Clear Semantics Research Articles

Overlapping object detection with adaptive Gaussian sample division and asymmetric weighted loss

Existing deep learning based detectors are mostly designed for scenes with sparsely distributed objects. However, in certain scenarios such as dense crowds, objects often overlap severely. The dense anchor arrangement in anchor-based detectors is not quite suitable for the overlapping object detection. Anchor-free detectors have the potential to achieve high-performance in overlapping object detection, but troubled by the extreme imbalance of positive and negative samples. To this end, we propose an anchor-free overlapping object detector. Our adaptive Gaussian sample division (AGSD) can effectively allocate positive and negative samples with clear semantics to overlapping objects. Secondly, asymmetric weighted loss (AW Loss) adapts to continuous positive and negative sample values, thereby improving the classification ability of the detector. Lastly, our global location distribution head (GLD head) can introduce the supervision of overlapping object distributions. To verify the effectiveness of our method, we construct a large-scale high-quality overlapping object detection dataset containing 6173 images and 17,725 annotations. Compared with mainstream object detector, our method achieves the best performance of AP50 at 96.71%.

A Symmetrical Fuzzy Neural Network Regression Method Coordinating Structure and Parameter Identifications for Regression

Fuzzy neural networks have both the interpretability of fuzzy systems and the self-learning ability of neural networks, but they will face the challenge of “rule explosion” when dealing with high-dimensional data. Moreover, the structure and parameter identifications of models are generally performed in two stages, and this always attends to one thing and loses another in terms of interpretability and predictive performance. In this paper, a fuzzy neural network regression method (FNNR) that coordinates structure identification and parameter identification is proposed. To alleviate the problem of rule explosion, the structure identification and parameter identification are coordinated in the training process, and the numbers of fuzzy rules and fuzzy partitions are effectively limited, while the parameters of fuzzy rules are optimized. The symmetrical architecture of the FNNR is designed for automatic structure identification. An alternate training strategy is adopted by treating discrete and continuous parameters differently, and thus the convergence efficiency of the algorithm is improved. To enhance interpretability, regularized terms are designed from fuzzy rule level and fuzzy partition level to guide the model to learn fuzzy rules with simple structures and clear semantics. The experimental results show that the proposed method has both a compact structure and high precision.

IPC-Net: Incomplete point cloud classification network based on data augmentation and similarity measurement

Existing point cloud classification researches are usually conducted on datasets with complete structure and clear semantics. However, in real point cloud scenes, the occlusion and truncation may destroy the completeness of objects affecting the classification performance. To solve this problem, we propose an incomplete point cloud classification network (IPC-Net) with data augmentation and similarity measurement. The proposed network learns the feature representation of incomplete point clouds and the semantic differences compared to the complete ones for classification. Specifically, IPC-Net adopts a random erasing-based data augmentation to deal with incomplete point clouds. IPC-Net also introduces an auxiliary loss function weighted by attention scores to measure the similarity between the incomplete and the complete point clouds. Extensive experiments verify that IPC-Net has the ability to classify incomplete point clouds and significantly improves the robustness of point cloud classification under different completeness.

