Exploration of Web search results by the use of Soft operators

Decanato - Facoltà di scienze informatiche

Data d'inizio: 13 Aprile 2010

Data di fine: 14 Aprile 2010

The Faculty of Informatics is pleased to announce a seminar given by Gloria Bordogna

DATE: Tuesday, April 13th, 2010
PLACE: USI Università della Svizzera italiana, room SI-006, Informatics building (Via G. Buffi 13)
TIME: 15.30

This seminar focuses on the well known problem of web searches derived by both the ambiguity of user queries to search engines, and the way search engines present the results through the ranked list, and proposes a novel paradigm to explore the search results based on the application soft manipulation operators.

Most of such queries are very short, consisting of just one or two terms, and as a consequence of this rough and ambiguous description of users' needs, the search engines can hardly guess what users are really searching for. Thus they retrieve and rank documents regardless of the possible distinct semantics of the query terms.
Furthermore, since most users analyze just the first one of two pages of ranked results, they potentially can miss many relevant documents already retrieved in further pages.

In trying to capture relevant documents in the top k ranked positions, users generally adopt two distinct behaviors: some of them choose to submit the same request to distinct search engines, while others prefer searching the Web by iterative trial and error cycles of query reformulation, where at each step, they submit slight variations of the original query, in order to better express their needs.
Nevertheless, the retrieval of distinct lists of documents produces as major effect an information overloading: it may happen that almost the same documents appear in the first positions of all the ranked lists, or in distinct positions.

It would be useful providing the user with a way to compare and explore there results to find out novel, recurrent, and correlated documents.

For some kind of searches, such as for bibliographic searches, the user intent is to obtain new documents, not already seen, and covering the topics of interest in the top positions of the ranked list.
For other searches it can be useful to filter out authoritative web pages, those that are top ranked by most search engines as a result of the same query.

Some users, not very confident in their ability to express their real information needs by a query, try to identify correlated web pages, which may highlight serendipitous information.

In the talk it will be described the proposal of a manipulation language defined for exploring clustered search results: the language comprises several soft operators that can suit distinct user's search intents, allowing re-ranking the web pages retrieved by a single or multiple searches based on distinct criteria, and combining the distinct lists in several ways so as to filter out relevant information.

Gloria Bordogna is a senior research scientist of "Consiglio Nazionale delle Ricerche" (CNR) within "Istituto per la Dinamica dei Processi Ambientali" (IDPA) located in Dalmine (BG),  Italy.
She received the Laurea degree in Physics from "Università degli  Studi di Milano" in 1984.
After a one year scholarship at "Politecnico di Milano"  and at "Olivetti Tecnost  S.p.A" in Ivrea,  in 1986 she joined CNR, where her research activity was devoted to digital image analysis, Information Retrieval and Geographic Information management.
Since 2003, she has been teaching Information Retrieval and Geographic Information Systems at the Faculty of Engineering of "Università degli Studi di Bergamo".
Her research area lies at the crossroads of Information Retrieval, and uncertain and imprecise information management by soft computing techniques.
Her research contributions include the definition of Fuzzy Information retrieval Models, Flexible Query Languages for Information Retrieval Systems and Geographic Information Systems, Soft techniques for Information access and retrieval on the Web, Fuzzy aggregation operators and fuzzy clustering, the results of which have been published in over 150 papers in journals , books and proceedings of international conferences.
She participated in several European funded research projects on the themes of Information retrieval such as e-.court, Peng, Ide-Univers.  She organized several events among which the special Track on "Information Access and Retrieval" of the ACM Symposium on Applied Computing since 2008, and in 2009 the IEEE ACM Web Intelligence and Intelligent Agent Technology Conference.

HOST: Prof. Fabio Crestani