User behavior analysis and relevance extraction modelling

MAJOR PROJECTFINAL

User Behavior Analysis and Relevance Extraction

Modelling

Problem Statement

Personalization of web search experience is so far largely dependent upon user’s activities on that particular search engine. Potential of search result personalization lies in monitoring user’s entire web activities including entire surfing interests and behavior, bookmarks, time spent on particular document etc (which is possible if we monitor user’s activities from client side) rather than just monitoring queries and clicks on search results.

Diagrammatical Representation of Problem

Assumptions

An assumption in this study is that the user is willing to express his/her interests and surfing/searching in a natural way.

Surfing behaviour of a user reflects areas of interests for that user.

Assumptions

A user will not stop searching/surfing world wide web until his information need is satisfied.

Expected Outcomes

1. A mechanism to collect user’s web related information (surfing interests, click/skip behavior, time spent, bookmarks etc), User behavior modeling and develop an algorithm to rank relevance of search results according to this analysis.

Expected Outcomes

2. An algorithm for fusion of relevance based ranking on client side and importance based ranking provided by search engine service and calculating final ranks for documents.

3. An extension of an open source Web Browser to implement above stated functionalities.

Approach to the Solution

Monitor the surfing/searching activities of user:

In order to determine user’s areas of interest and relevance of web pages for that particular user we will monitor all surfing/searching activities of that particular user.

Approach to the Solution

Probabilistic Modeling of User Behavior: Design a probabilistic model to depict user’s

tendency to click and skip search links, taking into consideration relevance and satisfaction factors.

Approach to the Solution

Merge the Relevance and Importance:

In order to personalize web search experience we will merge our relevance factor with search results ranking returned by a conventional search engine.

Algorithm

Tools and Technology

Microsoft Visual Studio

Google Search API

LIB SVM Classifier

JSON API

Implementation Plan

Design a customized web browser:

Implementation Plan

Monitor and Model User’s activities:

Thank You!

A Presentation by:Ambar Gupta (9910103470) F-3

Under Mentorship OfMR. SUDHANSHU KULSHRESTHA

ASSISTANT PROFESSORDEPARTMENT OF CSE AND IT

JIIT SEC 128, NOIDA