Catálogo de publicaciones - libros

We describe an experiment that measures the pedagogical usefulness of the results returned by the National Science Digital Library (NSDL) and Google. Eleven public school teachers from the state of Virginia (USA) were used to evaluate a set of 38 search terms and search results based on the Standards of Learning (SOL) for Virginia Public Schools. Evaluations of search results were obtained from the NSDL (572 evaluations) and Google (650 evaluations). In our experiments, teachers ranked the links returned by Google as more relevant to the SOL than the links returned by the NSDL. Furthermore, Google’s ranking of educational material also showed some correlation with expert judgments.

The access and reproduction policies of the digital collections of ten leading university digital libraries worldwide are classified according to factors such as the creation type of the material, acquisition method, copyright ownership etc. The relationship of these factors is analyzed, showing how acquisition methods and copyright ownership affect the access and reproduction policies of digital collections. We conclude with rules about which factors lead to specific policies. For example, when the library has the copyright of the material, the reproduction for private use is usually provided free with a credit to the source or otherwise mostly under fair use provisions, but the commercial reproduction needs written permission and fees are charged. The extracted rules, which show the common practice on access and reproduction policies, constitute the policy model. Finally, conventional policies are mapped onto digital policies.

The aim of the study was to determine how significance indicators assigned to different Web page elements (internal metadata, title, headings, and main text) influence automated classification. The data collection that was used comprised 1000 Web pages in engineering, to which Engineering Information classes had been manually assigned. The significance indicators were derived using several different methods: (total and partial) precision and recall, semantic distance and multiple regression. It was shown that for best results all the elements have to be included in the classification process. The exact way of combining the significance indicators turned out not to be overly important: using the F1 measure, the best combination of significance indicators yielded no more than 3% higher performance results than the baseline.

Peer-to-peer architectures are a potentially powerful paradigm for retrieving documents over networks of digital libraries avoiding single points of failure by massive federation of (independent) information sources. Today sharing files over P2P infrastructures is already immensely successful, but restricted to simple metadata matching. But when it comes to the retrieval of complex documents, capabilities as provided by digital libraries are needed. Digital libraries have to cope with compound documents. Though some document parts (like embedded images) can efficiently be retrieved using metadata matching, the text-based information needs different methods like full text search. However, for effective querying of texts, also information like inverted document frequencies are essential. But due to the distributed characteristics of P2P networks such ’collection-wide’ information poses severe problems, e.g. that central updates whenever changes in any document collection occur use up valuable bandwidth. We will present a novel indexing technique that allows to query using collection-wide information with respect to different classifications and show the effectiveness of our scheme for practical applications. We will in detail discuss our findings and present simulations for the scheme’s efficiency and scalability.

We propose a method for detecting survey articles in a multilingual database. Generally, a survey article cites many important papers in a research domain. Using this feature, it is possible to detect survey articles. We applied HITS, which was devised to retrieve Web pages using the notions of authority and hub. We can consider that important papers and survey articles correspond to authorities and hubs, respectively. It is therefore possible to detect survey articles, by applying HITS to databases and by selecting papers with outstanding hub scores. However, HITS does not take into account the contents of each paper, so the algorithm may detect a paper citing many principal papers in mistake for survey articles. We therefore improve HITS by analysing the contents of each paper. We conducted an experiment and found that HITS was useful for the detection of survey articles, and that our method could improve HITS.

Vertical search engines and web portals are gaining ground over the general-purpose engines due to their limited size and their high precision for the domain they cover. The number of vertical portals has rapidly increased over the last years, making the importance of a topic-driven (focused) crawler evident. In this paper, we develop a latent semantic indexing classifier that combines link analysis with text content in order to retrieve and index domain specific web documents. We compare its efficiency with other well-known web information retrieval techniques. Our implementation presents a different approach to focused crawling and aims to overcome the size limitations of the initial training data while maintaining a high recall/precision ratio.

Daffodil is a front-end to federated, heterogeneous digital libraries targeting at strategic support of users during the information seeking process. This is done by offering a variety of functions for searching, exploring and managing digital library objects. However, the distributed search increases response time and the conceptual model of the underlying search processes is inherently weaker. This makes query formulation harder and the resulting waiting times can be frustrating. In this paper, we investigate the concept of proactive support during the user’s query formulation. For improving user efficiency and satisfaction, we implemented annotations, proactive support and error markers on the query form itself. These functions decrease the probability for syntactical or semantical errors in queries. Furthermore, the user is able to make better tactical decisions and feels more confident that the system handles the query properly. Evaluations with 30 subjects showed that user satisfaction is improved, whereas no conclusive results were received for efficiency.

The wide adoption of the Z39.50 protocol from the Libraries exposes their abilities to participate in a distributed environment. In spite of the protocol specification of a unified global access mechanism, query failures and/or inconsistent answers are the pending issues when searching many sources due to the variant or poor implementations. The elimination of these issues heavily depends on the ability of the client to make decisions prior to initiating search requests, utilizing the knowledge of the supported search capabilities of each source. To effectively reformulate such requests, we propose a Datalog based description for capturing the knowledge about the supported search capabilities of a Z 39.50 source. We assume that the accessible sources can answer some but possibly not all queries over their data, and we describe a model for their supported search capabilities using a set of parameterized queries, according to the Relational Query Description Language (RQDL) specification.

This paper examines the difference and similarities between the two on-line computer science citation databases DBLP and CiteSeer. The database entries in DBLP are inserted manually while the CiteSeer entries are obtained autonomously via a crawl of the Web and automatic processing of user submissions. CiteSeer’s autonomous citation database can be considered a form of self-selected on-line survey. It is important to understand the limitations of such databases, particularly when citation information is used to assess the performance of authors, institutions and funding bodies. We show that the CiteSeer database contains considerably fewer single author papers. This bias can be modeled by an exponential process with intuitive explanation. The model permits us to predict that the DBLP database covers approximately 24% of the entire literature of Computer Science. CiteSeer is also biased against low-cited papers. Despite their difference, both databases exhibit similar and significantly different citation distributions compared with previous analysis of the Physics community. In both databases, we also observe that the number of authors per paper has been increasing over time.

The design, implementation and evaluation of a multi-document summarization system for sociology dissertation abstracts are described. The system focuses on extracting variables and their relationships from different documents, integrating the extracted information, and presenting the integrated information using a variable-based framework. Two important summarization steps – information extraction and information integration were evaluated by comparing system-generated output against human-generated output. Results indicate that the system-generated output achieves good precision and recall while extracting important concepts from each document, as well as good clusters of similar concepts from the set of documents.