INPHOVIS: Interactive visual analytics for smartphone-based digital phenotyping

Digital phenotyping is the characterization of human behavior patterns based on data from digital devices such as smartphones in order to gain insights into the users’ state and especially to identify ailments. To support supervised machine learning, digital phenotyping requires gathering data from study participants’ smartphones as they live their lives. Periodically, participants are then asked to provide ground truth labels about their health status. Analyzing such complex data is challenging due to limited contextual information and imperfect health/wellness labels. We propose INteractive PHOne-o-typing VISualization (INPHOVIS), an interactive visual framework for exploratory analysis of smartphone health data to study phone-o-types. Prior visualization work has focused on mobile health data with clear semantics such as steps or heart rate data collected using dedicated health devices and wearables such as smartwatches. However, unlike smartphones which are owned by over 85 percent of the US population, wearable devices are less prevalent thus reducing the number of people from whom such data can be collected. In contrast, the “low-level” sensor data (e.g., accelerometer or GPS data) supported by INPHOVIS can be easily collected using smartphones. Data visualizations are designed to provide the essential contextualization of such data and thus help analysts discover complex relationships between observed sensor values and health-predictive phone-o-types. To guide the design of INPHOVIS, we performed a hierarchical task analysis of phone-o-typing requirements with health domain experts. We then designed and implemented multiple innovative visualizations integral to INPHOVIS including stacked bar charts to show diurnal behavioral patterns, calendar views to visualize day-level data along with bar charts, and correlation views to visualize important wellness predictive data. We demonstrate the usefulness of INPHOVIS with walk-throughs of use cases. We also evaluated INPHOVIS with expert feedback and received encouraging responses.

Efficient Vision Transformer via Token Merger.

Vision Transformers (ViTs) split an image into fixed-size patches as tokens. This strategy has succeeded in computer vision tasks, but introduces considerable tokens similar in semantics and appearances. This work proposes Token Merger to spot redundant tokens and merge them into a compact representation to accelerate ViTs. For each forward inference, the Token Merger first identifies meta tokens to represent meaningful cues of the image content, then adaptively merges similar tokens into a uniform one referring to meta tokens. To pursue a reasonable tradeoff between accuracy and efficiency, we further introduce learnable gates to adaptively decide the token merge ratios of different layers. As a generalizable module, Token Merger can be easily plugged into different layers of ViTs to boost their efficiency. Visualizations show that Token Merger progressively merges tokens and finally learns a compact set of tokens representing clear semantics. Compared with token pruning methods, Token Merger is more effective in preserving meaning contextual cues, thus performs and generalizes substantially better in different vision tasks. Extensive experiments and comparisons with other state-of-the-art downsampling methods also demonstrate its promising performance. For instance, it reduces 95% tokens and accelerates the inference speed by 62%. Meanwhile, the ImageNet classification accuracy only drops by 0.4%. The code will be available.

Application of Intelligent Analysis Technology of Football Video Based on Online Target Tracking Algorithm of Motion Characteristics in Football Training.

Because the scene of football video is fixed and simple, the events in football video, such as shooting and offside , also have clear semantics. At the same time, they have sufficient domain knowledge and have broad application prospects. The research framework of sports video intelligence analysis is usually regarded as a three-level framework, namely, the low-level feature layer, the middle-level key primitive generation layer, and the high-level event analysis layer. A MF_O2T (Moving Feature Online Target Tracking) algorithm is proposed. First, based on the marked first frame image, this algorithm extracts a set of standardized local images from the target areas of visible and infrared images as target convolution filters, removes the main color of the stadium by using HSV color space nonuniform quantization algorithm, and extracts the histogram of the main color of players in upper and lower blocks. Experimental results show that the algorithm designed in this paper has strong robustness, can better adapt to player tracking in different scenes of football video, and meets the real-time requirements of football training.

Planning for potential: efficient safe reinforcement learning

Deep reinforcement learning (DRL) has shown remarkable success in artificial domains and in some real-world applications. However, substantial challenges remain such as learning efficiently under safety constraints. Adherence to safety constraints is a hard requirement in many high-impact application domains such as healthcare and finance. These constraints are preferably represented symbolically to ensure clear semantics at a suitable level of abstraction. Existing approaches to safe DRL assume that being unsafe leads to low rewards. We show that this is a special case of symbolically constrained RL and analyze a generic setting in which total reward and being safe may or may not be correlated. We analyze the impact of symbolic constraints and identify a connection between expected future reward and distance towards a goal in an automaton representation of the constraints. We use this connection in an algorithm for learning complex behaviors safely and efficiently. This algorithm relies on symbolic reasoning over safety constraints to improve the efficiency of a subsymbolic learner with a symbolically obtained measure of progress. We measure sample efficiency on a grid world and a conversational product recommender with real-world constraints. The so-called Planning for Potential algorithm converges quickly and significantly outperforms all baselines. Specifically, we find that symbolic reasoning is necessary for safety during and after learning and can be effectively used to guide a neural learner towards promising areas of the solution space. We conclude that RL can be applied both safely and efficiently when combined with symbolic reasoning.

