GB2346228A

GB2346228A - Intelligent adverts

Info

Publication number: GB2346228A
Application number: GB9901666A
Authority: GB
Inventors: Adrian Mark Colyer
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 1999-01-27
Filing date: 1999-01-27
Publication date: 2000-08-02
Also published as: GB9901666D0

Abstract

A gene server 25 which communicates across the Internet with a client computer 10, comprises a database 90 storing gene sets, each gene set comprising parameter values defining the display of an advertisement on a client computer. The server includes a script 70, responsive to a request from the client, for providing a gene set from the database to the client. A further script 80 is responsive to a report from said client for storing if an advertisement displayed according to the gene set has been selected by a user. A mutation program 100 periodically mutates the gene sets according to the reports of user selection. A MouseEvent class may be incorporated in an applet 60 for causing an event such as the mouse cursor entering or exiting the advert boundary to call a command for the web browser 40 to jump 9a to the advertiser's home page 120. When the applet 60' is instantiated from within the browser, a tracking class 65 is downloaded and instantiated. The tracking object maintains a timer which may be halted, for example when the user jumps to a completely independent web page. In response, the timer value and the user's failure to select the advertisement are reported to the server.

Description

INTELLIGENT ADVERTS The present invention relates to the field of genetic programming applied to advertising in a distributed fashion from a server to a plurality of clients which are connected together by a computer network, such as the Internet.

Advertising via the Internet permits the targeting of huge audiences with the minimum of effort. It is now quite common for businesses to have their own web sites providing information about themselves and generally aimed towards increasing visibility.

Furthermore, certain well-known organisations charge companies for advertising space on their server. The search engine Yahoo! is one such example, which is a frequently visited site likely to provide businesses with the sort of population coverage that they demand.

In this regard, US patent 5,796,952 discloses a tracking program for determining the success of an advert on the Internet, as well as a means for storing the information gathered and is incorporated herein by reference.

Advertising is an art, not a science. Banner adverts on the WWW come in a variety of shapes and sizes. The essential technology involves the display of an image on a web page, which when clicked upon links to the advertiser's own website. In all cases, the format of the advertisement (it's appearance) is manually decided in advance by the advertiser. Look and feel holds the key to how successful a particular campaign will be. Advertisers often do trial runs of various versions of the advert, changing various aspects such as key messages, colour, size and positioning in order to see which has the best response.

Advertising is however subjective and will vary according to the audience at which it is being targeted. Furthermore, such an industry is extremely dynamic. Tastes are continually changing and it is therefore difficult to predict which adverts will be the most successful in any given period of time.

WebConnect (http ://www. webconnect. com/newmedia/WhatWeDo. asp) have disclosed an innovative advertisment placement service. The service assists the advertiser in deciding whether changes are necessary to an advertising campaign in order to achieve maximum return on investment.

Nonetheless, it remains necessary to be able to control an advertising campaign as frequently as possible and with an increasing degree of target audience granularity.

Accordingly, the invention provides a server according to claim 1 and an advertising component according to claim 7.

The invention provides for an advertiser a system by which their advertising across the Internet, in particular, can be optimised. Each individual placing (advert host URL) for a given advert can be adapted over time to its most effective form for the target audience viewing that site. This most effective form may be different for the same advert displayed at different host sites, because of the differences in audience at those sites.

Furthermore, the body responsible for evolving the advertisement (i. e. the server owner) is a distinctly separate entity. It is not tied to any one advertiser and may therefore be charged with evolving advertisements for a plurality of vendors.

Nonetheless, the means to produce a truly intelligent machine, capable of thinking in the same way as a human-being, is still a little way off, however computers are becoming ever more capable of solving increasingly perplexing problems. Genetic programming and genetic algorithms provide a computational optimisation strategy for those problems which are difficult to solve using conventional mathematical techniques. Such a strategy is especially applicable to those problems which have a large or infinite number of possible solutions, thus hindering any attempt at formalisation.

Genetic algorithms mimic the evolutionary process of natural selection. A random population is generated, to give an initial subject population and each member of the population is evaluated according to some fitness function. The more successful members of the population are "bred", by applying mutations and cross-over functions to their genes, whilst the weaker members are destroyed. The process is iterative over time, with each cycle producing a new generation of the population until a predetermined generation or fitness level has been achieved. The aim is to reach an optimum configuration, propagating highly robust individuals capable of solving increasingly complex situations. Just as in the evolutionary process of nature, there is no purpose or direction.

