I have successfully completed my Ph.D in technology in year 2012. My thesis, titled "Evolution of Analytical Quantified User Dependent Models for improving satisfaction in a search session of Search Engine"
Thesis includes algorithms and techniques for user based evaluation of search results. It investigates several research problems which arise in modern Web Information Retrieval (WebIR). The Holy Grail of modern WebIR is to find determine user satisfaction for learning its future expectation and find a way to organize and to rank results accordingly. According to the research done in this direction “To judge user behaviour his intention needs to be identified. In many situations users may have different search goals even when they express them with the same query, users experience with a session may impact their future trend of using a search session. In order to address this need we will discuss the algorithmic idea behind the User Dependency Model which predicts future dependency of the user on Search Engine for informational needs. The user Dependency model developed uses the concept of Fuzzy logic to determine quantitative value of user dependency. The model extends the search capabilities of existing methods and can answer more complex search requests. There are also situations where users might desire to access fresh information. In such cases Search Engine updated ness, its technological innovations to match user expectation play an important role. In these cases it’s important to acquire user opinion regarding various capabilities of Search Engine and judge his future intention of using Search Engine for similar needs. To measure this currently available feedback models may not be useful. In order to address this necessity, we will discuss the algorithmic and numerical ideas behind a User Dependency algorithm which is suitable for acquiring user intent and can be used to rank search results as per user’s opinion.
Existing search services rely solely on a query's occurrence in the document collection to locate relevant documents. They typically do not perform query based user behavioral and do not leverage changes in user query patterns over time. User search pattern was identified to differ for various types of queries. ‘Tendency to click even when search is successful’ was observed as the most common behavior for open ended informational/ transactional queries. Provided within are a set of techniques and metrics for performing user behavioral based analysis on query logs. During this research work a rule based system has been proposed which is based on user click behavior. This system contains a set of rules which determine query based satisfaction of user for a given session. User Behavior Analysis for different type of queries has been conducted and their satisfaction with the session has been predicted using this rule based system. The metrics proposed for our rule based analysis are shown to be reasonable and informative, and can be used to detect changing patterns of user for different informational needs, thus providing valuable data to a search service. We continue with an algorithm for automatic rule based detection of User Satisfaction for different type of queries. Results are presented showing that our detection approach can be successfully applied to a significant portion of the query logs, making it possible for search services to leverage it for improving search effectiveness and efficiency.
Thesis includes algorithms and techniques for user based evaluation of search results. It investigates several research problems which arise in modern Web Information Retrieval (WebIR). The Holy Grail of modern WebIR is to find determine user satisfaction for learning its future expectation and find a way to organize and to rank results accordingly. According to the research done in this direction “To judge user behaviour his intention needs to be identified. In many situations users may have different search goals even when they express them with the same query, users experience with a session may impact their future trend of using a search session. In order to address this need we will discuss the algorithmic idea behind the User Dependency Model which predicts future dependency of the user on Search Engine for informational needs. The user Dependency model developed uses the concept of Fuzzy logic to determine quantitative value of user dependency. The model extends the search capabilities of existing methods and can answer more complex search requests. There are also situations where users might desire to access fresh information. In such cases Search Engine updated ness, its technological innovations to match user expectation play an important role. In these cases it’s important to acquire user opinion regarding various capabilities of Search Engine and judge his future intention of using Search Engine for similar needs. To measure this currently available feedback models may not be useful. In order to address this necessity, we will discuss the algorithmic and numerical ideas behind a User Dependency algorithm which is suitable for acquiring user intent and can be used to rank search results as per user’s opinion.
Existing search services rely solely on a query's occurrence in the document collection to locate relevant documents. They typically do not perform query based user behavioral and do not leverage changes in user query patterns over time. User search pattern was identified to differ for various types of queries. ‘Tendency to click even when search is successful’ was observed as the most common behavior for open ended informational/ transactional queries. Provided within are a set of techniques and metrics for performing user behavioral based analysis on query logs. During this research work a rule based system has been proposed which is based on user click behavior. This system contains a set of rules which determine query based satisfaction of user for a given session. User Behavior Analysis for different type of queries has been conducted and their satisfaction with the session has been predicted using this rule based system. The metrics proposed for our rule based analysis are shown to be reasonable and informative, and can be used to detect changing patterns of user for different informational needs, thus providing valuable data to a search service. We continue with an algorithm for automatic rule based detection of User Satisfaction for different type of queries. Results are presented showing that our detection approach can be successfully applied to a significant portion of the query logs, making it possible for search services to leverage it for improving search effectiveness and efficiency.