Information Extraction from Web Pages Using Presentation Regularities and Domain Knowledge

World Wide Web is transforming itself into the largest information resource making the process of information extraction (IE) from Web an important and challenging problem. In this paper, we present an automated IE system that is domain independent and that can automatically transform a given Web page into a semi-structured hierarchical document using presentation regularities. The resulting documents are weakly annotated in the sense that they might contain many incorrect annotations and missing labels. We also describe how to improve the quality of weakly annotated data by using domain knowledge in terms of a statistical domain model. We demonstrate that such system can recover from ambiguities in the presentation and boost the overall accuracy of a base information extractor by up to 20%. Our experimental evaluations with TAP data, computer science department Web sites, and RoadRunner document sets indicate that our algorithms can scale up to very large data sets.     

Automated analysis of images in documents for intelligent document search

Authors use images to present a wide variety of important information in documents. For example, two-dimensional (2-D) plots display important data in scientific publications. Often, end-users seek to extract this data and convert it into a machine-processible form so that the data can be analyzed automatically or compared with other existing data. Existing document data extraction tools are semi-automatic and require users to provide metadata and interactively extract the data. In this paper, we describe a system that extracts data from documents fully automatically, completely eliminating the need for human intervention. The system uses a supervised learning-based algorithm to classify figures in digital documents into five classes: photographs, 2-D plots, 3-D plots, diagrams, and others. Then, an integrated algorithm is used to extract numerical data from data points and lines in the 2-D plot images along with the axes and their labels, the data symbols in the figure’s legend and their associated labels. We demonstrate that the proposed system and its component algorithms are effective via an empirical evaluation. Our data extraction system has the potential to be a vital component in high volume digital libraries.     

Table form document analysis based on the document structure grammar

Structure analysis of table form documents is an important issue because a printed document and even an electronic document do not provide logical structural information but merely geometrical layout and lexical information. To handle these documents automatically, logical structure information is necessary. In this paper, we first analyze the elements of the form documents from a communication point of view and retrieve the grammatical elements that appear in them. Then, we present a document structure grammar which governs the logical structure of the form documents. Finally, we propose a structure analysis system of the table form documents based on the grammar. By using grammar notation, we can easily modify and keep it consistent, as the rules are relatively simple. Another advantage of using grammar notation is that it can be used for generating documents only from logical structure. In our system, documents are assumed to be composed of a set of boxes and they are classified as seven box types. Then the box relations between the indication box and its associated entry box are analyzed based on the semantic and geometric knowledge defined in the document structure grammar. Experimental results have shown that the system successfully analyzed several kinds of table forms.     

基于知识的多页文档逻辑结构的分析和理解

文档图像理解中最重要的部分是逻辑结构的提取。目前的研究主要集中在页面的布局分析上,少数对文档逻辑结构的研究只是针对单页文档或页面关系简单的多页文档。建筑标书的特殊性在于其层次式的逻辑组成结构没有明确的索引信息标识。本文提出了一种利用页面间引用关系获取文档逻辑结构的方法。该方法采用修正的树形结构表示文档的逻辑结构,逻辑树的创建过程就是逻辑结构的获取过程,而且有利于更高层的语义处理及还原输出。该方法已在标书自动处理系统中实现,保证了该系统的灵活和高效。     

Biblio: automatic meta-data extraction

Biblio is an adaptive system that automatically extracts meta-data from semi-structured and structured scanned documents. Instead of using hand-coded templates or other methods manually customized for each given document format, it uses example-based machine learning to adapt to customer-defined document and meta-data types. We provide results from experiments on the recognition of document information in two document corpuses: a set of scanned journal articles and a set of scanned legal documents. The first set is semi-structured, as the different journals use a variety of flexible layouts. The second set is largely free-form text based on poor quality scans of FAX-quality legal documents. We demonstrate accuracy on the semi-structured document set roughly comparable to hand-coded systems, and much worse performance on the legal documents.     

