with the documents in the set.
8. The method of claim 1, wherein the step of probabilistically
selecting a category comprises: selecting the category responsive
to weight values associated with the categories in the set.
9. The method of claim 2, wherein the step of selecting a category
comprises: selecting the category responsive to weight values
associated with the documents in the set.
10. The method of claim 1, wherein the step of selecting the
targeted message responsive to the selected category comprises:
selecting the message responsive to a number of shares of the
selected category owned by the message.
11. The method of claim 1, wherein there are no messages associated
with the selected category and wherein the step of selecting the
targeted message responsive to the selected category comprises:
selecting a general message associated with a general category.
12. The method of claim 2, wherein the one or more query terms are
received from a client and further comprising: sending references
to the identified documents and the selected targeted message to
the client.
13. The method of claim 12, further comprising: sorting the
identified documents by an associated category of each document;
and sending the references to the identified documents to the
client in the sorted order; wherein the category is selected
responsive to the category of the documents sent to the client.
14. A search engine for selecting a targeted message, the search
engine comprising: a receiving module for receiving one or more
query terms from a client; a web directory access module for
accessing a web directory holding references to documents, each
document associated with one or more categories, and for
identifying a set of documents referenced in the web directory that
match the one or more query terms; a category selection module for
probabilistically selecting a category responsive to the one or
more categories associated with the documents identified by the web
directory access module; and a message database access module for
accessing a database of messages, the messages each associated with
at least one category, and for selecting a message associated with
the category selected by the category selection module.
15. The search engine of claim 14, wherein the categories in the
web directory are arranged hierarchically.
16. The search engine of claim 15, wherein the category hierarchy
is substantially flat.
Description
BACKGROUND
[0001] 1. Field of the Invention
[0002] This invention pertains in general to advertising on the
Internet and in particular to selecting an advertising message for
display in response to a given search query.
[0003] 2. Background Art
[0004] Many web sites provide search engines which accept a
keyword-based query from a user and produce a list of matching
categories or documents to the user in return. For example, the
ALTAVISTA web site contains a search engine that enables a user to
execute a query on multiple keywords and receive matching
information, such as references to web pages, in response.
[0005] Such web sites typically generate revenue by displaying
advertising messages, such as banner ads, to the user along with
the returned search results. However, users often ignore the ads
and may even find the ads irritating. Therefore, web sites strive
to display ads related to the search query on the assumption that a
user is more likely to show interest in such ads. For example, if
the user searches for the phrase "Tahoe ski areas," then the user
may tend to respond to ads for products such as skis, hotels near
Lake Tahoe, etc. because the user is presumably interested in
skiing.
[0006] To provide this functionality, web sites typically sell
keywords to advertisers. An advertiser's ad is displayed when the
user searches on a keyword owned by that advertiser. Since multiple
advertisers may desire the same keyword, the web sites often
auction desirable keywords to the highest bidders. For example, a
web site might auction generic keywords such as "car" and
"automobile."
[0007] Still, web sites often display ads that are inappropriate in
view of the query received from the user. Sometimes, a particular
ad is generally appropriate for many queries and the advertiser
therefore buys a large number of keywords for the ad. As a result,
other advertisers are blocked from using the keywords for their
own, possibly more relevant, ads. In a competitive advertising
scenario, for example, a dealer of a first make of automobile might
buy the keywords corresponding to other makes, thereby intending to
entice buyers away from competing makes. As a result, a user may be
annoyed to see ads for the first make of automobile when the user
is seeking information about another make.
[0008] Moreover, advertisers often have little guidance when
selecting keywords and seldom know the characteristics of
real-world user queries. As a result, advertisers frequently buy
overly general keywords like "cars," on which users seldom query.
Consequently, often no advertisers have purchased the exact
keywords utilized by the users. Therefore, the web site displays
general rotation ads instead of targeted ads.
[0009] Therefore, there is a need in the art for a way to increase
the relevancy of the selected ads to the query terms received from
users. Preferably, a solution to this need will overcome the
deficiencies of selecting ads via keywords.
