Document Classification

Interest in the area of pattern recognition has been renewed recently due to emerging applications which are not only challenging but also computationally more demanding. These applications include data mining (identifying a "pattern", e.g., correlation, or an outlier in millions of multidimensional patterns), document classifications (efficiently searching text documents), organization and retrieval of multimedia database, and biometrics (personal identifications based on various physical attributes such as face and fingerprints). The three best known conventional approaches for pattern recognition are: template matching, statistical classification and syntactic or structural matching. The limitations and constraints of these conventional approaches have made researchers to look for alternate techniques based on Artificial Neural Networks. The main characteristics of neural networks are that they have the ability to learn complex nonlinear input-output relationship, use sequential training procedures, and adapt themselves to the data. In this paper, we discuss implementation and fault tolerance analysis of the most commonly used family of neural networks for pattern classification tasks i.e., the feed forward network, which includes a fully interconnected three layered [25-10-1] perceptron. The delta rule weight adjustment is implemented by taking the gradient of error function, which gives the direction in which weights have to be adjusted to get error value within the predefined threshold. Fault tolerance analysis is done for Gaussian and Uniform distribution of weights. The efficacy of neural network based pattern recognition is tested by the computer simulation results. Index terms-Pattern recognition, fault tolerance, neural network, back-propagation

In recent years, XML has been established as a major means for information management, and has been broadly utilized for complex data representation (e.g. multimedia objects). Owing to an unparalleled increasing use of the XML standard, developing efficient techniques for comparing XML-based documents becomes essential in the database and information retrieval communities. In this paper, we provide an overview of XML similarity/comparison by presenting existing research related to XML similarity. We also detail the possible applications of XML comparison processes in various fields, ranging over data warehousing, data integration, classification/clustering and XML querying, and discuss some required and emergent future research directions.

To help the growing qualitative and quantitative demands for information from the WWW, efficient automatic Web page classifiers are urgently needed. However, a classifier applied to the WWW faces a huge-scale dimensionality problem since it must handle millions of Web pages, tens of thousands of features, and hundreds of categories. When it comes to practical implementation, reducing the dimensionality is a critically important challenge. In this paper, we propose a fuzzy ranking analysis paradigm together with a novel relevance measure, discriminating power measure (DPM), to effectively reduce the input dimensionality from tens of thousands to a few hundred with zero rejection rate and small decrease in accuracy. The two-level promotion method based on fuzzy ranking analysis is proposed to improve the behavior of each relevance measure and combine those measures to produce a better evaluation of features. Additionally, the DPM measure has low computation cost and emphasizes on both positive and negative discriminating features. Also, it emphasizes classification in parallel order, rather than classification in serial order. In our experimental results, the fuzzy ranking analysis is useful for validating the uncertain behavior of each relevance measure. Moreover, the DPM reduces input dimensionality from 10,427 to 200 with zero rejection rate and with less than 5% decline (from 84.5% to 80.4%) in the test accuracy. Furthermore, to consider the impacts on classification accuracy for the proposed DPM, the experimental results of China Time and Reuter-21578 datasets have demonstrated that the DPM provides major benefit to promote document classification accuracy rate. The results also show that the DPM indeed can reduce both redundancy and noise features to set up a better classifier.

The web contains a wealth of product reviews, but sifting through them is a daunting task. Ideally, an opinion mining tool would process a set of search results for a given item, generating a list of product attributes (quality, features, etc.) and aggregating opinions about each of them (poor, mixed, good). We begin by identifying the unique properties of this problem and develop a method for automatically distinguishing between positive and negative reviews. Our classifier draws on information retrieval techniques for feature extraction and scoring, and the results for various metrics and heuristics vary depending on the testing situation. The best methods work as well as or better than traditional machine learning. When operating on individual sentences collected from web searches, performance is limited due to noise and ambiguity. But in the context of a complete web-based tool and aided by a simple method for grouping sentences into attributes, the results are qualitatively quite useful.