Genetic algorithms can provide for an environment in which individuals compete with one another for sought after resources and where only the fittest survive.

Genetic programming has been applied to a wide variety of diverse fields within the realms of artificial intelligence. A typical example is in the area of robotics, where it is especially suitable, since it is important that a robot is capable of adapting to both a continually changing environment and the needs of different users. Further industrial applications of genetic algorithms are described in"Genetic based Machine Learning-Applications in Industry and Commerce"by Terence C Fogarty, Neil S Ireson, and Lawrence Bull, in the Proceedings of the Conference on Adaptive Computing and Information Processing, in the UK, 25-27 Jan 1994, vol 2, page 509-530.

The Internet has also been widely used as a means for simulating the evolutionary process. It is possible to download a number of programs which incorporate genetic algorithms."ComputerAnts"is one such program, available from http://www. axcelis. com, which enables one to generate a population of red ants and observe the way in which they interact with one another. Just as if part of the natural living world, the ants can mate and reproduce. The genes attributed to each offspring will determine how successful their behaviour is. Only those who eat well and avoid poison will survive long enough to pass their good qualities on to the next generation of ants.

Genetic programming on the Internet has also been used to evolve new generations of sibling images according to user preference.

DancerDNA, found at http://www. dancerdna. com, is one such example. This program will generate surreal figures capable of dancing in time to music and allow one to mutate the species in order to achieve a population which best fits the type of music being played. Apparently, a somewhat similar approach has been taken in an attempt to generate aesthetically pleasing images, in which users are able to rank pictures in order of preference. The results are then recorded on the web server which uses this information to determine the subsequent generation of images.

It is therefore known to employ genetic algorithms to evolve images displayed over the Internet. However, in the case of the prior art, the user was aware that they were participating in a genetic experiment and that their subjective judgement was being used to spawn the next generation of images. The use of such a strategy in which it is the user's subconscious preference towards a particular image/set of images which is used to determine subsequent generations of images is not known.

Furthermore, neither is the use of genetic algorithms in the evolution of advertisements over the Internet known.

Canadian Patent Application No. 1338423 teaches a method for presenting on a local client machine, a page comprising a plurality of applications displayable in a first portion of the screen and an advertisement displayable in a second portion of the screen. A data collection manager gathers information regarding amongst others, service interaction, user demographics and geographical location and uses this information in order to supply the user with individualised advertising in the second portion of the screen. However, the data collection manager has no means of determining the success of a particular advert within a third-party page; evolving the advert based upon such a determination; or jumping to an advertiser's site.

There is the need for a mechanism by which the various permutations of an advert may be automatically evolved, according to user preference, in order that the advert reaches its most successful form.

Thus, in a further aspect the invention provides a server as claimed in claim 2 and an advertisement component as claimed in claim 8.

The invention caters for an audience that changes over time, since the advert evolves automatically to match. Just as in nature, the evolution of genetic adverts is on-going. The process never halts, since users'expectations and tastes are continually changing.

A preferred embodiment of the invention will now be described in detail by way of example only with reference to the following drawings: Figure 1 represents a schematic diagram of a preferred embodiment of the present invention.

Figure 2 illustrates the gene sets mutation process.

Figure 3 depicts the mutation process in more detail.

Figure 1 illustrates a network in which a plurality of clients 10 (only 1 shown) running web browsers 40 access a page 50 stored on a server 20 over the Internet or World Wide Web (WWW). Conventionally, such a page 50 includes a reference to an advertising banner comprising an applet 60 also stored on the server 20. The page 50 is downloaded to the browser at step la, the applet 60 is instantiated 60'at step 2a to display the advert on the client machine. A device commonly known as a "mouse listener"can be used register any interaction with the advert.

Thus, a MouseEvent class is incorporated in the applet 60 to record events such as the mouse cursor entering/exiting the advert boundary, the user pressing a mouse button etc.. The applet listens for a mouse click event and is thus responsive to the user clicking on the advert to call a showURL () command. This typically causes the browser to jump at step 9a to the advertiser's home page 120 stored on the advertiser's server 110.

