Web search personalisation by Shashank Gupta

10

Click here to load reader

description

Personalise your search using browser on client side predict intentions of user from history and display results based on these intentions.

Transcript of Web search personalisation by Shashank Gupta

Page 1: Web search personalisation by Shashank Gupta

WEB SEARCH

PERSONALIZATION

On the client side

Page 2: Web search personalisation by Shashank Gupta

WHY PERSONALIZATION ?

• Apple – Company Or Fruit

• Mouse– Animal Or Peripheral Device

Page 3: Web search personalisation by Shashank Gupta

WHY PERSONALIZATION ?

• Apple – Company Or Fruit

• Mouse– Animal Or Peripheral Device

Page 4: Web search personalisation by Shashank Gupta

WHY PERSONALIZATION ?

• Apple – Company Or Fruit

• Mouse– Animal Or Peripheral Device

Automatically animal is

crossed out

Page 5: Web search personalisation by Shashank Gupta

WHAT IS ALREADY EXISTING

• Google personalizes if you are logged in.

• Can start personalizing only when your session starts to a search engine(SE).

• How much time a user is browsing a particular page can’t be deduced.

• Page relevance criteria is based only on those WebPages that you access using SE.

Page 6: Web search personalisation by Shashank Gupta

WHAT IS THE SOLUTION!!!

• It is the Browser that knows about user’s intentions.

• So better shift the focus from server side to client side

• Keep track of all the web pages visited and rank them on the basis of their relevance

• Use criteria like Browse time & Download to give them relevance

• After making a set of such pages extract keywords from them.

Page 7: Web search personalisation by Shashank Gupta

KEYWORDS ????????

• Words of high frequency.

• Words in title, subtitle, meta tags and in the starting paragraph of page.

• Should not be STOP words i.e. words like “a, an, the” which are of high frequency.

Page 8: Web search personalisation by Shashank Gupta

CATEGORY DEDUCTION

• A model that will give category classification such that we have C1, C2, C3, C4 ……

• C1, C2, C3, C4 should span the entire web search query possibilities.

• C1, C2, C3, C4 should be independent.

• Apple – computers/hardware or living/food & cooking

• Mouse - computers/hardware or wildlife/nature

Ex .

Page 9: Web search personalisation by Shashank Gupta

CATEGORY DEDUCTION (CONTD.)

• So once keywords are made deduce all possible categories for them.

• Now take intersection of them i.e. pick out common from them.

• if we take intersection of Mouse and Apple we get – computers/hardware

• Now we know what user intentions actually are

Ex.

Page 10: Web search personalisation by Shashank Gupta

NEXT TIME WHEN USER GOES TO GOOGLE….

• Type query

• Results come to browser

• Give an option to rank the result according to user intentions

• If yes, send the retrieved result + category deduced earlier to the server hosted

• Server will re-rank them according to the category

• This re-ranked page is what user will see eventually.