The widespread use of information technologies for construction is considerably increasing the number of electronic text documents stored in construction management information systems. Consequently, automated methods for organizing and improving the access to the information contained in these types of documents become essential to construction information management. This paper describes a methodology developed to improve information organization and access in construction management information systems based on automatic hierarchical classification of construction project documents according to project components. A prototype system for document classification is presented, as well as the experiments conducted to verify the feasibility of the proposed approach.

Document image classification is an important step in Office Automation, Digital Libraries, and other document image analysis applications. There is great diversity in document image classifiers: they differ in the problems they solve, in the use of training data to construct class models, and in the choice of document features and classification algorithms. We survey this diverse literature using three components: the problem statement, the classifier architecture, and performance evaluation. This brings to light important issues in designing a document classifier, including the definition of document classes, the choice of document features and feature representation, and the choice of classification algorithm and learning mechanism. We emphasize techniques that classify single-page typeset document images without using OCR results. Developing a general, adaptable, high-performance classifier is challenging due to the great variety of documents, the diverse criteria used to define document classes, and the ambiguity that arises due to ill-defined or fuzzy document classes.

Frequent itemset mining (FIM) is a core operation for several data mining applications as association rules computation, correlations, document classification, and many others, which has been extensively studied over the last decades. Moreover, databases are becoming increasingly larger, thus requiring a higher computing power to mine them in reasonable time. At the same time, the advances in high performance computing platforms are transforming them into hierarchical parallel environments equipped with multi-core processors and many-core accelerators, such as GPUs. Thus, fully exploiting these systems to perform FIM tasks poses as a challenging and critical problem that we address in this paper. We present efficient multi-core and GPU accelerated parallelizations of the Tree Projection, one of the most competitive FIM algorithms. The experimental results show that our Tree Projection implementation scales almost linearly in a CPU shared-memory environment after careful optimizations, while the GPU versions are up to 173 times faster than standard the CPU version.

— With the increasing availability of electronic documents and the rapid growth of the World Wide Web, the task of automatic categorization of documents became the key method for organizing the information and know-ledge discovery. Proper classification of e-documents, online news, blogs, e-mails and digital libraries need text mining, machine learning and natural language processing tech-niques to get meaningful knowledge. The aim of this paper is to highlight the important techniques and methodologies that are employed in text documents classification, while at the same time making awareness of some of the interesting challenges that remain to be solved, focused mainly on text representation and machine learning techniques. This paper provides a review of the theory and methods of document classification and text mining, focusing on the existing litera-ture.

TWLT is an acronym of Twente Workshop(s) on Language Technology. These workshops on natural language theory and technology are organised by the Parlevink Project, a language theory and technology project of the . For each workshop proceedings are published containing the papers that were presented. TWLT 14, has been organised together with the German Research Center for Artificial Intelligence, DFKI Saarbrücken, Germany. The idea for this workshop grew out of a longstanding cooperation between the University of Twente, TNO-TPD in Delft and DFKI. This co-operation manifested itself for the first time in the Twenty-One project, which inspired a whole series of other projects, such as Pop-Eye and Olive, but which also led to a close contact and exchange with independently established projects such as Mulinex and MIETTA for which DFKI was responsible. All of these projects had in common that they were funded by the Telematics Application Programme of the European Commission, all, except for Twenty-One, by the Language Engineering Sector.

With the increasing availability of electronic documents and the rapid growth of the World Wide Web, the task of automatic categorization of documents became the key method for organizing the information and knowledge discovery. Proper classification of e-documents, online news, blogs, e-mails and digital libraries need text mining, machine learning and natural language processing techniques to get meaningful knowledge. The aim of this paper is to highlight the important techniques and methodologies that are employed in text documents classification, while at the same time making awareness of some of the interesting challenges that remain to be solved, focused mainly on text representation and machine learning techniques. This paper provides a review of the theory and methods of document classification and text mining, focusing on the existing literature.

