Catálogo de publicaciones - libros

The amount of semi-structured documents is tremendous online, such as business annual reports, online airport listings, catalogs, hotel directories, etc. List, which has structured characteristics, is used to store highly structured and database-like information in many semi-structured documents. This paper is about list data extraction from semi-structured documents. By list data extraction, we mean extracting data from lists and grouping it by rows and columns. List data extraction is of benefit to text mining applications on semi-structured documents. Recently, several methods are proposed to extract list data by utilizing the word layout and arrangement information [1, 2]. However, in the research community, few previous studies has so far sufficiently investigated the problem of making use of not only layout and arrangement information, but also the semantic information of words, to the best of our knowledge. In this paper, we propose a clustering based method making use of both the layout information and the semantic information of words for this extraction task. We show experimental results on plain-text annual reports from Shanghai Stock Exchange, in which 73.49% of the lists were extracted correctly.

In this paper, we discuss one of several optimization issues regarding our algebraic query model for keyword search over tree-structured documents. In particular, we focus on the properties of a class of filters. The filters in this class not only restrict the size of query results, but also are capable of reducing the cost of query processing.

A challenging problem in Web engineering is the integration of XML data sources. Even if these data sources conform to schemas, they may have their schemas and the correspongind XML documents structured differently. In this paper, we address the problem of integrating XML data sources (a) by adding semantic information to document schemas, and (b) by using a query language that allows a partial specification of tree patterns. The semantic information allows the grouping of elements into the so called schema dimensions . Our approach allows querying data sources with different schemas in an integrated way. Users posing queries have the flexibility to specify structural constraints fully, partially or not at all. Our approach was initially developed for arbitrarily structured data sources [1]. Here, we show how this approach can be applied to tree-structured data sources that comply to schemas.

We propose a method named WordRank to extract a few salient words from the target document and then use these words to retrieve similar documents based on popular retrieval functions. The set of extracted words is a concise and topic-oriented representation of the target document and reduces the ambiguous and noisy information in the document, so as to improve the retrieval performance. Experiments and results demonstrate the high effectiveness of the proposed approach.

We consider the Peer-To-Peer (P2P) database system with RDF ontologies and with the semantic characterization of P2P mappings based on logical views over local peer’s ontology. Such kind of virtual-predicate based mappings needs an embedding of RDF ontologies into a predicate first-order logic, or at some of its sublanguages as, for example, logic programs for deductive databases. We consider a peer as a local epistemic logic system with its own belief based on RDF tuples, independent from other peers and their own beliefs. This motivates the need of a semantic characterization of P2P mappings based not on the extension but on the meaning of concepts used in the mappings, that is, based on intensional logic. We show that it adequately models robust weakly-coupled framework of RDF ontologies and supports decidable query answering.The approach to use conventional first order logic (FOL) as the semantic underpinning for RDF has many advantages: FOL is well established and well understood. We will consider an RDF-ontology as finite set of triples < r , p , v >, where r is a resource name (for class, an instance or a value), p is a property (InstanceOf or Property in RDF, or Subclass or Property in RDFS), and v is a value (which could also be a resource name). We denote by $\mathcal{T}$ the set of all triples which satisfy such requirements.

This paper proposes a component-based architecture for adaptive Web-based education systems that support a particular kind of EML models. Our work is concerned with the development of an EML meta-model to provide an enhanced support for the modeling of collaborative practices. The proposal is based on the identification of perspectives.

Developing an intelligent recommendation system is a good way to overcome the problem of products information overload. We believe that the personalized recommendation system should be build according the special features of a certain kind of products, thereby forming professional recommendation systems for different products. In the paper, we propose a system for digital products, such as laptop, digital camera, PDA, etc. The approach utilizes fuzzy logic to retrieve optimal products based on the consumer’s current preferences from the system-user interactions. Experimental results show the promise of our systems.

We propose a framework for ontology-based consensus making, which is grounded on personalization and community support. Corresponding software is designed to be naturally deployed in community Web environments.

A chemistry-focused search engine, named ChemEngine, is developed to help chemists to get chemical information more conveniently and precisely on Internet. Text Categorization is used in ChemEngine to facilitate users’ search. The semantic similarity and noisy data in chemical web pages make traditional classifier perform poorly on them. To classify chemical web pages more accurately, a new text categorization approach based on dictionary and voting is proposed and integrated into the ChemEngine.

Web Services play a growing role in the geographic community. Efforts to establish a Spatial Data Infrastructure (SDI) are coordinated by the Open Geospatial Consortium (OGC). However, as the infrastructure gets established, and more content providers wish to offer their products, questions of security arise. Geographical services often stand to gain tremendously from composition and delegation of requests, where one piece of data is constructed from several sets of data, using different sources. However, current standards by the OGC regulate only message transfers, without taking considerations of access control, security and privacy into account.