 |
Description  |
|
|
TECHNICAL FIELD
This invention relates generally to the usage of a computer network by a
user as more specifically to the techniques of providing specialized
information to a network user based on accumulated user data.
BACKGROUND
The World Wide Web (WWW) of computers is a large collection of computers
operated under a client-server computer network model. In a client-server
computer network, a client computer requests information from a server
computer. In response to the request, the server computer passes the
requested information to the client computer. Server computers are
typically operated by large information providers, such as commercial
organizations, governmental units, and universities, and are typically
referred to as "web sites". Client computers are typically operated by
individuals.
To ensure interoperability in a client-server computer network, various
protocols are observed. For example, a protocol known as the Hypertext
Transport Protocol (HTTP) is used to move hypertext files across the WWW.
In addition, the WWW observes several protocols for organizing and
presenting information, two examples being the Hypertext Markup Language
(HTML) and the Extensible Markup Language (XML). The information delivered
by the server computer is typically referred to as a "web page".
A server computer can use a technique known as "dynamically-generated
customized pages" to create a web page in response to a request for
information from a client computer. A dynamically-generated customized
page results in a set of information in a particular format. For example,
a first client computer may support the ability to represent information
in a number of columns, while a second client computer may support the
ability to represent information in a table. Thus, a server computer
receiving a request from the first client computer can dynamically
generate the requested information in a format with columns. It can
respond to a request from the second client computer by dynamically
generating the requested information in table format. In this example, two
customized pages are created to represent the same information.
It is not unusual for a server computer on the WWW to contain thousands or
even tens of thousands of web pages. This large quantity of e makes it
difficult for a person, i.e., a "web site visitor", operating a client
computer to locate the information of most interest to them. In much the
same way that dynamically-generated customized pages can be used to
present the same information in a different presentation format for each
client computer, dynamically-generated customized pages can be used to
select the information to be displayed so that each web site visitor may
see information customized to their specific interests. This process is
known in the art as personalization.
Personalization can be achieved through current technology using survey
questions to ascertain the visitor's interests, and using
dynamically-generated customized pages compute customized pages for each
visitor. There are two disadvantages to this approach. First, web site
visitors frequently prefer to not fill out questionnaires when visiting a
web site, making it difficult for a site to gather the necessary visitor
preference data. Second, dynamic generation of every page on a server
computer does not scale well for large numbers of requests. In other
words, existing methods provide a relatively slow response when a large
number of requests are made for personalized pages. This slow response
time is attributable to the fact that in existing systems a computer
program must be executed to completely generate each dynamic page on every
single request.
In view of the foregoing, it would be highly desirable to provide a
technique to unobtrusively gather web site visitor preference data and
efficiently respond to a large number of requests for personalized pages.
SUMMARY OF THE INVENTION
The invention is a method and apparatus for learning in what a visitor is
interested and what demographics the visitor may demonstrate so as to
deliver personalized information to the visitor based upon accumulated
data, and to do so without requiring dynamic page generation for each
individual visitor.
For example, a visitor may demonstrate interest in football and, in
particular, his favorite football team. The present invention learns this
by observing the behavior of the visitor, i.e., which sports articles he
reads and if such articles are focused even further. If a tendency is
observed, the learned knowledge is then used to deliver more information
about that team to the visitor. Such preferred articles can be recycled by
having the invention deliver the same information to other visitors who
have the same favorite team.
Visitor interests can be tracked by including "keyword directives" in
content contained within the web site. These keyword directives specify a
keyword indicating the type of category of information represented by the
content. As the content is delivered to the visitor in the form of a web
page, the number of keyword directives attached to the content is
accumulated into a specified visitor profile. Over time, this visitor
profile can represent the types of information the visitor has viewed and
serve as an indicator of his or her preferences. In this way, the
invention can accumulate a visitor profile unobtrusively, without
requiring the visitors to fill out a survey or questionnaire. The profile
may also be augmented with explicit information the visitor provides over
time, such as a name or address provided when ordering a product from the
site.
The present invention then delivers personalized pages to the visitor by
examining such visitor's profile. Another directive, called a
personalization directive, may be placed into web pages that are to be
customized by the invention. These directives cause a personalization
function to be applied to the visitor's profile data. The result of the
personalization function defines an attribute to be used for locating
personalized page fragments, called "page components", that the invention
then assembles into a customized page for the visitor. In this manner,
each visitor may receive a page containing three different classes of
data: common data received by all visitors, personalized data received by
a similar group of visitors, and individual data received only by this one
visitor. The present invention assembles all of this data and delivers a
"personalized" page to the visitor.
The present invention stores personalized page components in a cache.