Instead of trying to produce a programme to simulate the adult mind, why not rather try to produce one which simulates the child's? If this were then subjected to an appropriate course of education one would obtain the adult brain. A mis padres y a mi hermano

Integrating Different Strategies for Cross-Language Information Retrieval in the MIETTA Project Paul Buitelaar, Klaus Netter, Feiyu Xu DFKI Language Technology Lab Stuhlsatzenhausweg 3, 66123 Saarbrücken, Germany {paulb, netter, feiyu}@ dfki. de ABSTRACT In this paper ...

Automatic classification has become an important research area due to the rapid increase of digital information. Evidently, manual classification of documents is a tough work due to occurrences of vocabulary ambiguities of classification schemes as well as the language used in the text in hand.

In our study, we made an attempt to resolve this matter. This research has developed a computer program that can automatically classify a given text document based on a well developed ontology. Therefore, the user gets correct options of classification just after feeding the document to the new system. The new ontology is a domain ontology which is based on the Dewey Decimal Classification scheme and the Sears List. Data was obtained for classification accuracy for both manual and automatic methods. Moreover, the relationship between the vagueness of language in documents and the inaccuracy of classification were determined against manual classification and manual classification with automatic aid.

The research revealed that the results were more accurate when the newly developed automatic system was used. In addition, it also revealed that the vagueness of documents affects more in manual classification than in manual classification with the automatic aid.

The use of ontology in order to provide a mechanism to enable machine reasoning has continuously increased during the last few years. This paper suggests an automated method for document classification using an ontology, which expresses terminology information and vocabulary contained in Web documents by way of a hierarchical structure. Ontology-based document classification involves determining document features that represent the Web documents most accurately, and classifying them into the most appropriate categories after analyzing their contents by using at least two predefined categories per given document features. In this paper, Web pages are classified in real time not with experimental data or a learning process, but by similar calculations between the terminology information extracted from Web pages and ontology categories. This results in a more accurate document classification since the meanings and relationships unique to each document are determined.

An increasing and overwhelming amount of biomedical information is available in the research literature mainly in the form of free-text. Biologists need tools that automate their information search and deal with the high volume and ambiguity of free-text. Ontologies can help automatic information processing by providing standard concepts and information about the relationships between concepts. The Medical Subject Headings (MeSH) ontology is already available and used by MEDLINE indexers to annotate the conceptual content of biomedical articles. This paper presents a domain-independent method that uses the MeSH ontology inter-concept relationships to extend the existing MeSHbased representation of MEDLINE documents. The extension method is evaluated within a document triage task organized by the Genomics track of the 2005 Text REtrieval Conference (TREC). Our method for extending the representation of documents leads to an improvement of 18.3% over a non-extended baseline in terms of normalized utility, the metric defined for the task.

Automatic document classification due to its various applications in data mining and information technology is one of the important topics in computer science. Classification plays a vital role in many information management and retrieval tasks. Document classification, also known as document categorization, is the process of assigning a document to one or more predefined category labels. Classification is often posed as a supervised learning problem in which a set of labeled data is used to train a classifier which can be applied to label future examples [1]. Document classification includes different parts such as text processing, feature extraction, feature vector construction and final classification. Thus improvement in each part should lead to better results in document classification. In this paper, we apply machine learning methods for automatic Persian news classification. In this regard, we first try to exert some language preprocess in Hamshahri dataset [2], and then we extract a feature vector for each news text by using feature weighting and feature selection algorithms. After that we train our classifier by support vector machine (SVM) and K-nearest neighbor (KNN) algorithms. In Experiments, although both algorithms show acceptable results for Persian text classification, the performance of KNN is better in comparison to SVM.

