| Abstract |
In a selective information dissemination (SDI) system, users submit profiles consisting of a number of long standing queries to represent their information needs. The system then continuously collects new documents from underlying information sources, filters them against the user profiles, and delivers relevant information to corresponding users. SDI systems are very important nowadays due to the vast amount of information that flows in the World Wide Web, as they inform users for relevant information, without requiring them to spend time to locate it. The majority of SDI systems are based on lexical search. In particular, they represent documents and user profiles as sets of terms and check the identicalness between these two sets, in order to make the decision for sending a document to a user. Users of SDI systems may use many different terms to express the same meaning, terms that are called synonyms. Simultaneously, when users seek information about a term, they are also interested in information about terms that are hyponyms (special terms) of this initial term. As a result, it is necessary for such a system to contain mechanisms that take into account the semantic relationships between terms during matching of user profiles with documents. In the present thesis, an SDI system has been implemented, which takes into account the semantic relationships between terms. In particular, a user profile is considered relevant with a document, if its terms or the synonyms or hyponyms of them appear in the document. The system deals with profiles that are represented in the two most popular models in information retrieval, namely the Boolean model and the Vector Space model. In order to improve the systems performance, an index structure of profiles rather than of documents has been created, as profile information constitutes a larger volume and is more static. When a document arrives from the information sources, a matching algorithm for the document and the profiles in the index structure is executed. This algorithm takes into account the semantics of terms. Finally documents that contain the terms of a profile or relative terms of profile terms are delivered to user. The system has been evaluated based on experiments that have been conducted.
|
Διασπορά πληροφορίας βασισμένη σε σημασιολογικές συσχετίσεις