Manifestation of Physis in Vilnius County Drymonymy

    The current research aims to shed some light on the conceptualization of physis1 in the drymonyms from present-day Vilnius County. The research is limited to the analysis of 65 names selected from the corpus of 787 drymonyms currently officially functioning in the territory of Vilnius County. Data were collected from printed (Onomasticon of Lithuania’s Forests (Part I, 1994)), electronic (Electronic Nature Catalogue (www.ezerai21.lt)) sources, and interactive maps (Portal of Spacial Information of Lithuania (www.geoporta.lt)). The study reveals that the domain of physis, in a broad sense, is an abundant motivation source. The drymonyms were divided into two groups, those of clear semantics and motivation and those of obscure semantics and motivation. The drymonyms of clear semantics and motivation conceptualize dominant vegetation species (21), fauna (9), forest type (15), and the peculiarities of the landscape (17). The group of obscure semantics and motivation presents an overview of names, the conceptualization of which is ambiguous due to the polysemy of the motivating appellative (1), the appellative or nominal nature of themotivating concept (5), or even both (3).

A Formal Framework for Complex Event Recognition

Complex event recognition (CER) has emerged as the unifying field for technologies that require processing and correlating distributed data sources in real time. CER finds applications in diverse domains, which has resulted in a large number of proposals for expressing and processing complex events. Existing CER languages lack a clear semantics, however, which makes them hard to understand and generalize. Moreover, there are no general techniques for evaluating CER query languages with clear performance guarantees.In this article, we embark on the task of giving a rigorous and efficient framework to CER. We propose a formal language for specifying complex events, calledcomplex event logic(CEL), that contains the main features used in the literature and has a denotational and compositional semantics. We also formalize the so-called selection strategies, which had only been presented as by-design extensions to existing frameworks. We give insight into the language design trade-offs regarding the strict sequencing operators of CEL and selection strategies.With a well-defined semantics at hand, we discuss how to efficiently process complex events by evaluating CEL formulas with unary filters. We start by introducing a formal computational model for CER, calledcomplex event automata(CEA), and study how to compile CEL formulas with unary filters into CEA. Furthermore, we provide efficient algorithms for evaluating CEA over event streams using constant time per event followed by output-linear delay enumeration of the results.

Retraction Note: Non-point source pollution in river basin based on Bayesian network and intelligent translation system of English books

With the rapid development of social science and industrial technology, more and more attention has been paid to the evaluation of water environmental pollution. Bayesian network has strong probability theory inference ability, clear semantics, easy to understand, and other technical characteristics. This provides a natural and effective way to express dependency and causality, which can find potential relationships and patterns in datasets. Therefore, it has its own advantages in data mining. Bayesian network technology is used to study non-point source pollution in river basin and intelligent translation system of English books. Water pollution is one of the main problems that human beings are facing. In China, water pollution is more serious. First of all, the acceleration of urbanization and the change of land use will lead to the change of surface vegetation, resulting in the loss of soil and water. Secondly, many fertilizers, pesticides, and other chemical reagents are used in agricultural activities, some of which are absorbed by crops, and most of the rest are released from the soil in the process of rainfall. “Machine translation + post-translation editing” mode has been widely used in various translation projects, and the English Chinese subtitle translation project of “a little English” English learning software has adopted this mode. In practice, it is found that “machine translation + mode can improve the efficiency of subtitle E-C translation, but there are also limitations. In this paper, by analyzing the cases in the subtitle English-Chinese translation project of the English learning software, the advantages and typical errors of machine translation in subtitle English-Chinese translation are summarized, and the corresponding post-translation editing strategies are proposed for different error types. The main errors at the lexical level are the wrong choice of word meaning, no translation of abbreviations, and no translation of coined words. The proposed post-editing strategies are to locate polysemous words.