A method of document comparison based on a hierarchical dictionary of topics (concepts) is described. The hierarchical links in the dictionary are supplied with the weights that are used for detecting the main topics of a document and for determining the similarity between two documents. The method allows for the comparison of documents that do not share any words literally but do share concepts, including comparison of documents in different languages. Also, the method allows for comparison with respect to a specific “aspect, ” i.e., a specific topic of interest (with its respective subtopics). A system Classifier using the discussed method for document classification and information retrieval is discussed. 1. Introduction*

It is well known that links are an important source of information when dealing with Web collections. However, the question remains on whether the same techniques that are used on the Web can be applied to collections of documents containing citations between scientific papers. In this work we present a comparative study of digital library citations and Web links, in the context of automatic text classification. We show that there are in fact differences between citations and links in this context. For the comparison, we run a series of experiments using a digital library of computer science papers and a Web directory. In our reference collections, measures based on co-citation tend to perform better for pages in the Web directory, with gains up to 37% over text based classifiers, while measures based on bibliographic coupling perform better in a digital library. We also propose a simple and effective way of combining a traditional text based classifier with a citation-link based classifier. This combination is based on the notion of classifier reliability and presented gains of up to 14% in micro-averaged F1 in the Web collection. However, no significant gain was obtained in the digital library. Finally, a user study was performed to further investigate the causes for these results. We discovered that misclassifications by the citation-link based classifiers are in fact difficult cases, hard to classify even for humans.

The combination of multiple features or views when representing documents or other kinds of objects usually leads to improved results in classification (and retrieval) tasks. Most systems assume that those views will be available both at training and test time. However, some views may be tooexpensive'to be available at test time. In this paper, we consider the use of Canonical Correlation Analysis to leverageexpensive'views that are available only at training time. Experimental results show that this information may ...

Pattern classification has been successfully applied in many problem domains, such as biometric recognition, document classification or medical diagnosis. Missing or unknown data are a common drawback that pattern recognition techniques need to deal with when solving real-life classification tasks. Machine learning approaches and methods imported from statistical learning theory have been most intensively studied and used in this subject.

The amount of narrative clinical text documents stored in Electronic Patient Records (EPR) of Hospital Information Systems is increasing. Physicians spend a lot of time finding relevant patient-related information for medical decision making in these clinical text documents. Thus, efficient and topical retrieval of relevant patient-related information is an important task in an EPR system. This paper describes the prototype of a medical information retrieval system (MIRS) for clinical text documents. The open-source information retrieval framework Apache Lucene has been used to implement the prototype of the MIRS. Additionally, a multi-label classification system based on the open-source data mining framework WEKA generates metadata from the clinical text document set. The metadata is used for influencing the rank order of documents retrieved by physicians. Combining information retrieval and automated document classification offers an enhanced approach to let physicians and in the near future patients define their information needs for information stored in an EPR. The system has been designed as a J2EE Web-application. First findings are based on a sample of 18,000 unstructured, clinical text documents written in German.

With the increased use of Internet, a large number of consumers first consult on line resources for their healthcare decisions. The problem of the existing information structure primarily lies in the fact that the vocabulary used in consumer queries is intrinsically different from the vocabulary represented in medical literature. Consequently, the medical information retrieval often provides poor search results. Since consumers make medical decisions based on the search results, building an effective information retrieval system becomes an essential issue. By reviewing the foundational concepts and application components of medical information retrieval, this paper will contribute to a body of research that seeks appropriate answers to a question like “How can we design a medical information retrieval system that can satisfy consumer’s information needs?”

In this work, we jointly apply several text mining methods to a corpus of legal documents in order to compare the separation quality of two inherently different document classification schemes. The classification schemes are compared with the clusters produced by the K-means algorithm. In the future, we believe that our comparison method will be coupled with semi-supervised and active learning techniques. Also, this paper presents the idea of combining K-means and Principal Component Analysis for cluster visualization. The described idea allows calculations to be performed in reasonable amount of CPU time.

