In this paper, we present an effective saliency detection method based on color contrast and similarity distribution information. We define a salient object by measuring its color contrast to surroundings and compactness similarity distribution. In the method, both global and local color contrasts are considered to measure the color distinctiveness, and similarity distribution is used to measur...

This paper describes a Chinese word segmentation system that is based on majority voting among three models: a forward maximum matching model, a conditional random field (CRF) model using maximum subword-based tagging, and a CRF model using minimum subwordbased tagging. In addition, it contains a post-processing component to deal with inconsistencies. Testing on the closed track of CityU, MSRA ...

Detection is a fundamental problem in many applications. In many cases, knowing the presence of underlying objects is of significant importance. Multiscale methods have been demonstrated to be advantageous in solving this problem. Besides theoretical results that have been achieved, this paper discusses how the ‘most powerful’ detection can be realized, for a set of specifically organized under...

This paper describes the MSRA-USTC-SJTU experiments for TRECVID 2007. We performed the experiments in high-level feature extraction and automatic search tasks. For high-level feature extraction, we investigated the benefit of unlabeled data by semi-supervised learning, and the multi-layer (ML) multi-instance (MI) relation embedded in video by MLMI kernel, as well as the correlations between con...

This paper briefly describes our system in The Fourth SIGHAN Bakeoff. Discriminative models including maximum entropy model and conditional random fields are utilized in Chinese word segmentation and named entity recognition with different tag sets and features. Transformation-based learning model is used in part-of-speech tagging. Evaluation shows that our system achieves the F-scores: 92.64% ...

Bu çalışmada, Isparta ilinde satışı yapılan Tulum peyniri örneklerinde Staphylococcus aureus yayğınlığının tespiti ve S. izolatlarında antibiyotik direnç profillerinin genlerinin belirlenmesi amaçlanmıştır. 75 örneğinden toplam 141 muhtemel kolonisi izole edilmiştir. Gram boyama katalaz testi sonucu izolatların 100’ünün pozitif kok morfolojisinde özellik gösterdiği tespit türü üyesi olup olmadı...

Currently most of state-of-the-art methods for Chinese word segmentation are based on supervised learning, whose features aremostly extracted from a local context. Thesemethods cannot utilize the long distance information which is also crucial for word segmentation. In this paper, we propose a novel neural network model for Chinese word segmentation, which adopts the long short-term memory (LST...

Device-to-Device (D2D) has been a potential technology to improve the sum-rate of cellular networks, especially in local communication. By reusing the resource of cellular user equipment (UE), D2D can enhance the spectrum efficiency, but at the cost of introducing extra co-channel interference. In this paper, we adopt a resource reusing mechanism in which multiple D2D pairs can share multiple r...

In this paper, we present an image saliency detection method by constructing graph model. We extract color, texture and compactness features and segment superpixels from an input image to construct a graph model. Then, saliency for each is measured by calculating random walker probability on the node. Extensive results on MSRA dataset containing 1000 test images with ground truths demonstrate t...

This paper presents systems submitted to the close track of Fourth SIGHAN Bakeoff. We built up three systems based on Conditional Random Field for Chinese Word Segmentation, Named Entity Recognition and Part-Of-Speech Tagging respectively. Our systems employed basic features as well as a large number of linguistic features. For segmentation task, we adjusted the BIO tags according to confidence...