Subsequent delivery of the same page components is satisfied by retrieving
the information from the cache, rather than by dynamically generating it
each time. The present invention can therefore take advantage of a common
situation where large groups of visitors share similar interests and
should receive the same data. Since previously generated personalized page
components need not be re-generated for every visitor, computational
overhead is reduced tremendously by supplying such pre-generated page
components.
For example, a home page for a large web site might include a
personalization directive describing the inclusion of an article related
to a visitor's favorite NFL team. The personalization directive function
examines the visitor profile, determines the favorite team, and includes
the appropriate page with information about that team. In this way, each
visitor to the web site might receive a different introductory web page,
customized for their preferences. Even though every visitor receives a
page that appears to be customized for them, since, in fact, there are
only 30 or so NFL teams; the caching mechanism of the invention ensures
that the dynamic page generation only occurs at most 30 or so times. If
one million visitors come to the site, most of the visitors simply receive
a web page that was already dynamically generated for a previous visitor.
In essence, the invention allows "personalized" pages to be constructed by
choosing from a set of previously computed pages, rather than by
dynamically computing each page for every visitor.
It is a primary object of the present invention to provide an efficient
mechanism for gathering visitor preference and behavior information and
storing it in a visitor profile.
Another object of the invention is categorizing content in a web site and
associating viewed categorized content with a user to develop a visitor
profile.
It is another object of the present invention to provide a highly efficient
and scalable mechanism for assembling personalized pages based on
information contained in the visitor profile, without requiring a full
dynamically-generated customized page computation for each visitor.
It is still another object of the present invention to allow for specific
data from the visitor profile to be directly inserted into personalized
pages.
Yet another object of the invention is to insert pre-customized content
into various areas of a single web page.
It is a further object of the invention to allow for visitor profile data
to be based on the actual content viewed by the visitors.
It is another object of the invention to allow for visitor profile data to
be gathered and updated efficiently even in the case where multiple web
servers are operating simultaneously to deliver information to users in
parallel.
It is another object of the invention to provide efficient management and
storage of visitor profile data for large web sites that may have as many
as 10 million visitors or more.
The above objects of the invention and the brief description of the
preferred embodiment should be constructed to be merely illustrative of
some of the more prominent features and applications of the invention.
Many other beneficial results can be attained by applying the disclosed
invention in a different manner or modifying the invention as will be
described. Accordingly, other objects and a fuller understanding of the
invention may be had by referring to the following Detailed Description of
the preferred embodiment.
BRIEF DESCRIPTION OF THE DRAWINGS
For a more complete understanding of the present invention and the
advantages thereof, reference should be made to the following Detailed
Description taken in connection with the accompanying drawings in which:
FIG. 1 illustrates a client-server computer network that may be operated in
accordance with the present invention;
FIG. 2 is an example page delivered by a web server;
FIG. 3 illustrates a relationship diagram of the primary components in the
present invention; and
FIG. 4 illustrates the invention configured for use with multiple server
computers.
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 illustrates a client-server computer network 100 that may be
operated in accordance with the present invention. For the preferred
embodiment, the network 100 includes at least one client computer 110 and
at least one server computer 130. The client computer 110 and the server
computer 130 arc connected by a transmission channel 120, which may be any
wire or wireless transmission channel.
The client computer 110 may be a standard computer including a Central
Processing Unit (CPU) 112 connected to a memory (primary and/or secondary)
114. The memory 114 stores a number of computer programs, including a
"browser" 116. As known in the art, a browser is used to communicate with
remote server computers 130 and to visually present the information
received from such computers. The client computer 110 establishes network
communications through a standard network connection device 118.
The server computer 130 includes standard server computer components,
including a network connection device 138, a CPU 132, and a memory
(primary and/or secondary) 134. The memory 134 stores a set of computer
programs to implement the processing associated with the invention. These
programs arc collectively referred to as a the web server software 136.
The invention may be used with any web server software, including, but not
limited to, Netscape Enterprise Server from Netscape Inc., Internet
Information Server from Microsoft, or Apache from the Apache HTTP Server
Project.
FIG. 2 illustrates a typical web page 200. The web page contains graphical
information and textual information. Web page design varies greatly, but
usually follows a general pattern of being divided up into sections of
related information. In the provided example, there are four areas of
information 210, 220, 230, and 240. In the terminology of the invention,
each of the distinct sections of the web page, such as 210, 220, 230 and
240, are called `components`. The component on top 210 contains a company
logo graphic 212. Below it is a component 220 containing sports news
stories intended to be of interest to the web site visitor. At the bottom
230 is what is called in the art a "navigation bar" containing hyperlinks
232, 234 to other web pages on the site. In the preferred embodiment, a
hyperlink is defined by HTML (or any other appropriate markup language) as
a point-and-click mechanism implemented on a computer that allows a viewer
to link (or jump) from one screen display where a topic is referred to
(called the `hyperlink source`) to other screen displays where more
information about that topic exists (called the `hyperlink destination`).