The amount of narrative clinical text documents stored in Electronic Patient Records (EPR) of Hospital Information Systems is increasing. Physicians spend a lot of time finding relevant patient-related information for medical decision making in these clinical text documents. Thus, efficient and topical retrieval of relevant patient-related information is an important task in an EPR system. This paper describes the prototype of a medical information retrieval system (MIRS) for clinical text documents. The open-source information retrieval framework Apache Lucene has been used to implement the prototype of the MIRS. Additionally, a multi-label classification system based on the open-source data mining framework WEKA generates metadata from the clinical text document set. The metadata is used for influencing the rank order of documents retrieved by physicians. Combining information retrieval and automated document classification offers an enhanced approach to let physicians and in the near future patients define their information needs for information stored in an EPR. The system has been designed as a J2EE Web-application. First findings are based on a sample of 18,000 unstructured, clinical text documents written in German.

The goal of the reported research is the development of a computational approach that could help a cognitive scientist to interactively represent a learner's mental models, and to automatically validate their coherence with respect to the available experimental data. In a reported case-study, the student's mental models are inferred from questionnaires and interviews collected during a sequence of teaching sessions. These putative cognitive models are based on a theory of knowledge representation, derived from psychological results and educational studies, which accounts for the evolution of the student's knowledge over a learning period. The learning system WHY, able to handle (causal) domain knowledge, shows how to model the answers and the causal explanations given by the learner.

Classifier for the Internet Resource Discovery (ACIRD), which uses machine learning techniques to organize and retrieve Internet documents. ACIRD consists of a knowledge acquisition process, document classifier and two-phase search engine. T he knowledge acquisition process of ACIRD automatically learns classification knowledge from classified Internet documents. The document classifier applies learned classification knowledge to classify newly collected Internet documents into one or more classes.

Quantifying the concept of co-occurrence and iterated co-occurrence yields indices of similarity between words or between documents. These similarities are associated with a reversible Markov transition matrix, the formal properties of which enable us to define euclidean distances, allowing in turn to perform words-documents correspondence analysis as well as words (or documents) classifications at various co-occurrences orders.

We propose a simple Bayesian network-based text classifier, which may be considered as a discriminative counterpart of the generative multinomial naive Bayes classifier. The method relies on the use of a fixed network topology with the arcs going form term nodes to class nodes, and also on a network parametrization based on noisy or gates. Comparative experiments of the proposed method with naive Bayes and Rocchio algorithms are carried out using three standard document collections.

We propose a method which, given a document to be classified, automatically generates an ordered set of appropriate descriptors extracted from a thesaurus. The method creates a Bayesian network to model the thesaurus and uses probabilistic inference to select the set of descriptors having high posterior probability of being relevant given the available evidence (the document to be classified). Our model can be used without having preclassified training documents, although it improves its performance as long as more training data become available. We have tested the classification model using a document dataset containing parliamentary resolutions from the regional Parliament of Andalucía at Spain, which were manually indexed from the Eurovoc thesaurus, also carrying out an experimental comparison with other standard text classifiers.

This paper uses Systemic Functional Linguistic (SFL) theory as a basis for extracting semantic features of documents. We focus on the pronominal and determination system and the role it plays in constructing interpersonal distance. By using a hierarchical system model that represents the author's language choices, it is possible to construct a rich and informative feature representation. Using these systemic features, we report clear separation between registers with different interpersonal distance.

Email has become an important means of electronic communication but the viability of its usage is marred by Un-solicited Bulk Email (UBE) messages. UBE poses technical and socio-economic challenges to usage of emails. Besides, the definition and understanding of UBE differs from one person to another. To meet these challenges and combat this menace, we need to understand UBE. Towards this end, this paper proposes a classifier for UBE documents. Technically, this is an application of un-structured document classification using text content analysis and we approach it using supervised machine learning technique. Our experiments show the success rate of proposed classifier is 98.50%. This is the first formal attempt to provide a novel tool for UBE classification and the empirical results show that the tool is strong enough to be implemented in real world.