DISCLOSURE OF THE INVENTION
[0010] The above need is met by a search engine that
probabilistically maps a user query into a category, and then uses
the category to select a targeted message. The search engine
receives the query terms from the user's client computer via the
Internet. In response, the search engine executes a search on a web
directory to locate zero or more documents that match the query
terms. The search engine may also search other directories in order
to identify web pages or other documents that match the query
terms.
[0011] If the search engine has located one or more documents
matching the query terms, the search engine determines the
categories corresponding to the retrieved documents. In a preferred
embodiment of the present invention, each document in the web
directory is assigned to a category in a hierarchical directory. In
one embodiment, the hierarchical directory is derived from
information available through the Open Directory Project (ODP). In
general, the ODP is a hierarchical directory of web pages assembled
by human operators who review the web pages and assign the pages to
certain categories.
[0012] In an alternative embodiment of the present invention, the
hierarchy in the web directory is completely or substantially flat.
A flatter hierarchy simplifies the process of selecting a message
because a documents belongs to only one or a few categories,
instead of belonging to a long chain of ancestor categories.
[0013] The search engine preferably next selects one of the
categories corresponding to the retrieved documents. In one
embodiment of the present invention, the search engine selects the
category of one of the returned documents at random. Since
categories having more documents matching the query terms are more
likely to be selected, the random selection probabilistically
selects the category best matching the query terms. In alternative
embodiments, the search engine uses one or more other heuristics to
bias the category selection.
[0014] Once the search engine selects a category, it preferably
accesses a message database and selects a message associated with
the selected category. The message database preferably holds
multiple messages, with each message belonging to one or more of
the categories enumerated in the web directory. In a preferred
embodiment, the messages are banner ads provided by advertisers.
Preferably, the advertisers purchase "shares" in one or more of the
categories for each banner ad. When the search engine selects the
category in which a given ad has shares, the search engine selects
that ad a number of times proportional to the shares of that
category owned by the ad.
[0015] The search engine preferably returns a web page containing
the document references retrieved from the web directory and the
one or more messages selected from the message database to the
client. Typically, the web page includes a subset of the document
references (e.g., the 10 best matches) and at least one banner ad
(i.e., the message). Obviously, the web page can include other
message types in addition to or instead of the banner ad and/or a
different number of references or messages.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] FIG. 1 a block diagram illustrating a typical computing
environment according to an embodiment of the present
invention;
[0017] FIG. 2 is a high-level block diagram illustrating an
exemplary embodiment of a computer system for use as a client or
web server;
[0018] FIG. 3 is a block diagram illustrating a lower-level view of
the web server containing the search engine according to an
embodiment of the present invention; and
[0019] FIG. 4 is a flowchart illustrating steps performed by the
search engine according to an embodiment of the present
invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0020] FIG. 1 is a block diagram illustrating a typical computing
environment 100 according to an embodiment of the present
invention. FIG. 1 illustrates three client computers 110A, 110B,
110C in communication with the Internet 112 using known
communications technologies. Multiple other web servers 114, 116
are also in communication with the Internet using known
communications technologies. One of the web servers 114 includes a
search engine 118. As is known in the art, the Internet 112 is a
publicly accessible network of computers and supports
communications among the various computers in communication with
it.
[0021] Although FIG. 1 illustrates only three clients 110,
embodiments of the present invention can have thousands, or even
millions, of clients simultaneously in communication with the web
servers 114, 116 via Internet 112. As is known in the art, a client
is typically a personal computer system such as an IBM PC- or Apple
Macintosh-compatible computer. A client 110 typically use a web
browser 120 such as NETSCAPE NAVIGATOR from Netscape Communications
Corp. or INTERNET EXPLORER from Microsoft Corp. to exchange
information with the web servers 114, 116. In a preferred
embodiment of the present invention, this information takes the
form of hypertext markup language (HTML) web pages transmitted via
the hypertext transport protocol (HTTP). However, alternative
embodiments of the present invention utilize different transmission
techniques and/or exchange different types of information.
