Carmine Cesarano

An intelligent search agent system for semantic information retrieval on the internet

In this paper we describe a prototype system for information retrieval on the Internet. Our idea is that the Web has to be searched both semantically and syntactically. In order to automatically categorize the web pages on the fly we propose a novel approach based on ontology and semantic networks and we describe a prototype system based on the Intelligent Agent Paradigm. Preliminary experiments are shown and discussed while describing open problems and on-going research.

Story creation from heterogeneous data sources

There are numerous applications where there is a need to rapidly infer a story about a given subject from a given set of potentially heterogeneous data sources. In this paper, we formally define a story to be a set of facts about a given subject that satisfies a “story length” constraint. An optimal story is a story that maximizes the value of an objective function measuring the goodness of a story. We present algorithms to extract stories from text and other data sources. We also develop an algorithm to compute an optimal story, as well as three heuristic algorithms to rapidly compute a suboptimal story. We run experiments to show that constructing stories can be efficiently performed and that the stories constructed by these heuristic algorithms are high quality stories. We have built a prototype STORY system based on our model—we briefly describe the prototype as well as one application in this paper.

A system for summary-document similarity in notary domain

In this paper we propose a methodology to perform a comparison between a legal document and its related handwritten summary. We thus describe the algorithms that verify when a human-produced summary is consistent with its source document. We first analyze both documents in order to extract only the relevant information then we compare such information in order to obtain a measure of correlation indicating the consistence between the two documents. Eventually, we briefly report on the performance of these algorithms.If we try to increase the level of automation in enterprise application integration (EAI) scenarios, we confront challenges related to the resolution of data heterogeneities, service discovery and process composition. In this paper, we propose the enterprise interoperability ontology (ENIO) that provides a shared, common understanding of data, services and processes within enterprise application integration scenarios. ENIO consists of an upper EAI ontology, which is based on the DOLCE-SUMO alignment, with extensions called facets that cover several dimensions of the EAI domain. Each facet contains a relative meta-model that utilizes widely adopted standards.

Improving Usability of Web Pages for Blinds

Warranting the access to Web contents to any citizen, even to people with physical disabilities, is a major concern of many government organizations. Although guidelines for Web developers have been proposed by international organisations (such as the W3C) to make Web site contents accessible, the wider part of todays Web sites are not completely usable by peoples with sight disabilities. In this paper, two different approaches for dynamically transforming Web pages into aural Web pages, i.e. pages that are optimised for blind peoples, will be presented. The approaches exploit heuristic techniques for summarising Web pages contents and providing them to blind users in order to improve the usability of Web sites. The techniques have been validated in an experiment where usability metrics have been used to assess the effectiveness of the Web page transformation techniques.

A Semantic Topic Identification System for Document Retrieval on the Web

Effective search and retrieval are fundamental for realizing the full potential of the Web. Although nowadays search engines perform much better than few years ago, big improvements are still needed with respect to the relevance of the retrieved documents to the users query and the presentation of the results. In this paper we present the prototype of a News retrieval system which exploits Wordnets semantics in identifying the topic of retrieved documents and ranking them according to their relevance to the query. Also the system provides a short summary of each document, helping the user in browsing the result collection.

A multimedia data base browsing system

Browsing large multimedia databases is becoming a challenging problem, due to the availability of great amounts of data and the complexity of retrieval. In this paper we propose a system that assists a user in browsing a digital collection making useful recommendations. The system combines computer vision techniques and taxonomic classifcations to measure the similarity between objects and adopts an innovative strategy to take into account user behavior.

An indexing approach for representing multimedia objects in high-dimensional spaces based on expectation maximization algorithm

In this paper we introduce a new indexing approach to representing multimedia object classes generated by the Expectation Maximization clustering algorithm in a balanced and dynamic tree structure. To this aim the EM algorithm has been modified in order to obtain at each step of its recursive application balanced clusters. In this manner our tree provides a simple and practical solution to index clustered data and support efficient retrieval of the nearest neighbors in high dimensional object spaces.

Video segmentation based on active vision

In this paper we present a novel approach for video shot segmentation based on active vision techniques. A continuity function is introduced taking into account the results from biological vision and experiments and discussions are carried out.