A new algorithm based on learning vector quantisation classifier is presented based on a modified proximity-measure, which enforces a predetermined correct classification level in training while using sliding-mode approach for stable variation in weight updates towards convergence. The proposed algorithm and some well-known counterparts are implemented by using Python libraries and compared in a task of text classification for document categorisation. Results reveal that the new classifier is a successful contender to those algorithms in terms of testing and training performances.

ABSTRACT: Improvements in hardware, communication technology and database have led to the explosion of multimedia information repositories. In order to provide the quality of information retrieval and the quality of services, it is necessary to consider both retrieval techniques and database architecture. This paper presents the project named VLSHDS-Very Large Scale Hypermedia Delivery System. The quality of textual information search is enhanced by using NLP techniques. The quality of service over a

Feature selection is of paramount concern in document classification process which improves the efficiency and accuracy of text classifier. Vector Space Model is used to represent the "Bag of Word" BOW of the documents with term weighting phenomena. Documents representing through this model has some limitations that is, ignoring term dependencies, structure and ordering of the terms in documents. To overcome this problem semantic base feature vector is proposed. That is used to extracts the concept of term, co-occurring and associated terms using ontology. The proposed method is applied on small documents dataset, which shows that this method outperforms then term frequency/ inverse document frequency (TF-IDF) with BOW feature selection method for text classification.

In this paper we propose a matching algorithm for measuring the structural similarity between an XML document and a DTD. The matching algorithm, by comparing the document structure against the one the DTD requires, is able to identify commonalities and differences. Differences can be due to the presence of extra elements with respect to those the DTD requires and to the absence of required elements. The evaluation of commonalities and differences gives raise to a numerical rank of the structural similarity. Moreover, in the paper, some applications of the matching algorithm are discussed. Specifically, the matching algorithm is exploited for the classification of XML documents against a set of DTDs, the evolution of the DTD structure, the evaluation of structural queries, the selective dissemination of XML documents, and the protection of XML document contents. r

We present a novel approach for classifying documents that combines different pieces of evidence (e.g., textual features of documents, links, and citations) transparently, through a data mining technique which generates rules associating these pieces of evidence to predefined classes. These rules can contain any number and mixture of the available evidence and are associated with several quality criteria which can be used in conjunction to choose the "best" rule to be applied at classification time. Our method is able to perform evidence enhancement by link forwarding/backwarding (i.e., navigating among documents related through citation), so that new pieces of link-based evidence are derived when necessary. Furthermore, instead of inducing a single model (or rule set) that is good on average for all predictions, the proposed approach employs a lazy method which delays the inductive process until a document is given for classification, therefore taking advantage of better qualitative evidence coming from the document. We conducted a systematic evaluation of the proposed approach using documents from the ACM Digital Library and from a Brazilian Web directory. Our approach was able to outperform in both collections all classifiers based on the best available evidence in isolation as well as state-of-the-art multi-evidence classifiers. We also evaluated our approach using the standard WebKB collection, where our approach showed gains of 1% in accuracy, being 25 times faster. Further, our approach is extremely efficient in terms of computational performance, showing gains of more than one order of magnitude when compared against other multi-evidence classifiers.

In this paper we present the Dual Support Apriori for Temporal data (DSAT) algorithm. This is a novel technique for discovering Jumping Emerging Patterns (JEPs) from time series data using a sliding window technique. Our approach is particularly effective when performing trend analysis in order to explore the itemset variations over time. Our proposed framework is different from the previous work on JEP in that we do not rely on itemsets borders with a constrained search space. DSAT exploits previously mined time stamped data by using a sliding window concept, thus requiring less memory, minimum computational cost and very low dataset accesses. DSAT discovers all JEPs, as in "naïve" approaches, but utilises less memory and scales linearly with large datasets sets as demonstrated in the experimental section.