Typically, the advert on the banner ad is encoded as a Java applet, using technology available from Sun Microsystems Inc.. A java applet is a computer program embedded in an html web page, which can be downloaded in the form of byte code from a server and executed locally on the client machine. It is platform-independent and hence provides web designers with a great degree of flexibility. It enables them to develop more sophisticated web pages, incorporating animation, sound, special effects etc.. Most importantly, a java applet is capable of modifying its own appearance and transmitting data between the web browser and the server.

The present invention makes use of the conventional structure of a Java applet which comprises a class having the following four methods; i) init () ii) start () iii) stop () iv) destroy () In the present embodiment, the Init () method of an advert class is called, once only, upon the instantiation of the applet. It is responsible for receiving any parameters which can then be used to configure the advert to be displayed by the applet. Start () is first called by init () and thereafter whenever the applet becomes visible.

According to the preferred embodiment it causes the advert to be displayed within the applet. Stop () is instantiated whenever the applet becomes invisible. This could be as a result of: i) Scrolling down the page; ii) Closing the browser; iii) jumping to a completely independent web page; or iv) jumping to the advertiser's own web site Destroy () releases any resources that the applet is using and is called some time after stop ().

According to the present invention, the advert itself comprises a number of variable components making up a set of parameters defining the display of the advertisement. In the preferred embodiment, these include: i) The key message text ii) The font and font size iii) The text colour iv) The background colour v) The sound effect (s) vi) Animation (on/off) vii) Delay lengths in the appearance of advert elements viii) Animation rate It is up to the advertiser to decide upon the most appropriate parameters to use within a gene set and it should therefore be understood that the invention is not limited to the above. It is these genes that describe the various advert permutations and comprise the data received by the init () method.

Conventionally, in order to invoke the applet advert or any other Java applet, a Web page of html text contains an < APPLET > tag, which identifies the URL containing the applet. A browser responds to this tag by retrieving and running the applet. Also defined is a < PARAM > tag, which is contained within a pair of corresponding < APPLET > and < /APPLET > tags, and which can be used to specify parameters which are passed to the applet at run-time. (Note that the APPLET and PARAM tags are not formally incorporated into the html standard, but are nevertheless recognised by many Web browsers). Further information about the Java programming language may be obtained from"Teach Yourself Java in 21 days"by Laura Lemay and Charles Perkins (Sama. net Publishing, Indianapolis, USA, 1996).

In the preferred embodiment, the HTML code for invoking the applet advert looks as follows: < APPLET CODE="Advert. class" WIDTH=int HEIGHT=int > . The PARAM tag is also used in order to specify the address of a gene server 25 with which the applet is to communicate: < PARAM NAME "servername"VALUE="serveraddress" > According to the preferred embodiment, the gene server 25 creates a random population of genes for each combination of advert and host-site.

The gene server includes a database 90. The database comprises a plurality of records, each having a number of attributes, inter alia; i) The URL of the page in which the advert is incorporated ; ii) Its associated gene set; and iii) Results for the gene set.

In the present embodiment, the HTML page is requested and the advert 60' instantiated at steps 1 and 2 in a conventional manner. When the banner advert 60 is downloaded from the host-site, the gene server is contacted at step 3 and a gene set &num;1 requested at step 4. The applet does this by issuing an HTTP GET, or an equivalent HTTP POST, with a URL of the form : http://www. geneserver. com/getGenes ?"advert~host~pageW This URL invokes a Common-Gateway Interface (CGI) script or servlet program, GetGenes 70, running on the gene server 25. The script 70 is passed as an argument the URL of the page 50 in which the advert is incorporated,"adverthostpage". The GetGenes program uses the URL to: i) reference the correct entry in the gene sets'database 90; ii) request the current gene set at step 4; and iii) return this set to the applet at step 5.

The instantiation 60'of the banner ad applet 60 running on the web browser then interprets this gene set and uses the set as parameters to configure its display attributes.

An additional function of the gene server is to record the success of the adverts that its gene sets depict. According to the preferred embodiment, a tracking class 65 resides on the ad-hosting server 20.