[0022] As is known in the art, a web server 114, 116 is typically
comprised of one or more computer systems adapted to simultaneously
interact with multiple clients 110 via the Internet 112. The web
servers preferably contain one or more documents referenced via
uniform resource locators (URLs), such as
"http://www.altavista.com". In one embodiment, a document is a web
page containing text, images and/or other types of media. Those of
ordinary,skill in the art will recognize that the present invention
can also be utilized with other types of documents referenced using
techniques other than URLs.
[0023] As described above, one 114 of the web servers preferably
contains a search engine 118. A user preferably uses the web
browser 120 to download a web page from the search engine 118
containing a search form. The web browser 120 allows the user to
enter query terms into the search form and then transmit the query
terms to the search engine 118 via HTTP or another communications
protocol. The query terms are typically one or more words or other
text strings, but can include other information indicating the
types of documents desired by the user.
[0024] Upon receiving the search terms, the search engine 118
preferably searches for documents matching the query terms and
provides references (e.g., URLs) to the matching documents to the
client 110. In a preferred embodiment of the present invention, the
search engine 118 derives the search results from documents
available from web servers 114, 116 connected to the Internet.
However, the exact data domains searched by the search engine 118
are not material to the present invention. The search engine is
preferably the ALTAVISTA search engine from AltaVista Co. and
available at http://www.altavista.com.
[0025] The search engine 118 preferably provides the search results
to the client 110 in the form of one or more web pages containing
the references to the matching documents. The web pages are
typically displayed by the browsing software 120 at the client 110.
In addition to the search results, each web page preferably
contains one or more messages selected by the search engine 118. A
"message" is a communication including text, graphics, sounds
and/or any other type of information.
[0026] Preferably, one or more of the messages are "targeted" to
the query terms in order to make the messages more relevant to the
user who initially provided the query. As used herein, the term
"targeted" means that the messages are selected and/or generated
such that the messages are related to the query terms and/or other
information known about the user or client 110.
[0027] In a preferred embodiment, the targeted messages contain
advertising and are displayed on behalf of one or more advertisers.
In one embodiment, the messages are displayed as banner ads that
accompany the search results. In alternative embodiments, the
messages are displayed in a "popup" window separate from the search
results, displayed as search results, and/or displayed in any other
practical location or manner.
[0028] FIG. 2 is a high-level block diagram illustrating an
exemplary embodiment of a computer system for use as a client 110
or web server 114, 116. Illustrated are at least one processor 202
coupled to a bus 204. Also coupled to the bus 204 are a memory 206,
a storage device 208, a keyboard 210, a graphics adapter 212, a
pointing device 214, and a network adapter 216. A display 218 is
coupled to the graphics adapter 212.
[0029] The at least one processor 202 may be any specific or
general-purpose processor such as an INTEL x86 or
POWERPC-compatible central processing unit (CPU). The storage
device 208 may be any device capable of holding large amounts of
data, like a hard drive, compact disk read-only memory (CD-ROM),
DVD, or some other form of fixed or removable storage device. The
memory 206 holds instructions and data used by the processor 202.
The pointing device 214 may be a mouse, track ball, light pen,
touch-sensitive display, or other type of pointing device and is
used in combination with the keyboard 210 to input data into the
computer system 200. The network adapter 216 couples the computer
system 200 to the Internet 112 or another network.
[0030] Modules 220 for providing the functionality described herein
are preferably stored on the storage device 208, loaded into the
memory 206, and executed by the processor 202. Alternatively,
hardware or software modules may be stored elsewhere within the
computer system 200. As used herein, the term "module" refers to
computer program logic and/or any hardware or circuitry utilized to
provide the functionality attributed to the modules. The types of
hardware and software within the computer system 200 may vary
depending upon how the computer system is utilized. For example, a
computer system used as a web server 114, 116 is likely to have
greater processing power and storage capacity than a computer
system used as a client 110.
[0031] FIG. 3 is a block diagram illustrating a lower level view of
the web server 114 containing the search engine 118. FIG. 4 is a
flowchart illustrating steps performed by the search engine 118
according to an embodiment of the present invention. It will be
understood by those of skill in the art that the search engine 118
is a module and that the actions and functionality attributed to
the search engine herein can be performed by the search engine
itself or by other modules within the web server 114. Accordingly,
the term "search engine" is intended to include any module or other
entity that performs the attributed functions.