SBGN Bricks Ontology as a tool to describe recurring concepts in molecular networks.

A comprehensible representation of a molecular network is key to communicating and understanding scientific results in systems biology. The Systems Biology Graphical Notation (SBGN) has emerged as the main standard to represent such networks graphically. It has been implemented by different software tools, and is now largely used to communicate maps in scientific publications. However, learning the standard, and using it to build large maps, can be tedious. Moreover, SBGN maps are not grounded on a formal semantic layer and therefore do not enable formal analysis. Here, we introduce a new set of patterns representing recurring concepts encountered in molecular networks, called SBGN bricks. The bricks are structured in a new ontology, the Bricks Ontology (BKO), to define clear semantics for each of the biological concepts they represent. We show the usefulness of the bricks and BKO for both the template-based construction and the semantic annotation of molecular networks. The SBGN bricks and BKO can be freely explored and downloaded at sbgnbricks.org.

TS-Extractor: large graph exploration via subgraph extraction based on topological and semantic information.

Exploring large graphs is difficult due to their large size and semantic information such as node attributes. Extracting only a subgraph relevant to the user-specified nodes (called focus nodes) is an effective strategy for exploring a large graph. However, existing approaches following this strategy mainly focus on graph topology and do not fully consider node attributes, resulting in the lack of clear semantics in the extracted subgraphs. In this paper, we propose a novel approach called TS-Extractor that can extract a relevant subgraph around the user-selected focus nodes to help the user explore the large graph from a local perspective. By combining the graph topology and the user-selected node attributes, TS-Extractor can extract and visualize a connected subgraph that contains as many nodes sharing the same/similar attribute values with the focus nodes as possible, thereby providing the user with clear semantics. Based on TS-Extractor, we develop a Web-based graph exploration system that allows users to interactively extract, analyze and expand subgraphs. Through two case studies and a user study, we demonstrate the usability and effectiveness of TS-Extractor.Graphic abstract

NATIVE LAYER OF AVAR PERSONAL NAMES

An attempt is made to show that in the course of historical development and under the influence of other anthroponymic systems, the bulk of the original names of the Avars were lost. A small part of them that have been preserved shows that this category of names was mainly formed by suffixes added to the name, to the verb base, to the adverb, or by the word structure, and not all of them have a clear semantics. The socio-historical, ethnographic and cultural state of a people is recorded in anthroponymy only indirectly and is held by tradition for many hundreds of years after the phenomena that brought them to life have already been lost. Thus, models of personal names with the term kinship, formed probably in the bowels of the tribal system, are kept in the Avar code names in our time.

UX-Painter: An Approach to Explore Interaction Fixes in the Browser

Usability and user interaction improvement is a central task in web development to guarantee the success of a web application. However, designers are barely able to keep up with the current development cycle because their practices are too costly, while interaction issues accumulate in applications that end-users keep suffering. In this work, we propose a method for designers to rapidly explore solutions through visual programming to the interaction problems of an application under development, even when it has been already deployed. The method is realized by a tool called UX-Painter, an exploratory tool for designers to apply quick fixes to interaction issues at the client-side of a web application without the need of any script programming knowledge. The palette of available fixes in UX-Painter are client-side web refactorings, i.e., changes to web page elements that solve specific user interaction problems without changing the underlying functionality. UX-Painter allows designers to quickly set up new versions of a web application by combining refactorings to create alternative designs for user testing or an inspection review. UX-Painter also provides the means to communicate design improvements, as a sequence of refactorings with clear semantics. We show the feedback provided by interviews with designers about UX-Painter's functionality and the results of a user test about its usability.

