Search

This paper reports on the RMIT group's approach to XML retrieval while participating in INEX 2003. We indexed XML documents using Lucy, a compact and fast text search engine designed and written by the Search Engine Group at RMIT University. For each INEX topic, up to 1000 highly ranked documents were then loaded and indexed by eXist, an open source native XML database. A query translator converts the INEX topics into corresponding Lucy and eXist query expressions, respectively. These query expressions may represent traditional information retrieval tasks (unconstrained, CO topics), or may focus on retrieving and ranking specific document components (constrained, CAS topics). With respect to both these expression types, we used eXist to extract final answers (either full documents or document components) from those documents that were judged highly relevant by Lucy. Several extraction strategies were used that differently in uenced the ranking order of the fnal answers. The final INEX results show that our choice for a translation method and an extraction strategy leads to a very effective XML retrieval for the CAS topics. We observed a system limitation for the CO topics resulting in the same or similar choice to have little or no impact on the retrieval performance.

This paper gives an overview of tools and methods for Cross-Language Information Retrieval (CLIR) that are developed within the Twenty-One project. The tools and methods are evaluated with the TREC CLIR task document collection using Dutch queries on the English document base. The main issue addressed here is an evaluation of two approaches to disambiguation. The underlying question is whether a lot of effort should be put in finding the correct translation for each query term before searching, or whether searching with more than one possible translation leads to better results? The experimental study suggests that the quality of search methods is more important than the quality of disambiguation methods. Good retrieval methods are able to disambiguate translated queries implicitly during searching.

This paper reports on the RMIT group's approach to XML retrieval while participating in INEX 2003. We indexed XML documents using Lucy, a compact and fast text search engine designed and written by the Search Engine Group at RMIT University. For each INEX topic, up to 1000 highly ranked documents were then loaded and indexed by eXist, an open source native XML database. A query translator converts the INEX topics into corresponding Lucy and eXist query expressions respectively. These query expressions may represent traditional information retrieval tasks(unconstrained, CO topics), or may focus on retrieving and ranking specific document components (constrained, CAS topics). With respect to both these expressions types, we used eXist to extract final answers (either full documents or document components) frome those documents that were judged highly relevant by Luy. Several extraction strategies were used that diffeently influenced the ranking order of the final answers. The final INEX results show that our choice for a translation method and an extraction stategy leads to a very effective XML retrieval for the CAS topics. We observed a system limitation for CO topics resulting in the same or similar choice to have little or no impact on the retrieval performance.

Increasingly, people find information about corporations through their web sites, as opposed to reading printed brochures. Typically, the only tools they have to find that information is searching and browsing. But can we do better with a modest user model and coherent delivery? Can we provide, through the web, more relevant and coherent information about a corporation, information specially tailored to specific clients and presented in a useful and understandable manner? Our conjecture was that a tailored hypermedia system offering tailored and coherent information would indeed provide a good alternative to the current search-and-browse facilities, also allowing corporations to be more responsive to their clients' needs. In this paper, we present a preliminary study aimed at testing this hypothesis. We designed an experiment aimed at investigating the effect of delivering coherent and tailored information in response to user's query, in comparison to delivering the results of conventional search engines. We employed our experimental system, PERCY, which exploits information retrieval, user modeling and discourse-oriented technologies to retrieve and select parts of existing documents from CSIRO website and assemble them back into a coherent and tailored webpage. Our experiment consisted of asking users to evaluate and compare the results produced by PERCY and those produced by a typical search engine, given a query about the some project or research topic at CSIRO. Our experimental results indicate that users preferred the output of Percy, as it attracted more their attention and they understood the results better.

This paper explores the needs of XML-search and makes comparisons between the INEX experience in evaluating XML retrieval and the recently proposed W3C requirements for extending XQuery and XPath to handle full-text retrieval. Our experience in INEX gave insight into some problems combining ranking with XML structured queries and more generally the needs from the users' perspective. INEX has focused on how to express the user needs through topics that may involve structural XML elements. How such user needs can be translated into XML queries has an impact on the requirements for an XML-search language. The analysis in this paper shows that the W3C requirements for extending XQuery and XPath with full-text search seem to address most of the needs identified through our first experiments with INEX. In particular, being able to specify an implementation dependent scoring function that allows ranking of answers. However, there are still issues concerning how the W3C requirements cater for combining structural requirements with ranking.