[0032] Initially, the search engine 118 receives 410 the query
terms. In response, the search engine executes 412 a search on a
web directory 312 to locate zero or more documents that match the
query terms. As used herein, a document "matches" the query terms
if the document is in some way related to the query terms. For
example, in one embodiment, documents that contain one or more of
the query terms match the query terms. Those of skill in the art
will recognize that different embodiments of the present invention
can use different thresholds of relatedness to determine whether a
document matches the query terms.
[0033] Preferably, the web directory 312 contains a searchable
index of terms contained in documents available from web servers
114, 116 on the Internet 112, and references to those documents. In
one embodiment, the web directory 312 contains only a small subset
of the total number of documents available. The web directory 312
is preferably optimized to allow fast retrieval of the references
to documents matching the query terms.
[0034] In a preferred embodiment of the present invention, each
document in the web directory 312 is assigned to a category in a
hierarchical directory. In one embodiment, the hierarchical
directory and category assignments are derived from information
available through the Open Directory Project (ODP). In general, the
ODP is a hierarchical directory of web pages assembled by human
operators who review the web pages and assign the pages to certain
categories. More information about the ODP is available at
http://www.dmoz.org. In another embodiment, the web directory 312
is derived from one or more alternative data sets.
[0035] For example, a document related to snow skiing might be
located at "Sports: Skiing: Regional: North America" in the
hierarchy. Likewise, a document related to automobiles might be
located at "Home: Consumer Information: Automobiles." Note that the
hierarchies may be arranged so that certain documents are stored at
multiple locations. For example, documents related to "skis" can be
found in the categories "Home: Consumer Information: Sports and
Recreation: Skiing" and "Sports: Skiing: Backcountry: Telemark:
Equipment."
[0036] In an alternative embodiment of the present invention, the
hierarchy in the web directory 312 is completely or substantially
flat. One embodiment flattens the hierarchy by removing regional-
and/or geography-based branches. Likewise, another embodiment trims
the hierarchy by removing certain broad categories into which many
documents fall.
[0037] In another embodiment of the present invention, the web
directory 312 holds documents in a specialized message hierarchy.
The message hierarchy resembles the ODP hierarchy described above
except that the hierarchical levels and categories are defined in
view of the types of messages available for inclusion with the
search results. Thus, in the embodiment wherein the messages
contain advertising, the hierarchy categories depend upon the types
of ads available. For example, if the ads are placed in broad
categories such as "healthcare," "sporting equipment," and/or
"automobiles," then the web directory 312 holds the web pages in
similar categories. As described below, more specific message
categories can lead to more targeted messages.
[0038] The search 412 performed by the search engine 118 identifies
414 zero or more documents matching the query terms and references
to those documents. A focused query might return documents from
only a single category, while less focused queries will probably
return documents from several categories. Since queries are often
vague, the latter type of query should occur more frequently than
the former.
[0039] The search engine 118 preferably next selects 418 one or
more of the categories corresponding to the identified documents.
In one embodiment of the present invention, the search engine 118
randomly selects a category from the identified documents. This
embodiment is referred to as the "dart board" because it operates
in a manner similar to simply throwing a dart at the identified
documents and selecting the category of the document hit by the
dart.
[0040] An advantage of the dart board approach is that the
probability that a category is selected corresponds to the degree
to which the query terms match the category. Consider an extreme
example: If the query terms match 99 documents in a first category
and one document in a second category, then the query terms can be
said to match the first category much better than the second
category. Correspondingly, there is a 99% chance that the search
engine 118 will select the first category and a one percent chance
that it will select the second category. Since the category is
selected probabilistically, the search engine will occasionally
select different categories in response to the same query terms
(assuming matching documents are found in multiple categories).
[0041] In alternative embodiments, the search engine 118 uses one
or more other heuristics to select 418 the category. In one
embodiment, the documents and/or categories in the web directory
312 have weight values that the search engine 118 uses to bias the
selection. For example, the heuristic can bias the selection
towards the category containing the greatest number of documents.