Enhanced semantic representation of coaxiality with double material requirements

SummaryIn order to improve the assemblability and reduce the cost of the parts, the material requirements can be applied to the toleranced feature and the datum feature of the coaxiality at the same time in the practical application. How to realize the information exchange between different heterogeneous systems and how to read and explain the semantics of this complex tolerance specification automatically is one of the urgent problems to be solved in intelligent digital manufacturing. Ontology described in OWL2 DL and SWRL language has strict mathematical basis of description logic, and the knowledge represented by it has clear semantics and can be read and understood directly by computer. In addition, because of its consistency checking, knowledge reasoning, and flexible query ability, it becomes the most potential knowledge representation technology in the future. On the other hand, the semantic interpretation of tolerance specification by the new generation GPS theory based on skin model has the advantages of clear semantics, no ambiguity, and consistency at different stages of the product life cycle. In this article, the enhanced semantic representation ontology for this kind of coaxiality tolerance specification is constructed based on the advantages of the new GPS theory and ontology technology and the effectiveness of the method is verified and analyzed through a case.

Vector Generation of Reflection and Shearing of 2D Objects: Sequencing in Column Major and Row Major Pattern

Transformations are the best and effective techniques of producing changed objects in terms of Displacement, Enlargements and Orientations. In other words the Transformation are the most challenging way of shifting or changing the dimensions and orientations of images in the most effective way.But This modifying and repositikoning of the existing images is calculated by a standard convention of Matrix Calculations.The usual practice of doing so is straight forward. The transformed object can be obtained by coupling (Matrix Multiplication) original object Matrix with the transformation vectors. The main challenge is how to evaluate it. The usual practice is standard Column Major Pattern. The alternative Row Major Pattern is also known approach but what matters is the sequence of operations that make these both approaches worth mentioning.Visualization can be enhanced if either of the approach is adopted. But what makes it more exciting is the calculations that come with both of these approaches. A normal practice is to use the standard Column Major Pattern for processing objects .An Alternative to this Technique is Row Major Pattern which equates to the same result only if sequence of operations are not compromised. This paper describes the transformation (Reflection and Shearing) in both Column and Row Major Pattern and at the same time aims in putting down clear semantics in justifying the sequence of operations.

In-Air Gesture Interaction: Real Time Hand Posture Recognition Using Passive RFID Tags

In-air gesture interaction enables a natural communication between a man and a machine with its clear semantics and humane mode of operation. In this paper, we propose a real-time recognition system on multiple gestures in the air. It uses the commodity off-the-shelf (COTS) reader with three antennas to detect the radio frequency (RF) signals of the passive radio frequency identification (RFID) Tags attached to the fingers. The idea derives from the crucial insight that the sequential phase profile of the backscatter RF signals is a reliable and well-regulated indicator insinuating space-time situation of the tagged object, which presents a close interdependency with tag's movements and positions. The KL divergence is utilized to extract the dynamic gesture segment by confirming the endpoints of the data flow. To achieve the template matching and classification, we bring in the dynamic time warping (DTW) and k-nearest neighbors (KNN) for similarity scores calculation and appropriate gesture recognition. The experiment results show that the recognition rates for static and dynamic gestures can reach 85% and 90%, respectively. Moreover, it can maintain satisfying performance under different situations, such as diverse antenna-to-user distances and being hidden from view by nonconducting obstacles.