When the applet 60'is instantiated from within the web browser on the local client machine, the tracking class is downloaded at step 2c and instantiated via the start () method at step 6.

The tracking object maintains a timer and in the preferred embodiment, the timer is halted via the stop () method when the user either: i) Scrolls down the page; ii) jumps to a completely independent web page; iii) jumps to the advertiser's own web site; or iv) Closes the browser With regard to i) and ii), the applet waits a predetermined amount of time before reporting failure to the gene server. Should the user either scroll back up the page or return to the advert-hosting page, the start () method of the applet is responsible for restarting the timer again. It is the total accumulated time that is eventually reported back to the gene server.

The total time that the applet was displayed and if the advert was successful, i. e. whether or not it was clicked upon, is reported to the gene server at step 7. In this manner, the gene server can build over time a measure of : a) The number of hits per ad display; or b) The number of hits per minute of display Success or failure is reported to the gene server by posting to a URL of the form: http ://www. geneserver. com/geneReport? genes="genes" & time="seconds" & success ="true/false" Should the web browser leave the page without clicking on the banner ad, the advert traps this in the"stop ()" method of the applet, and then submits a geneReport with the success parameter set to false. Should the user click on the banner ad, the advert detects the mouse event, and then submits a geneReport with the success parameter set to true.

The URL includes three parameters passed to geneReport : the gene set the advert used, the time it was displayed, and whether or not it was clicked on.

These parameters are stored at step 8 as results in this attribute of the gene sets'database 90. The GeneReport program preferably feeds back to the applet that all the information has been successfully recorded at step 9 and the browser can then either link at step 9a to the advertiser's site 110 or to an independent page at step 9b, depending on the course of action taken by the user.

It is vital to note that the exact format of the getGenes and geneReport URLs is not important, but the fact that the communication with the gene server is by URL is important, as an applet is not permitted any other form of communication to a host other than the one from which it was downloaded (the advert hosting site).

In the preferred embodiment, each gene set is allocated 20 minutes of display time. Those with the greatest number of hits recorded during this timeslot are deemed to be the most successful and are used to evolve the subsequent generation. According to the preferred embodiment, the information within the gene sets'database 90 is passed to a mutation program at step 9c which discards the unsuccessful adverts, whilst genetically evolving the successful ones. The newly evolved gene sets are returned to the database at step 9d.

According to an alternative preferred embodiment the database stores not only the current results for each gene set, but also the results for the previous generation of gene sets. In this way the mutation program can perform a comparison between the two and determine the success of the mutation process. It can then initiate"rollback" should the need arise.

In the preferred embodiment the mutation process is shown by Figure 2 and works as follows: The 40 most successful gene sets (adverts) are identified and a subset (two) randomly extracted at steps 300,301. The top half of genes from the 1st set 100.1 are swapped with the top half from the 2nd set 100.2 at steps 302,303 and finally, a random disturbance factor is applied. This process is repeated for all further subsets until a new generation of gene sets has evolved.

Figure 3 depicts the effect of mutation on one single subset. Gene sets 100. la and 100.2a are the result of this process. It should be noted that the due to the random disturbance factor, there is not a direct correlation between the original subset and the two newly evolved gene sets. However, according to the preferred embodiment the difference is minimal.

According to the preferred embodiment, the level of response by web browsers to a particular image in a page, relative to its siblings viewed by other browsers, is also used to determine how much the host site should charge for displaying the advert in question. The information illustrating the success of each advert is obtained from the database 90 mentioned above. Cost information can then be stored in separated billing database (not shown) on the gene server.

The advantages of this technique are that the GetGenes program can separate this advert from additional copies hosted at other sites. Each individual placing (advert host URL) for a given advert can evolve over time to its most effective form for the target audience viewing that site. This most effective form may be different for the same advert displayed at different host sites, because of the differences in audience at those sites. This genetic technique also copes with an audience that changes over time, as the advert evolves to match. Just as in nature, the evolution of genetic adverts is on-going. The process never halts, since users'expectations and tastes are continually changing.

It should be understood that although the invention is described here in the context of an advert encoded as a Java applet, a person skilled in the art will realise that the invention is not limited to such and that the advert may be encoded, for example, using ActiveX technology available from the Microsoft Corporation.

