Entity Recognition (ER) is a key component of relation extraction systems and many other natural-language processing applications. Unfortunately, most ER are restricted to produce labels from small set entity classes, e.g., person, organization, location or miscellaneous. In order intelligently understand text extract wide range information, it useful more precisely determine the semantic classes entities mentioned in unstructured text. This paper defines fine-grained 112 tags, formulates...

Recent research on entity linking (EL) has introduced a plethora of promising techniques, ranging from deep neural networks to joint inference. But despite numerous papers there is surprisingly little understanding the state art in EL. We attack this confusion by analyzing differences between several versions EL problem and presenting simple yet effective, modular, unsupervised system, called Vinculum, for linking. conduct an extensive evaluation nine data sets, comparing Vinculum with two...

Since about 100 years ago, to learn the intrinsic structure of data, many representation learning approaches have been proposed, including both linear ones and nonlinear ones, supervised unsupervised ones. Particularly, deep architectures are widely applied for in recent years, delivered top results tasks, such as image classification, object detection speech recognition. In this paper, we review development data methods. Specifically, investigate traditional feature algorithms...

Since Pearson developed principal component analysis (PCA) in 1901, feature learning (or called representation learning) has been studied for more than 100 years. During this period, many “shallow” methods have proposed based on various criteria and techniques, until the popular deep research recent In advanced review, we describe historical profile of shallow introduce important developments models. Particularly, survey architectures with benefits from optimization their width depth, as...

Angli Liu, Stephen Soderland, Jonathan Bragg, Christopher H. Lin, Xiao Ling, Daniel S. Weld. Proceedings of the 2016 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2016.

The speed of data retrieval qualitatively affects how analysts visually explore and analyze their data. To ensure smooth interactions in massive time series datasets, one needs to address the challenges computing ad hoc queries, distributing query load, hiding system latency. In this paper, we present ATLAS, a visualization tool for temporal that addresses these issues using combination high performance database technology, predictive caching, level detail management. We demonstrate ATLAS...

Clouds and accompanying shadows, which exist in optical remote sensing images with high possibility, can degrade or even completely occlude certain ground-cover information images, limiting their applicabilities for Earth observation, change detection, land-cover classification. In this paper, we aim to deal cloud contamination problems the objective of generating cloud-removed images. Inspired by low-rank representation together sparsity constraints, propose a coarse-to-fine framework...

To ensure the performance of online service systems, their status is closely monitored with various software and system metrics. Performance anomalies represent degradation issues (e.g., slow response) systems. When performing anomaly detection over metrics, existing methods often lack merit interpretability, which vital for engineers analysts to take remediation actions. Moreover, they are unable effectively accommodate ever-changing services in an fashion. address these limitations, this...

ABSTRACT Building damage assessment in the face of natural disasters is crucial for economic development, disaster relief, and post‐disaster reconstruction. However, existing algorithms often overlook impact class when extracting difference features from high‐resolution pre‐ image pairs obtained through satellite remote sensing, without considering influence type, that is, different ways which affect buildings. To address this limitation, we propose U2DDS‐Net, a two‐stage model based on...

This paper presents a network traffic analysis system that couples visual with declarative knowledge representation. The supports multiple iterations of the sense-making loop analytic reasoning by allowing users to save discoveries as they are found and reuse them in future iterations. We show how representation can be used improve both representations basic analytical tasks filtering changing level detail. describe produce models patterns, results from classifying one day our laboratory

The increasing number of vehicles in high density, urban areas is leading to significant parking space shortages. While systems have been developed enable visibility into vacancies for drivers, most rely on costly, dedicated sensor devices that require installation costs. proliferation inexpensive Internet Things (IoT) enables the use compute platforms with integrated cameras could be used monitor occupancy. However, even camera-captured images, manual specification locations required before...

Water body extraction from remote sensing imagery is an essential and nontrivial issue due to the complexity of spectral characteristics various kinds water bodies redundant background information. An automatic multifeature (MFWE) method integrating spatial features proposed in this letter for GF-1 multispectral unsupervised way. This first discusses a feature index, called pixel region index (PRI), describe smoothness local area surrounding pixel. PRI advantageous assisting normalized...

Limited by the noise, missing data and varying sampling density of point clouds, planar primitives are prone to be lost during plane segmentation, leading topology errors when reconstructing complex building models. In this paper, a pipeline recover broken (TopoLAP) is proposed reconstruct level details 3 (LoD3) Firstly, segmented from incomplete clouds feature lines detected both images. Secondly, structural contours each segment reconstructed subset selection intersections these lines....

Optical remote sensing has emerged as a crucial technique for earth observation. However, interference of clouds and fog can adversely affect the spatial spectral information images, presenting significant challenges in interpreting data limiting its availability. Moreover, existing methods addressing issue cloud occlusion primarily rely on either physical models or neural networks, lacking comprehensive integration advantages offered by both approaches. So, we propose an end-to-end...

Nadir viewing satellite image is an effective data source to generate orthomosaics. Because of the georeferencing error images, block adjustment first step orthomosaic generation over a large area. However, geometric relationship neighboring orbits nadir images not rigid enough. This paper proposes new rational function model (RFM) approach that constrains tie point elevation enhance relative rigidity. By interpolating elevations points in digital (DEM) and estimating priori errors...

Automatic registration of multimodal remote sensing images, which is a critical prerequisite in range applications (e.g. image fusion, mosaic, and analysis), continues to be fundamental challenging problem. In this paper, we propose novel extended phase correlation algorithm based on Log-Gabor filtering (LGEPC) for the images with nonlinear radiometric differences geometric rotation, scale, translation). Our focuses two problems that traditional algorithms cannot well handle: 1) significant...

Anthony Chen, Pallavi Gudipati, Shayne Longpre, Xiao Ling, Sameer Singh. Proceedings of the 59th Annual Meeting Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

This paper presents a novel image matching method for multi-source satellite images, which integrates global Shuttle Radar Topography Mission (SRTM) data and segmentation to achieve robust numerous correspondences. first generates the epipolar lines as geometric constraint assisted by SRTM data, after seed points are selected matched. To produce more reliable results, region segmentation-based propagation is proposed in this paper, whereby segmentations extracted considered be spatial...

Due to the inevitable existence of clouds and their shadows in optical remote sensing images, certain ground-cover information is degraded or even appears be missing, which limits analysis utilization. Thus, cloud removal great importance facilitate downstream applications. Motivated by sparse representation techniques have obtained a stunning performance variety applications, including target detection, anomaly so on; we propose two-pass robust principal component (RPCA) framework for...