A hyperlink thus provides a computer-assisted way for a human user to
efficiently jump between various web pages containing related information.
Hyperlinks can be graphical 234, stylized text 232, or even plain text
224, conventionally formatted with underlining.
In the example of FIG. 2, the small component 240 on the page illustrates
personalized information as provided in the manner of the present
invention. The first line 242 shows an example of `monogramming`, where
the generic information on the page has been customized with information
specific to a particular web site visitor. The next line 244 shows an
example of the results of a personalization directive. The information on
the page has been customized to reflect the fact that this visitor,
preferably based on prior visits, has demonstrated interest in the Round
Rock Rocker's football team; therefore, a custom hyperlink 244 has been
added to the page to provide the visitor with a quick way of obtaining
more information about their favorite team.
The main story component 220 shows another example of personalization.
Visitors interested in football can be shown a set of football stories
221, 223, 225; whereas other visitors may be shown basketball or baseball
stories.
This type of personalization can be achieved in the prior art only by
forcing the user to explicitly answer survey questions and creating
individualized pages. For example, a survey would ask the visitor whether
the visitor preferred to see football or baseball stories, and then ask
the visitor for their favorite teams in order to obtain profile
information. Furthermore, current technology would require that every page
on the web site be generated dynamically for each visitor, which results
in slow response times and poor performance.
The present invention solves the problem of explicit questions and the
performance problem. In the preferred embodiment, the method is
implemented on a web site server. When the web site is being developed,
"Web Content Items" are created by the developers of the web site. Web
Content Items can be an entire web page, a component of a web page, an
insertion into a web page, a graphic link and/or any other items that can
be accessed and viewed by a user. Often times a content item is a
self-contained story or fragment of data; for example, the individual
stories 221, 223, 225 are each a Web Content Item. Web Content Items can
reside at more than one URL. The Web Content Items are preferably defined
through a markup language, including, but not limited to, HTML.
In the preferred embodiment the developer can then assign at least one
category and/or a keyword to each of the Web Content Items. These
categories and key words are used to determine visitor interest when they
access Web Content Items on a Web Site.
In such a preferred embodiment, the developer thereby defines all the
categories that can be used within the system. The categories might be
broad definitions and/or include keywords. The developer can then devise a
set of Web Content Items that can `personalize` the Web Site for the
visitor the next time the visitor accesses the web site. This
personalization can be done according to the accumulated data in the
visitor's file, gathered implicitly by observing which Web Content Items,
and therefore which categories have been of interest to the visitor in the
past. The `personalization` will not be a one-time dynamically generated
customized web page, which would be too resource intensive and therefore
slow, but will be based on predetermined Web Content Items that are
developed and then cached into memory.
The accumulation process functions when a visitor accesses a URL and the
associated Web Content Items. At that point the program registers the
representative categories belonging to the web page. If this is a new
visitor, a new "visitor file" for that visitor is created; otherwise, a
previous visitor file is accessed. In either case, the statistics on the
accessed categories is updated in the visitor's file.
The visitor file contains a Lining tally of the visitor's interest
preferably based on accessed Web Contents Items. In a preferred
embodiment, an algorithm is included that gives greater weight to more
recently accessed Web Content Items, thereby accounting for changing
interests and tastes.
When a visitor accesses a web site that has an existing file for that
visitor, the program determines from the file and the tallied categories,
which pre-customized content, i.e., the personalized page components, to
provide to the visitor.
Such predetermined content is cached in memory and is, preferably, designed
by a web site to appeal to interests in certain topics.
The benefits of the present invention are immediately evident. The present
invention gives the visitor the impression of a customized page visitor
when in actuality it presents pre-customized pages and/or page components
that have been cached. The system thereby conserves computing resources
and retains a higher access speed on a server as opposed to those systems
that dynamically generate customized pages for each visitor.
In the alternative embodiment, the pre-customized pages have at least one
base Web Content Item and insert areas wherein personalized page
components are provided and inserted to make each page appropriate for a
given preference. In another alternative embodiment, the entire page can
be obtained from the cache.
Returning to FIG. 2, the page is illustrative of how a base page is
pre-customized to make it seemingly customized for a given visitor.