Claims

CLAIMS 1. A server adapted to communicate across a network with a client computer, said server comprising: means for generating a set of parameter values defining the display of an advertisement on said client computer; means, responsive to a request from said client, for providing said set to said client; results means, responsive to a report from said client, for storing if an advertisement displayed according to said set has been selected by a user; and means for periodically updating the set of parameters provided to a client according to said reports of user selection.
2. A server according to claim 1 wherein said generating means generates a plurality of sets of parameters values and stores said sets as a generation of gene sets and said means for periodically updating mutates said generation of gene sets to generate a next generation of gene sets.
3. A server as claimed in claim 1 wherein said providing means comprises a CGI script.
4. A server as claimed in claim 1 where said results means comprises a CGI script.
5. A server as claimed in claim 1 wherein said client report comprises a timer value and an indication of whether or not said user has selected said advertisement.
6. A server as claimed in claim 2 wherein said client report is stored in a database comprising an entry for each gene set, and wherein each entry comprises: the address of said first web page; and client reports for said first web page.
7. An advertisement component instantiable from a first web page adapted to be displayed within a web browser on a client computer in a network, said component comprising: means for requesting from a server a set of parameter values; means, responsive to receiving said set of parameter values defining the display of said advertisement on said client computer, for displaying said advert within said web page according to said set; and means, responsive to said client selecting said advertisement, for reporting said user selection to said server and for causing said client computer to display a second web page located at an advertisers web server.
8. An advertisement component as claimed in claim 7 wherein said set of parameter values is a set of genes from a generation of gene sets.
9. An advertisement component as claimed in claim 7 wherein said component is an applet comprising Java byte code.
10. An advertisement component as claimed in claim 7 wherein said component is an Active-X component.
11. An advertisement component as claimed in claim 7 comprising means for instantiating a first process on said server and for receiving said set from said first process.
12. An advertisement component as claimed in claim 11 wherein said first process comprises a CGI script.
13. An advertisement component as claimed in claim 7 wherein said first web page is located on a web site remote from said advertiser's web server.
14. An advertisement component as claimed in claim 7 wherein said first web page is located on a web server remote from said server.
15. An advertisement component as claimed in claim 7 comprising: means, responsive to said browser jumping to display said first web page, for starting a timer; and means, responsive to said browser jumping to display said another web page, for reporting the value of said timer and said user's failure to select said advertisement to said server.
16. An advertisement component as claimed in claim 15 comprising: means, responsive to closing of said browser during display of said first web page, for reporting the value of said timer and said user's failure to select said advertisement to said server.
17. An advertisement component as claimed in claim 15 wherein said reporting means is adapted to report the value of said timer when reporting said user selection to said server.
18. An advertisement component as claimed in claim 15, wherein said reporting means is responsive to said browser returning to display said first page for re-starting said timer.
19. An advertisement component as claimed in claim 15, wherein said reporting means is responsive to said browser scrolling said first web page to display said advertisement for re-starting said timer.
20. An advertisement component as claimed in claim 15 comprising means for instantiating a second process on said server and transmitting said timer value and said user selection value to said second process.
21. An advertisement component as claimed in claim 20 wherein said second process comprises a CGI script.
22. A method operable in a server adapted to communicate across a network with a client computer, said method comprising the steps of: generating a set of parameter values defining the display of an advertisement on said client computer ; responsive to a request from said client, providing said set to said client; responsive to a report from said client, storing if an advertisement displayed according to said set has been selected by a user; and periodically updating the set of parameters provided to a client according to said reports of user selection.
23. A method of advertising in a first web page adapted to be displayed within a web browser on a client computer in a network, said method comprising the steps of: requesting from a server a set of parameter values ; responsive to receiving said set of parameter values defining the display of said advertisement on said client computer, displaying said advert within said web page according to said set; and responsive to said client selecting said advertisement, reporting said user selection to said server and for causing said client computer to display a second web page located at an advertisers web server.
24. A computer program product comprising computer program code stored on a computer readable storage medium for, when executed on a computer, generating, providing and updating sets of parameters for advertisements, the program code comprising a server as claimed in claim 1.
25. A computer program product comprising computer program code stored on a computer readable storage medium for, when executed on a computer, displaying an advertisement, the program code comprising an advertising component as claimed in claim 7.