Catálogo de publicaciones - libros

In this paper we propose to see the information seeking behaviour of domain novices as search and learn processes. We explore the concept of naïve ontology as the bases for designing browsing/navigation interface of search and learn for domain novices of digital libraries. Naïve ontology is a type of information access interface which allows domain novices to refine their knowledge interactively by acquiring information chunk-by-chunk as they encounter it in digital libraries. Through elicitation and analysis of search and learn processes of domain novices in history and geography, we identified several implications for designing naïve ontology.

Identifying meaningful document fragments is a major advantage achieved by encoding documents in XML. In scholarly articles, such document fragments include sections, subsections and paragraphs. XML information retrieval systems need to search document fragments relevant to queries from a set of XML documents in a digital library. We present Kikori-KS, an effective and efficient XML information retrieval system for scholartic articles. Kikori-KS accepts a set of keywords as a query. This form of query is simple yet useful because users are not required to understand XML query languages or XML schema. To meet practical demands for searching relevant fragments in scholartic articles, we have developed a user-friendly interface for displaying search results. Kikori-KS was implemented on top of a relational XML database system developed by our group. By carefully designing the database schema, Kikori-KS handles a huge number of document fragments efficiently. Our experiments using INEX test collection show that Kikori-KS achieved an acceptable search time and with relatively high precision.

XML is the most popular platform-independent data expression language which is used to specify various digital content such as web content, multimedia content, bio-chemical data, etc. These various forms of XML data are continuously increasing by a large amount and there is a strong demand on effectively managing such data in digital libraries or archives. The most popular query language to search and retrieve information from such semi-structured XML digital libraries is XQuery. XQuery has a very powerful syntax which allows users to iterate over data items and perform calculation, string matching, and output formatting. However, it lacks a simple and easy way to group and provide summaries on vast amounts of XML data. This grouping and summary function is especially important for large digital archives where users like to obtain an overview or summary of the contents in the digital library. Our work is focused on providing an easy way for grouping in XQuery at the query language level. We provide several cases where this can be considered to be effective. We have also implemented an XQuery processing system with grouping functions based on the eXist Native XML Database.

Digital Libraries make Japan’s rich cultural heritage available worldwide for the first time, opening new areas for scholarship. Nonetheless, access, discovery and navigation challenges abound. This paper presents case studies of three formats (poetry, company histories, manga ) characteristic of Japanese print publishing, identifying specific support structures that are evolving to enhance scholarly access to and use of these resources in digital libraries. In particular, the role of researchers in design and construction of access resources is explored. Better awareness and understanding of the role of access resources in digital libraries has implications for information literacy programs and other kinds of instruction.

In this paper, we present the results of experiments identifying the drug treatment relation and drug treatment attributes like dosage, treatment frequency and duration from abstracts of medical publications using linguistic patterns. The approach uses an automatic linguistic pattern construction algorithm after the dataset has been semantically annotated. The automatically constructed patterns were able to identify treatment relations and their attributes with varying success. We observe that the simple (or naïve) treatment patterns performs much better than the non-naïve treatment patterns in identifying sentences with drug treatment relationship in both cancer and non-cancer drug therapy domain. However the drug dosage, frequency and duration patterns performed much better in the identification of relationships in the cancer drug therapy domain than the non-cancer drug therapy domain.

This paper presents findings from a questionnaire survey that aimed to identify the issues around the use and linkage of source and output repositories and the expectations of the chemistry research community about their use. In the context of the StORe project (http://jiscstore.jot.com/ WikiHome), which sought to develop new ways of linking academic publications with repositories of research data, thirty eight (38) members of academic and research staff from institutions across the UK provided valuable feedback regarding the nature of the research that they conduct, the type of data that they produce, the sharing and availability of research data and the use and expectations of source and output repositories.

This paper proposes the architecture of the functional composition of Web databases (WebDBs). Unlike a general search engine which receives keywords and returns a list of URLs, a WebDB receives a complex query and returns a list of records. The complex query specifies the condition of each field of the records. The process of composing WebDBs is described as a script, where a user chooses the target WebDBs and describes how to connect the output from one WebDB to the input of another WebDB and how to generate outputs. The novelty of the proposal is that both the WebDBs and output formats are considered as components of the same level and that the reuse of new keywords is represented as a connection (CGI links). Once the process is described as a script, the user can use the script for a new WebDB of his own.

In this paper, we address the problem of integrating Wikipedia, an online encyclopedia, and G-Portal, a web-based digital library, in the geography domain. The integration facilitates the sharing of data and services between the two web applications that are of great value in learning. We first present an overall system architecture for supporting such an integration and address the metadata extraction problem associated with it. In metadata extraction, we focus on extracting and constructing metadata for geo-political regions namely cities and countries. Some empirical performance results will be presented. The paper will also describe the adaptations of G-Portal and Wikipedia to meet the integration requirements.

Hierarchical summarization technique summarizes a large document based on the hierarchical structure and salient features of the document. Previous study has shown that hierarchical summarization is a promising technique which can effectively extract the most important information from the source document. Hierarchical summarization has been extended to summarization of multiple documents. Three hierarchical structures were proposed to organize a set of related documents. This paper investigates the impact of document structure on hierarchical summarization. The results show that the hierarchical summarization of multiple documents organized in hierarchical structure outperforms other multi-document summarization systems without using the hierarchical structure. Moreover, the hierarchical summarization by event topics extracts a set of sentences significantly different from hierarchical summarization of other hierarchical structures and performs the best when the summary is highly-compressed.

Recommendation systems help users find the information, products and services they most want to find. Collaborative filtering is the method of making automatic predictions about the interest of a user by collecting interest information from many users, which has been very successful recommendation technique for recommendation systems in both research and practice. However, the traditional collaborative filtering is slow to detect the interest of a user changing with time as a case of user behavior and to adapt the changes, because the traditional collaborative filtering uses Pearson’s correlation coefficient between users with the numerous values of property. In this paper, we apply the wavelet analysis to collaborative filtering in order to reveal the trends hidden in the interest of a user and propose the wavelet-based collaborative filtering for adapting changes in user behavior. The results of the performance evaluation show that the proposed wavelet-based collaborative filtering makes the improvement in the personalized recommendations.