National and Kapodistrian University of Athens
Εξατομίκευση επερωτήσεων σύμφωνα με τις προτιμήσεις του χρήστη
Abstract
dc:descriptionThis thesis studies Query Personalization, a process of dynamically enhancing a query with preferences stored in a user profile with the purpose of generating a customized focused and hopefully smaller answer for a specific user. We present a preference model that combines expressivity and concision and provides a direct way to personalize queries. User preferences are stored as degrees of interest in atomic query elements (individual selection and join conditions), which may be used to transform a query. The degree of interest expresses the interest of a person to include the associated condition into the qualification of a given query. Specific logic is introduced for derivation of preferences combining stored atomic ones. We propose a personalization framework according to which personalized results should satisfy at least L out of the top K most important preferences that are related to the given query. Parameters K and L can be specified with the use of various criteria. Personalized results are ranked based on the estimated degree of interest in the combination of preferences they satisfy. Preference selection from a structured user profile is formulated as a graph computation problem. Efficient algorithms that derive the top K preferences from a user profile that are related to a query are described. The top K preferences derived from the user profile may be integrated into the initial query so that the resulting one should return results satisfying at least L from the top K preferences. Alternative query rewriting schemes have been examined. A specialized algorithm that allows for progressive generation of personalized results is also proposed. Query personalization is formulated as a constrained optimization problem, where constraints may concern the personalized query execution time and result size. This is the first realistic approach to personalization that takes into account real-time factors and conditions in order to return personalized results. Constrained query personalization is formulated as a state space search problem. Each personalized query that is a potential solution is a state in a space (a node in a graph), characterized by degree of interest, execution cost, and result size. Transitions (edges) arrange states in partial orders based on each of the aforementioned query parameters. State-space search algorithms are devised that take advantage of these partial orders to solve these problems efficiently
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2005
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Koutrika, Georgia
- Κούτρικα, Γεωργία
Subjects
dc:subject × 14- Εξατομίκευση
- Εξατομικευμένη αναζήτηση
- Μοντελοποίηση χρηστών
- Χρήστες, Προφίλ
- Μοντελοποίηση προτιμήσεων
- Personalization
- Personalized search
- User modeling
- User profiles
- Preference modelling
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/22576