The main source of competitive advantage of an organization is its knowledge assets and its learning capacity. We argue that finding knowledge in an organization is not just about finding documents, it is also about finding the right person. In this paper we present PeopleFinder, a Web based system which automatically identifies experts in an area, based on the documents already published on an organization's intranet or part of the corporate data. The system can be queried like a standard Web search engine, but instead of returning documents it returns a list of experts. Each expert listing includes contact details and supporting evidence. The prototype leverages on organizational data, such the project descriptions and membership to infer which documents can be used as best evidences of expertise. The paper also includes preliminary evaluations

This paper reports on the CSIRO group's participation in INEX. We indexed documents and document fragments using PADRE, the core of CSIRO's Panoptic Enterprise Search Engine. A query translator converts the INEX topics into queries containing selection and projection constraints for the results. Answers are extracted from ranked documents and document fragments based on the projection constraints in the query.

he notion of wrapping a web server to produce XML documents from unstructed web pages is driven by the need to produce structured data that can be used by a variety of applications. The web contains vast amounts of information that cannot be used by most applications as it targets a human audience. A solution to this is to automate the browsing process and convert the unstructured extracted information into a more structured format such as XML. This is called wrapping. We have used two different tools to wrap several tourist sites into XML The tools we have used are Norfolk, a system developed by the CSIRO TED group and W4F, initially developed at the University of Pennsylvania and now a commercial product. This report describes our practical experience with the tools and compares them. The comparison highlights features required by a wrapper system to support real applications.

The Web is creating exciting new possibilities for direct and instantaneous publishing of information. However, the apparent ease with which one can publish documents on the Web hides more complex issues such updating and maintaining Web pages. We believe one of the crucial requirements to document delivery is the ability to extract and reuse information from other documents or sources. In this paper we present a descriptive language that allows users to write virtual documents, where dynamic information can be retrieved from various sources, transformed, and included along with static information in HTML documents. The language uses a tree-like structure for the representation of information, and defines a database-like query language for extracting and combining information without a complete knowledge of the structure or the types of information. The data structures and the syntax of the language are presented along with examples.

The goal of our work is to use a set of reports and extract named entities, in our case the names of Industrial or Academic partners. Starting with an initial list of entities, we use a first set of documents to identify syntactic patterns that are then validated in a supervised learning phase on a set of annotated documents. The complete collection is then explored. This approach is similar to the ones used in data extraction from semi-structured documents (wrappers) and do not need any linguistic resources neither a large set for training. As our collection of documents would evolve over years, we hope that the performance of the extraction would improve with the increased size of the training set., The bibteX file has been replaced with the correct one

International audience; This article is a report concerning the two years of the XML Mining track at INEX (2005 and 2006). We focus here on the classification and clustering XML documents. We detail these two tasks and the corpus used for this challenge and then present a summary of the different methods proposed by the participants. We last compare the results obtained during the two years of the track.

www.cepadues.com; In this work, we propose a new clustering document representation for semi-structured documents collections. Our approach consists on a representation of XML documents based on their sub-paths, defined according to some criteria (length, root beginning, leaf ending) using the structure only or both the structure and the content. By considering those sub-paths as words, we can use standard methods for vocabulary reduction, and simple clustering methods such as K-means that scale up well. We actually use an implementation of the clustering algorithm known as \textit{dynamic clouds} that can work with distinct groups of independent variables. This is necessary in our model since embedded sub-paths are not independent. For validation and evaluation of our method, two collections are used: the INEX corpus and the INRIA activity reports, and a set of metrics well-known in Information Retrieval.

http://www.cs.utwente.nl/~tdm/Proceedings/TDMProc.pdf; Three approaches to content-and-structure XML retrieval are analysed in this paper: first by using Zettair, a full-text information retrieval system; second by using eXist, a native XML database, and third by using a hybrid XML retrieval system that uses eXist to produce the final answers from likely relevant articles retrieved by Zettair. INEX 2003 content-and-structure topics can be classified in two categories: the first retrieving full articles as final answers, and the second retrieving more specific elements within articles as final answers. We show that for both topic categories our initial hybrid system improves the retrieval effectiveness of a native XML database. For ranking the final answer elements, we propose and evaluate a novel retrieval model that utilises the structural relationships between the answer elements of a native XML database and retrieves Coherent Retrieval Elements. The final results of our experiments show that when the XML retrieval task focusses on highly relevant elements our hybrid XML retrieval system with the Coherent Retrieval Elements module is 1.8 times more effective than Zettair and 3 times more effective than eXist, and yields an effective content-and-structure XML retrieval.