Microsoft is engaged in "demographic predictions"
The demo version of the system of demographic forecasts was created at the center of adCenter Labs . Since its founding in January 2006, he has been developing an innovative advertising platform, MSN adCenter.
Enter any URL or keyword into the form, and the demographic prediction system will immediately show how much “male” or “female” this search query is in percent, as well as what is the distribution of users by age. In addition to the “general” age distribution, the system also predicts how this distribution will change in the future.
For example, for the search query [xbox], a male audience (71%) is shown, mostly young.
')
On the page with the demo version there is not a word about how it works. It is only said that for analysis the MSN Search search query database for one month is used.
Testing the system allows, nevertheless, to make some assumptions about the algorithm of work. First, it is clear that the system is not able to analyze absolutely any request. For example, it does not cope with meaningless character sets: honestly writes that there is no such query in the database. Thus, it can be assumed that as the result of the analysis, just statistical information from the database is displayed.
Apparently, to compile a demographic system, the developers took not only the database of search queries, but also information about the linking of requests to users authorized by MS Passport (with indication of gender and age). Thus, a sex-age distribution was obtained for all search queries, which is demonstrated to us. What technologies are used to predict changes in the audience in the future is not clear.
The database contains Russian-language queries as well. True, one can doubt the representativeness of the sample of Russian-speaking users registered in the MS Passport system. For example, [lipstick] is recognized as a male request (71%). In English, everything is normal: [lipstick] - 66% of women. Interestingly, the query [vodka] interests more women, as well as [blogging]. And the request [tampon], on the contrary, is considered masculine.
In addition to analyzing search queries, the system allows you to analyze the age and gender of various sites. For example, for our site, the system shows a male audience (76%) aged 35-49 years.
By the way, on the site adCenter Labs posted more than a dozen other demo versions . There are also many interesting things there: for example, graphical visualization of “search tunnels” , that is, the most typical ways of query evolution during one search session.
Enter any URL or keyword into the form, and the demographic prediction system will immediately show how much “male” or “female” this search query is in percent, as well as what is the distribution of users by age. In addition to the “general” age distribution, the system also predicts how this distribution will change in the future.
For example, for the search query [xbox], a male audience (71%) is shown, mostly young.
')
On the page with the demo version there is not a word about how it works. It is only said that for analysis the MSN Search search query database for one month is used.
Testing the system allows, nevertheless, to make some assumptions about the algorithm of work. First, it is clear that the system is not able to analyze absolutely any request. For example, it does not cope with meaningless character sets: honestly writes that there is no such query in the database. Thus, it can be assumed that as the result of the analysis, just statistical information from the database is displayed.
Apparently, to compile a demographic system, the developers took not only the database of search queries, but also information about the linking of requests to users authorized by MS Passport (with indication of gender and age). Thus, a sex-age distribution was obtained for all search queries, which is demonstrated to us. What technologies are used to predict changes in the audience in the future is not clear.
The database contains Russian-language queries as well. True, one can doubt the representativeness of the sample of Russian-speaking users registered in the MS Passport system. For example, [lipstick] is recognized as a male request (71%). In English, everything is normal: [lipstick] - 66% of women. Interestingly, the query [vodka] interests more women, as well as [blogging]. And the request [tampon], on the contrary, is considered masculine.
In addition to analyzing search queries, the system allows you to analyze the age and gender of various sites. For example, for our site, the system shows a male audience (76%) aged 35-49 years.
By the way, on the site adCenter Labs posted more than a dozen other demo versions . There are also many interesting things there: for example, graphical visualization of “search tunnels” , that is, the most typical ways of query evolution during one search session.
Source: https://habr.com/ru/post/4831/
All Articles