Personalization of the Search
Earlier the web search engine used to return the same ranked list of the web pages based on the search performed irrespective of the user. Most of the returned results were irrelevant to the user. But after some time, search engine started to modify the search result by incorporating the following basic details of the user:
- Current Location, IP Address, Language
But the collected data was not enough to personalize the search result specific to the user. Eventually, the search engine started collecting more information about the user.
Personalized Search[4]
The personalization of the search can also be performed on the user's local system. Each returned result by the search engine can be personalized according to the user's interest.
The information about the user can be collected from the user’s online profile (for search engine side personalization) or from the local system (for client-side personalization) he/she is working on.
On Web Server’s Side
Personalised web search engine ranks the web pages based on the query terms and some additional information of the user. Most of the current search engine modifies the search result based on the following information of the users:
- Location of the User
- Time, Date, Month of the Year
- User’s Browser history, cache
They provide the custom-tailored result for each user. They keep track of the sites visited by the user in the past and other user’s details.
Disadvantages
- Search Engine keeps the record of everything and it might lead to privacy issues.
- Search engines still are not able to provide the top result personalized to the user. A search of “PGM” on any search engine (example Google) would return various wide diversity of results. Example: PGM result on Wikipedia, PGM Précision (French rifle manufacturing company), Pragmatic General Multicast (network protocol), etc. But are these results beneficial for all type of users? A data science student would be more interested in "Probabilistic Graphical Model" rather than PGM Précision or Pragmatic General Multicast. Similarly, a firearm dealer (with Federal Firearms License (FFL)) would be more interested in PGM precision.
Google search result of “PGM”
On User’s side
The result returned by the web search engine without personalization or by accounting the irrelevant user's details can be re-ranked by incorporating the user's interest through the data present in the local system of the user.
Local information present on user's system that helps to incorporate the user‘s interest would consist of the files or the documents viewed by the user in the recent past, email messages sent by the user, events created, links visited by the user in the past, etc. Through the re-ranking on user's side, the privacy of the user is maintained. These information helps to return the top k documents relevant to the user.
Disadvantages
- What if the user performs the query on the remote system? Personalisation would not be possible since the no user’s information is available on the remote system (and if available can cause privacy issues).
- The re-ranking can be performed only on the returned web pages of the web search engine which might be irrelevant to the user.
User's Side Personalization + Web Server's Side Personalization
Both the personalization methods could be merged in order to perform better personalization. First, the personalization would be performed on the web server engine based on the user's profile and again the client system would personalize the web pages returned by the search engine for getting the better user specific result.
Although the privacy issue remains unresolved on the server side, the top k returned result contains web pages of wide diversity which can be re-ranked according to the user’s interest.
Since the returned result contains wide diversity web pages, the user would be able to get the pages related to his/her interest and also have high chances of finding new information.
References
[1] Jaime Teevan, Susan T.Dumais, and Eric Horvitz. Personalizing search via automated analysis of interests and activities. In Proc. of SIGIR, 2005
[2] https://www.redweb.com/agency/blog/personalisation-the-future-of-search-engines
[3] http://tentacleinbound.com/articles/personalized-search
[4] https://seosherpa.com/google-personalised-search/
Comments
Post a Comment