基于Bert预训练模型的虚假新闻文本检测

虚假新闻剥夺了人们获取真相的权利,也给社会和国家带来了许多危害。文中以虚假新闻文本为例,分析和验证了多种预训练语言模型在虚假新闻文本分类上的分类效果。经实验证明,相较于其他语言模型,Bert预训练语言模型取得了较好的结果,预测准确率为86.97%,召回率为88.12%,F1值为87.54%。     

Influence of expertise on semantic categorization of wine odors

Aromatic characterization is a key element of enhancing one’s knowledge of wine. While several studies have investigated the importance of wine expertise in the ability to perform odor-related sensory tasks, little attention has been paid to the influence of expertise on the semantic categorization of wine odors. To bridge this gap, this study aimed to explore the influence of a subject’s expertise on the semantic representation of wine odors by means of a free sorting task. For this purpose, 156 subjects were recruited. Their level of expertise was measured using a questionnaire and the data analysis revealed four clusters of subjects with a gradual level of expertise. Subjects also performed a sorting task on 96 odor terms. From the number and the size of odor groups formed, as well as the additive tree representation and the consensus partition between the terms for each expertise level, we observed that all subjects, regardless of their experience, had largely the same semantic categorization of wine-odor attributes, which was mainly shaped by the sources of the odorants. It appeared that level of wine expertise played a minor role in creating the semantic representation of wine odors, affecting mainly the knowledge of specialized terms.     

Spatially constrained sparse coding scheme for natural scene categorization

Coding and pooling, the major two sequential procedures in sparse coding based scene categorization systems, have drawn much attention in recent years. Yet improvements have been made for coding or pooling separately, this paper proposes a spatially constrained scheme for sparse coding on both steps. Specifically, we employ the m-nearest neighbors of a local feature in the image space to improve the consistency of coding. The benefit is that similar image features will be encoded with similar codewords, which reduced the stochasticity of a conventional coding strategy. We also show that the Viola–Jones algorithm, which is well-known in face detection, can be tailored to learning receptive fields, embedding the spatially constrained information on the pooling step. Extensive experiments on the UIUC sport event, 15 natural scenes and the Caltech 101 database suggests that scene categorization performance of several popular algorithms can be ubiquitously improved by incorporating the proposed two spatially constrained sparse coding scheme.     

DFD-SS: Document Forgery Detection using Spectral – Spatial Features for Hyperspectral Images

In the present era of machines and edge-cutting technologies, still document frauds persist. They are done intuitively by using almost identical inks, that it becomes challenging to detect them—this demands an approach that efficiently investigates the document and leaves it intact. Hyperspectral imaging is one such a type of approach that captures the images from hundreds to thousands of spectral bands and analyzes the images through their spectral and spatial features, which is not possible by conventional imaging. Deep learning is an edge-cutting technology known for solving critical problems in various domains. Utilizing supervised learning imposes constraints on its usage in real scenarios, as the inks used in forgery are not known prior. Therefore, it is beneficial to use unsupervised learning. An unsupervised feature extraction through a Convolutional Autoencoder (CAE) followed by Logistic Regression (LR) for classification is proposed (CAE-LR). Feature extraction is evolved around spectral bands, spatial patches, and spectral-spatial patches. We inspected the impact of spectral, spatial, and spectral-spatial features by mixing inks in equal and unequal proportion using CAE-LR on the UWA writing ink hyperspectral images dataset for blue and black inks. Hyperspectral images are captured at multiple correlated spectral bands, resulting in information redundancy handled by restoring certain principal components. The proposed approach is compared with eight state-of-art approaches used by the researchers. The results depicted that by using the combination of spectral and spatial patches, the classification accuracy enhanced by 4.85% for black inks and 0.13% for blue inks compared to state-of-art results. In the present scenario, the primary area concern is to identify and detect the almost similar inks used in document forgery, are efficiently managed by the proposed approach.     

