This paper presents a new method of functional classification of text blocks on a document. It is based on texture analysis and unsupervised classification. Texture is used here to define different classes of text blocks in the document and to direct a possible way of exploration from the most eye-catching data to the less significant text block. The typographical properties of blocks are chara...

We study an open text mining problem – discovering concept-level event associations from a text stream. We investigate the importance and challenge of this task and propose a novel solution by using event sequential patterns. The proposed approach can discover important event associations implicitly expressed. The discovered event associations are general and useful as knowledge for application...

0

This paper demonstrates a new method for leveraging unstructured annotations to infer semantic document properties. We consider the domain of product reviews, which are often annotated by their authors with free-text keyphrases, such as “a real bargain” or “good value.” We leverage these unstructured annotations by clustering them into semantic properties, and then tying the induced clusters to...

Due to their expressive powers, tables are popularly used in documents. The presence of tables in documents creates needs for table-processing algorithms. Over the years, many systems were developed , and their performance, usually in terms of recalls and precisions, were reported. According to current practice, systems are tested using some propriety data. This raises some important questions:...

Previous work introduced transition-based algorithms to form a unified architecture of parsing rhetorical structures (including span, nuclearity and relation), but did not achieve satisfactory performance. In this paper, we propose that transition-based model is more appropriate for parsing the naked discourse tree (i.e., identifying span and nuclearity) due to data sparsity. At the same time, ...

Abstract For text categorization, one kind of FCM, two-level FCM, is proposed. The FCM includes two kinds of nodes that are attribute nodes and class nodes forming two levels, and has the intelligence advantages of structured knowledge representing and numeric matrix inference. The FCM mining method combines mutual information and gradient descent, and considers two stable states in FCM learnin...

Partially supervised text classification has received great research attention since it only uses positive and unlabeled examples as training data. This problem can be solved by automatically labeling some negative (and more positive) examples from unlabeled examples before training a text classifier. But it is difficult to guarantee both high quality and quantity of the new labeled examples. I...

Subspace detection and processing is receivingmore attention nowadays as a method to speed up searchand reduce processing overload. Subspace Learningalgorithms try to detect low dimensional subspaces in thedata which minimize the intra-class separation whilemaximizing the inter-class separation. In this paper wepresent a novel technique using the maximum significance...

Text complexity can be reduced by making different choices at the lexical and grammatical levels. However, discourse-level choices may also affect a text’s complexity. In a coherent text, explicit discourse relations (e.g. CAUSE, CONDITION) are expressed using discourse markers (e.g. since, because, etc.) that may be preferred for texts at different readability levels. In this paper, we investi...