Extracellular carbohydrate polymers from five desert soil algae with different cohesion in the stabilization of fine sand grain

Extracellular polymeric substances (EPS) from four filamentous cyanobacteria Microcoleus vaginatus, Scytonema javanicum, Phormidium tenue and Nostoc sp. and a coccoid single-cell green alga Desmococcus olivaceus that had been separated from desert algal crusts of Tegger desert of China, were investigated for their chemical composition, structure,and physical properties. The EPS contained 7.5-50.3% protein (in polymers ranging from 14 to more than 200 kD, SDS-PAGE) and 16.2-46.5% carbohydrate (110-460 kD, GFC). 6-12 kinds of monosaccharides, including 2-O-methyl rhamnose, 2-O-methyl glucose, and N-acetyl glucosamine were found. The main carbohydrate chains from M. vaginatus and S. javanicum consisted mainly of equal proportion of Man, Gal and Glc, that from P. tenue consisted mainly of arabinose, glucose and rhamnose. Arabinose was present in pyranose form, mainly alpha-L 1 --> 3 linked, with branches on C4 of almost half of the units. Glucose was responsible for the terminal units, in addition of having some units as beta1 --> 3 and some as beta1 --> 4 linked. Rhamnose was mainly 1 --> 3 linked with branches on C2 on half of the units. The carbohydrate polymer from D. olivaceus was composed mainly of beta1 --> 4 linked xylose, galactose and glucose. The galactose part was present both in beta-pyranose and -furanose forms. Arabinose in alpha-L-furanose form was mainly present as 1 --> 2 and 1 --> 2, 5 linked units, rhamnose only as alpha 1 --> 3 and xylose as beta 1 --> 4. The backbone of the polysaccharide from Nostoc sp. was composed of beta-1 --> 4 linked xylose, galactose and glucose. Most of the glucose was branched on position C6, terminal glucose and 2-O-methyl glucose units are also present. The relationship between structure, physical properties and potential biological function is discussed. (C) 2003 Elsevier Ltd. All rights reserved.

Study on text-dependent speaker recognition based on Biomimetic Pattern Recognition

We studied the application of Biomimetic Pattern Recognition to speaker recognition. A speaker recognition neural network using network matching degree as criterion is proposed. It has been used in the system of text-dependent speaker recognition. Experimental results show that good effect could be obtained even with lesser samples. Furthermore, the misrecognition caused by untrained speakers occurring in testing could be controlled effectively. In addition, the basic idea "cognition" of Biomimetic Pattern Recognition results in no requirement of retraining the old system for enrolling new speakers.

Study on text-dependent speaker recognition based on Biomimetic Pattern Recognition

We studied the application of Biomimetic Pattern Recognition to speaker recognition. A speaker recognition neural network using network matching degree as criterion is proposed. It has been used in the system of text-dependent speaker recognition. Experimental results show that good effect could be obtained even with lesser samples. Furthermore, the misrecognition caused by untrained speakers occurring in testing could be controlled effectively. In addition, the basic idea "cognition" of Biomimetic Pattern Recognition results in no requirement of retraining the old system for enrolling new speakers.

Smoothing LDA Model for Text Categorization

Abstract. Latent Dirichlet Allocation (LDA) is a document level language model. In general, LDA employ the symmetry Dirichlet distribution as prior of the topic-words’ distributions to implement model smoothing. In this paper, we propose a data-driven smoothing strategy in which probability mass is allocated from smoothing-data to latent variables by the intrinsic inference procedure of LDA. In such a way, the arbitrariness of choosing latent variables'priors for the multi-level graphical model is overcome. Following this data-driven strategy,two concrete methods, Laplacian smoothing and Jelinek-Mercer smoothing, are employed to LDA model. Evaluations on different text categorization collections show data-driven smoothing can significantly improve the performance in balanced and unbalanced corpora.

semantic oriented ontology cohesion metrics for ontology-based systems

Ontologies play a core role to provide shared knowledge models to semantic-driven applications targeted by Semantic Web. Ontology metrics become an important area because they can help ontology engineers to assess ontology and better control project management and development of ontology based systems, and therefore reduce the risk of project failures. In this paper, we propose a set of ontology cohesion metrics which focuses on measuring (possibly inconsistent) ontologies in the context of dynamic and changing Web. They are: Number of Ontology Partitions (NOP), Number of Minimally Inconsistent Subsets (NMIS) and Average Value of Axiom Inconsistencies (AVAI). These ontology metrics are used to measure ontological semantics rather than ontological structure. They are theoretically validated for ensuring their theoretical soundness, and further empirically validated by a standard test set of debugging ontologies. The related algorithms to compute these ontology metrics also are discussed. These metrics proposed in this paper can be used as a very useful complementarity of existing ontology cohesion metrics.