Automatic document classification and indexing in high-volume applications

In this paper a system for analysis and automatic indexing of imaged documents for high-volume applications is described. This system, named STRETCH (STorage and RETrieval by Content of imaged documents), is based on an Archiving and Retrieval Engine, which overcomes the bottleneck of document profiling bypassing some limitations of existing pre-defined indexing schemes. The engine exploits a structured document representation and can activate appropriate methods to characterise and automatically index heterogeneous documents with variable layout. The originality of STRETCH lies principally in the possibility for unskilled users to define the indexes relevant to the document domains of their interest by simply presenting visual examples and applying reliable automatic information extraction methods (document classification, flexible reading strategies) to index the documents automatically, thus creating archives as desired. STRETCH offers ease of use and application programming and the ability to dynamically adapt to new types of documents. The system has been tested in two applications in particular, one concerning passive invoices and the other bank documents. In these applications, several classes of documents are involved. The indexing strategy first automatically classifies the document, thus avoiding pre-sorting, then locates and reads the information pertaining to the specific document class. Experimental results are encouraging overall; in particular, document classification results fulfill the requirements of high-volume application. Integration into production lines is under execution. Received March 30, 2000 / Revised June 26, 2001     

Machine Learning for Intelligent Processing of Printed Documents

A paper document processing system is an information system component which transforms information on printed or handwritten documents into a computer-revisable form. In intelligent systems for paper document processing this information capture process is based on knowledge of the specific layout and logical structures of the documents. This article proposes the application of machine learning techniques to acquire the specific knowledge required by an intelligent document processing system, named WISDOM++, that manages printed documents, such as letters and journals. Knowledge is represented by means of decision trees and first-order rules automatically generated from a set of training documents. In particular, an incremental decision tree learning system is applied for the acquisition of decision trees used for the classification of segmented blocks, while a first-order learning system is applied for the induction of rules used for the layout-based classification and understanding of documents. Issues concerning the incremental induction of decision trees and the handling of both numeric and symbolic data in first-order rule learning are discussed, and the validity of the proposed solutions is empirically evaluated by processing a set of real printed documents.     

Parallelization of a dynamic SVD clustering algorithm and its application in information retrieval

We describe an implementation of a parallel document clustering scheme based on latent semantic indexing, which uses singular value decomposition. Given a set of documents, the clustering algorithm is dynamic in the sense that it automatically infers the number of clusters to be output. The parallel version has been implemented on a LAN and on a dual‐core system. Experimental evaluation of the algorithm shows an average speed‐up of 6.22 for the LAN implementation and an average speed‐up of 3.71 for the dual‐core implementation, while still maintaining a precision and recall in the range [0.85, 1]. To put these implementations in the context of information retrieval, we use the parallel clustering algorithm and develop a document similarity search system. The similarity search system shows good performance in terms of precision and recall. Copyright © 2010 John Wiley & Sons, Ltd.     

Semantic adaptation of multimedia documents

Multimedia documents have to be played on multiple device types. Hence, usage and platform diversity requires document adaptation according to execution contexts, not generally predictable at design time. In an earlier work, a semantic framework for multimedia document adaptation was proposed. In this framework, a multimedia document is interpreted as a set of potential executions corresponding to the author specification. To each target device corresponds a set of possible executions complying with the device constraints. In this context, adapting requires to select an execution that satisfies the target device constraints and which is as close as possible from the initial composition. This theoretical adaptation framework does not specifically consider the main multimedia document dimensions, i.e., temporal, spatial and hypermedia. In this paper, we propose a concrete application of this framework on standard multimedia documents. For that purpose, we first define an abstract structure that captures the spatio-temporal and hypermedia dimensions of multimedia documents, and we develop an adaptation algorithm which transforms in a minimal way such a structure according to device constraints. Then, we show how this can be used for adapting concrete multimedia documents in SMIL through converting the documents in the abstract structure, using the adaptation algorithm, and converting it back in SMIL. This can be used for other document formats without modifying the adaptation algorithm.