Assuming that a visitor frequents a sports-oriented web site in the
preferred embodiment, the main story on the page could be the same for all
the pre-customized pages, for example, a Super Bowl story; however, the
additional stories on the page can be adjusted with inserts of
personalized page components items according to the visitor's preferences,
such as individual team information. Assuming that visitor A in prior
visits has frequented a number of Web Content Items with a keyword of
"football", then when visitor A returns to the web site a page with
personalized page components will appear where the page components (e.g.,
221, 223, 225) are Web Content Items comprising football-related stories.
FIG. 3 shows a relationship diagram for the invention. Requests begin when
a browser 310 operating on a client computer (as in 110 in FIG. 1) makes a
request to the web site server (as in 130 in FIG. 1). When the site is
being accessed, the server request handler 320 analyzes the incoming
request and the corresponding pages, and invokes the monogrammer 330 and
the component assembler 340 as necessary.
The component assembler 340 examines the visitor file, if any, to determine
if there is a preference to be associated with the accumulated categoly
and keyword counts of the visitor. The visitor file is obtained from the
visitor data manager 350, which serves as a central coordination point for
retrievals and updates of visitor data within a single web server. If
there is no file for this visitor, the program generates a file based on
the visitor so as to determine the visitors reference for the next page
requested.
If a visitor file exists for the current visitor, the program accesses such
visitor file to determine the visitor's interests as determined by the
keywords associated with prior Web Content Items served, and, in one
embodiment, there may be a weighing factor or other algorithmic
determination for the additional Web Content Items viewed by the visitor
during the most recent usage. The program then selects a pre-customized
page or pre-customized page components which should reflect this interest.
These selections can be assembled by a component assembler 340, and may be
further subject to personal modification by a monogrammer 330 to make
changes such as inserting the visitor's name onto the page.
The component assembler uses the pre-customized file handler 360, to
retrieve the Web Content Items, formatted as pre-customized pages, that
are appropriate for this visitor. Pre-customized pages can be cached in a
pre-customized file store 365, or can be dynamically generated on demand
by the dynamic page generator 380.
The visitor may select any hyperlink on such page to access additional
interesting content.
In addition, the visitor can still be shown other content not necessarily
directly related to his or her interests. The visitor can still access
these hyperlinks and URLs; therefore, in the preferred embodiment, the
visitor file is an evolving file, since the visitor's interests can change
over time for a number of reasons. Therefore, the present invention can
allow an option to give greater weight to recently accessed Web Content
Items.
The server request handler 320 can then update the visitor file data with
the categories and keyword counts for the information assembled into the
final page that is returned to the visitor's browser. The updated visitor
file data is delivered back to the visitor data manager 350 and stored in
the visitor data file store 375 by the visitor file manager 370.
FIG. 4 shows another embodiment 400 of the invention wherein there are
multiple instances of the Server request handler and associated machinery.
Web sites often use this form of functional replication to achieve higher
performance by sharing the load across multiple server machines. A load
balancer, such as a Cisco Local Director, a DNS round robin, or equivalent
technology exists between the web site visitor's browser 410 and a set of
server request handlers 431, 432, 433. Each server request handler is a
complete copy and typically each one operates on a separate machine. The
server request handlers each have their own visitor data manager 441, 442,
443. As a visitor makes multiple requests to the web site, each individual
request may be redirected by the load balancer to a different request
handle and visitor data manager. Therefore, as the category and keyword
counts are updated by each individual server, some special mechanism must
be used to ensure that updates are not lost by having one set of visitor
data overwrite the results of another. This is the reason for having the
visitor file manager 470 as a separate mechanism within the invention.
There is only one visitor file manager and it serves as the collection
point for all updated data generated by the individual visitor data
managers 441, 442, 443. A further refinement is that the visitor data
managers communicate an incremental update value to the visitor file
manager. For example, consider the case where a visitor makes two requests
to the web site, with each request being for a page containing keyword
"A". The first request might be handled by server request handler 432 (and
visitor data manager 442). The second request might be handled by server
request handler 443 (and visitor data manager 443). Each one of these data
managers has a visitor profile stating that the visitor saw one instance
of the keyword "A". However, when each reports its results back to the
visitor file manager 470, the visitor file manager sums the results
together thus obtaining the correct value of two instances for the keyword
"A". The final results is written into the visitor data file store 475 and
made available for future operations.
It should be appreciated by those skilled in the art that the specific
embodiments disclosed above may be readily utilized as a basis for
modifying or designing other methods for carrying out the same purposes of
the present invention. It should also be realized by those skilled in the
art that such equivalent constructions do not depart from the spirit and
scope of the invention as set forth in the appended claims.
* * * * *
|
|
 |
|
 |
Description |
|
