Using perception in managing unstructured documents

Advances in Pattern Recognition, 2007

2006

Since the system is general and flexible, it can be embedded as a document management engine into many different domain-specific applications. Here, we present the exploitation of the system on the Conference Management domain, and show how DOMINUS can usefully support some of the more critical and knowledge-intensive tasks involved by the organization of a scientific conference.

2002

Military organizations have to deal with an increasing number of documents coming from different sources and in various formats (paper, fax, e-mail messages, electronic documents). These documents have to be screened, analyzed and categorized in order to interpret their content and gain situation awareness. These documents should be categorized according to their content to enable efficient storage and retrieval. In this context, intelligent techniques and tools should be provided to support this information management process that is currently partly manual. Integrating the recently acquired knowledge in different fields in a system for analyzing, diagnosing, filtering, classifying and clustering documents with a limited human intervention would improve efficiently the quality of information management with reduced human resources. A better categorization and management of information would facilitate correlation of information from different sources, avoid information redundancy, improve access to relevant information, and thus better support decision-making processes. The RDDC-Valcartier's ADAC system (Automatic Documents Analyzer and Classifier) incorporates several techniques and tools for document summarizing and semantic analysis based on ontology of a certain domain (e.g. terrorism), and algorithms of diagnostic, classification and clustering. In this paper, we describe the architecture of the system and the techniques and tools used at each step of the document processing. For the first prototype implementation, the focus has been concentrated on the terrorism domain to develop document corpus and related ontology.

In information retrieval systems, the indexation task is usually conducted irrespective of the document structure. We introduce here a module which allows preprocessing of documents prior to indexation techniques. It detects the physical lay-out of text and labels text zones, such as title and text-body. The method relies on pattern recognition and applies to a wide range of documents. This ensures the correct routing of selected parts of a document towards adequate tools for analysis. Examples of text identification and segmentation at rough and minute grain are presented. Document structure identification offers an opportunity to balance the calculation of the inputs of an index according to the nature of the zones in which the terms appear (title or text-body for instance).

2005

Every day, the volume of the world's digital data increases considerably. Over 75% of these data are non-structured. This paper is about restructuring graphic information contained in Portable Document Format (PDF) files and/or vector files. These documents are generally held by ''Smart Factory'' services: design offices, methods departments, new work departments and company maintenance services. To restructure these data, we propose using Knowledge Discovery in Databases (KDD) methods. Although, theoretically, the user is present during the KDD, in practice, this is not the case. This was observed by Fayard in 2003 at the KDD conference. Generally, the user is only present during the validation phase. We show why, in data restructuring, the user must be at the heart of the process and present at all stages. We can talk about (A)KDD for the Anthropocentric Knowledge Discovery in Databases .The first stage of this restructuring consists of extracting graphic and t...

Given a large hierarchical concept dictionary (thesaurus, or ontology), the task of selection of the concepts that describe the contents of a given document is considered. A statistical method of document indexing driven by such a dictionary is proposed. The method is insensible to inaccuracies in the dictionary, which allow for semi-automatic translation of the hierarchy into different languages. The problem of handling non-terminal and especially top-level nodes in the hierarchy is discussed. Common sense-complaint methods of automatically assigning the weights to the nodes and links in the hierarchy are presented. The application of the method in the Classifier system is discussed.

Data, Text and Web Mining Applications, 2011

Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark. Library of Congress Cataloging-in-Publication Data Visual analytics and interactive technologies : data, text, and web mining applications / Qingyu Zhang, Richard Segall, and Mei Cao, editors. p. cm. Includes bibliographical references and index. Summary: "This book is a comprehensive reference on concepts, algorithms, theories, applications, software, and visualization of data mining, text mining, Web mining and computing/supercomputing, covering state-of-the-art of the theory and applications of mining"-Provided by publisher.