In another example, the heuristic can bias the selection by
utilizing weight values determined from the number of times a
document is selected by a user after being returned as part of a
search result. In yet another example, the results of the dartboard
approach can be monitored and the selections biased to ensure that
each category is selected a certain percentage of the time.
[0042] Another alternative embodiment of the present invention uses
a heuristic that occasionally selects a hierarchical ancestor of
the category in which a document was identified. For example, if a
document is found in the category "Home: Consumer Information:
Automobiles," this embodiment occasionally selects the "Consumer
Information" category. This heuristic accounts for the potential
situation wherein a category having children is rarely selected
because the query terms frequently match documents in the child
category.
[0043] Once a category is selected, the search engine 118
preferably accesses a message database 314 and selects 420 a
message from the selected category. The message database 314
preferably holds multiple messages, with each message belonging to
one or more of the categories enumerated in the web directory 312.
Since the category is selected probabilistically, and the message
is selected from the category, the same query terms can result in
messages selected from different categories.
[0044] In a preferred embodiment, the messages are banner ads
provided by advertisers. Preferably, the advertisers purchase
"shares" in one or more of the categories for each banner ad. Then,
when the category in which the ad owns shares is selected, the
search engine 118 selects the ad a number of times proportional to
the number of shares owned by the ad. For example, if an ad owns
10% of the shares in the "Snow Skiing" category, the search engine
118 will select that ad 10% of the time that the "Snow Skiing"
category is selected.
[0045] If 422 no messages are associated with the category selected
by the search engine 118, or the search engine does not identify
414 any documents that match the query terms, in one embodiment of
the present invention the search engine selects 424 a message from
a "general" category. In this embodiment, messages own shares in
the general category in addition to the other categories. In this
manner, the search engine 118 is adapted to provide a non-targeted
message if there are no documents or messages related to the query
terms.
[0046] In an alternative embodiment of the present invention, if no
messages own shares in the selected category, the search engine 118
searches the category's hierarchical ancestors for messages. If the
search engine 118 does not find a message through this search, then
it preferably selects a general message.
[0047] After selecting the message, a preferred embodiment of the
search engine also searches 426 one or more other databases 316 to
identify references to documents matching the query terms. These
other databases 316 are similar to the web directory 312, except
that the indexed documents in the other databases are not
necessarily associated with categories. In this embodiment, the web
directory 312 may contain only a small subset of the documents
available from servers 114, 116 on the Internet 112 while the other
databases contain a much larger subset of the documents. Therefore,
the search engine 118 determines the category of the ad from the
small set of categorized documents but identifies the majority of
the search results from one or more databases 316 of uncategorized
documents.
[0048] As discussed above, the search engine 118 preferably returns
428 a web page containing the references to the documents matching
the query terms and the one or more messages selected from the
message database 314 to the client 110. Typically, the web page
includes a subset of the document references (e.g., the 10 best
matches) and at least one banner ad (i.e., the message). In an
alternative embodiment of the present invention, the references
returned by the search engine 118 are sorted by category and the
selected message is targeted to the category of the returned
references. For example, if the search engine 118 identifies 100
references in eight different categories, the search engine returns
to the client 110 all or some of the documents in the first
selected category along with a message selected from that category.
If the user requests additional results pages, the search engine
118 returns documents from a second category, along with a message
selected from the second category, etc.
[0049] In sum, the present invention selects messages by mapping a
practically infinite set of potential query terms to a finite set
of categories. The probability that the present invention selects a
category is determined by how well the category matches the query
terms, and the probability that the present invention selects a
message is determined by the number of shares in the selected
category owned by the message. Thus, the present invention provides
targeted messages without the deficiencies that occur when merely
associating ads with keywords.
[0050] The above description is included to illustrate the
operation of the preferred embodiments and is not meant to limit
the scope of the invention. The scope of the invention is to be
limited only by the following claims. From the above discussion,
many variations will be apparent to one skilled in the relevant art
that would yet be encompassed by the spirit and scope of the
invention.
* * * * *
References