Gas-cooled reactor thermal hydraulic analyses with MELCOR

Pursuant to the Energy Policy Act of 2005, the High Temperature Gas-Cooled Reactor (HTGR) has been selected as the reference design for the Next Generation Nuclear Plant (NGNP). Stemming from a U.S. Nuclear Regulatory Commission (NRC) HTGR research initiative, a need was identified for validation of systems-level computer code modeling capabilities in anticipation of the eventual need to perform licensing analyses. Because the NRC has used MELCOR for light water reactors (LWR) in the past and because MELCOR was recently updated to include gas-cooled reactor (GCR) physics models, MELCOR is among the system codes of interest to the NRC. This paper describes MELCOR modeling of the General Atomics' Modular High Temperature Gas-Cooled Reactor (MHTGR). The MHGTR is a suitable design for demonstration of MELCOR GCR modeling competency for two reasons: 1) the MHTGR is a predecessor to the more advanced General Atomics’ Gas-Turbine Modular High Temperature Reactor (GTMHR), and 2) experimental data useful for benchmark calculations may soon become available. Using the most complete literature references available for the MHTGR design, researchers at Texas A&M University (TAMU) constructed a MELCOR input deck for the MHTGR to partially validate MELCOR GCR modeling capabilities. Normal and off-normal system operating conditions were modeled with appropriate boundary and initial conditions. MELCOR predictions of system response were obtained for steady-state, pressurized conduction cool-down (PCC), and depressurized conduction cool-down (DCC) scenarios. Code results were checked against nominal MHTGR design parameters, physical intuition, and anticipated GCR thermal hydraulic response. No inherent deficiencies in MELCOR modeling capability were observed, suggesting that the newly-implemented GCR models are adequate for systems-level analysis. If and when experimental benchmark data becomes available, further validation activities may proceed given the modeling efforts discussed herein.     

Topic Tomographies (TopTom): a visual approach to distill information from media streams

In this paper we present Top Tom, a digital platform whose goal is to provide analytical and visual solutions for the exploration of a dynamic corpus of user‐generated messages and media articles, with the aim of i) distilling the information from thousands of documents in a low‐dimensional space of explainable topics, ii) cluster them in a hierarchical fashion while allowing to drill down to details and stories as constituents of the topics, iii) spotting trends and anomalies. Top Tom implements a batch processing pipeline able to run both in near‐real time with time stamped data from streaming sources and on historical data with a temporal dimension in a cold start mode. The resulting output unfolds along three main axes: time, volume and semantic similarity (i.e. topic hierarchical aggregation). To allow the browsing of data in a multiscale fashion and the identification of anomalous behaviors, three visual metaphors were adopted from biological and medical fields to design visualizations, i.e. the flowing of particles in a coherent stream, tomographic cross sectioning and contrast‐like analysis of biological tissues. The platform interface is composed by three main visualizations with coherent and smooth navigation interactions: calendar view, flow view, and temporal cut view. The integration of these three visual models with the multiscale analytic pipeline proposes a novel system for the identification and exploration of topics from unstructured texts. We evaluated the system using a collection of documents about the emerging opioid epidemics in the United States.     

Analyzing Dynamic Change in Social Network Based on Distribution-Free Multivariate Process Control Method

Social organizations can be represented by social network because it can mathematically quantify and represent complex interrelated organizational behavior. Exploring the change in dynamic social network is essential for the situation awareness of the corresponding social organization. Social network usually evolves gradually and slightly, which is hard to be noticed. The statistical process control techniques in industry field have been used to distinguish the statistically significant change of social network. But the original method is narrowed due to some limitation on measures. This paper presents a generic framework to address the change detection problem in dynamic social network and introduces a distribution-free multivariate control charts to supervise the changing of social network. Three groups of network parameters are integrated together in order to achieve a comprehensive view of the dynamic tendency. The proposed approaches handle the non-Gaussian data based on categorizing and ranking. Experiments indicate that nonparametric multivariate procedure is promising to be applied to social network analysis.     

The emotional quality of speech in voice services

The emotional quality of speech is defined as the global qualitative and hedonic impressions experienced by listeners. This research investigated the emotional quality of speech samples used in voice services. In a first experiment listening tests were conducted using 200 messages generated by 20 female speakers who pronounced two sentences in five elocution styles. Listeners grouped the messages according to similarities in terms of the impression of the messages. Verbal comments regarding hedonic effect on the listener and acoustic parameters of the voices' timbre and intonation were analysed. In a second experiment, the 200 messages were evaluated according to 20 criteria extracted from the first experiment. The results produced a precise perceptive portrait for each sequence, giving a full picture of the listeners' impressions of what they heard. The results can be applied to the design of voice services, as was done for the voicemail of France Telecom Orange.     

Hierarchical behavior categorization using correlation based adaptive resonance theory

This paper introduces a new model for robot behavior categorization. Correlation based adaptive resonance theory (CobART) networks are integrated hierarchically in order to develop an adequate categorization, and to elicit various behaviors performed by the robot. The proposed model is developed by adding a second layer CobART network which receives first layer CobART network categories as an input, and back-propagates the matching information to the first layer networks. The first layer CobART networks categorize self-behavior data of a robot or an object in the environment while the second layer CobART network categorizes the robot's behavior with respect to its effect on the object. Experiments show that the proposed model generates reasonable categorization of behaviors being tested. Moreover, it can learn different forms of the behaviors, and it can detect the relations between them. In essence, the model has an expandable architecture and it contains reusable parts. The first layer CobART networks can be integrated with other CobART networks for another categorization task. Hence, the model presents a way to reveal all behaviors performed by the robot at the same time.