Crisis and Disaster Situations on Social Media Streams: An Ontology-Based Knowledge Harvesting Approach

Aim/Purpose: Vis-à-vis management of crisis and disaster situations, this paper focuses on important use cases of social media functions, such as information collection & dissemination, disaster event identification & monitoring, collaborative problem-solving mechanism, and decision-making process. With the prolific utilization of disaster-based ontological framework, a strong disambiguation system is realized, which further enhances the searching capabilities of the user request and provides a solution of unambiguous in nature. Background: Even though social media is information-rich, it has created a challenge for deriving a decision in critical crisis-related cases. In order to make the whole process effective and avail quality decision making, sufficiently clear semantics of such information is necessary, which can be supplemented through employing semantic web technologies. Methodology: This paper evolves a disaster ontology-based system availing a framework model for monitoring uses of social media during risk and crisis-related events. The proposed system monitors a discussion thread discovering whether it has reached its peak or decline after its root in the social forum like Twitter. The content in social media can be accessed through two typical ways: Search Application Program Interfaces (APIs) and Streaming APIs. These two kinds of API processes can be used interchangeably. News content may be filtered by time, geographical region, keyword occurrence and availability ratio. With the support of disaster ontology, domain knowledge extraction and comparison against all possible concepts are availed. Besides, the proposed method makes use of SPARQL to disambiguate the query and yield the results which produce high precision. Contribution: The model provides for the collection of crisis-related temporal data and decision making through semantic mapping of entities over concepts in a disaster ontology we developed, thereby disambiguating potential named entities. Results of empirical testing and analysis indicate that the proposed model outperforms similar other models. Findings: Crucial findings of this research lie in three aspects: (1) Twitter streams and conventional news media tend to offer almost similar types of news coverage for a specified event, but the rate of distribution among topics/categories differs. (2) On specific events such as disaster, crisis or any emergency situations, the volume of information that has been accumulated between the two news media stands divergent and filtering the most potential information poses a challenging task. (3) Relational mapping/co-occurrence of terms has been well designed for conventional news media, but due to shortness and sparseness of tweets, there remains a bottleneck for researchers. Recommendations for Practitioners: Though metadata avails collaborative details of news content and it has been conventionally used in many areas like information retrieval, natural language processing, and pattern recognition, there is still a lack of fulfillment in semantic aspects of data. Hence, the pervasive use of ontology is highly suggested that build semantic-oriented metadata for concept-based modeling, information flow searching and knowledge exchange. Recommendation for Researchers: The strong recommendation for researchers is that instead of heavily relying on conventional Information Retrieval (IR) systems, one can focus more on ontology for improving the accuracy rate and thereby reducing ambiguous terms persisting in the result sets. In order to harness the potential information to derive the hidden facts, this research recommends clustering the information from diverse sources rather than pruning a single news source. It is advisable to use a domain ontology to segregate the entities which pose ambiguity over other candidate sets thus strengthening the outcome. Impact on Society: The objective of this research is to provide informative summarization of happenings such as crisis, disaster, emergency and havoc-based situations in the real world. A system is proposed which provides the summarized views of such happenings and corroborates the news by interrelating with one another. Its major task is to monitor the events which are very booming and deemed important from a crowd’s perspective. Future Research: In the future, one shall strive to help to summarize and to visualize the potential information which is ranked high by the model.

FUZZY SUPERVISOR FOR PID CONTROLLER

Introduction: PID regulators are an important industrial automation tool. However, the traditional ways of their customization involve experiments with the plant, reducing the control system efficiency when the operating conditions change. To solve this problem, you have to use an adaptation loop which automatically changes the regulator parameters when the quality indicators deteriorate. Purpose: Developing a structure and an algorithm for a fuzzy supervisor of a PID controller in order to improve the operation quality in the context of uncontrolled changes in the plant parameters. Results: We propose two options for organizing a fuzzy supervisor. The first option involves continuous changing of the controller parameters based on the information about the current control error. The supervisor is trained offline with the help of a genetic algorithm. This approach makes the application universal, but the resulting structure is a black box. The second option is using transient process quality estimates; it can be used when the input signal periodically changes. The controller parameters are changed using fuzzy rules with clear semantics. It is shown that systems with a fuzzy supervisor can reduce the overshoot or static error which both can occur when the control object parameters change. The system operation was simulated using MatLab Simulink. Practical relevance: The use of PID controllers with fuzzy supervisors can be useful in the design of control systems for a wide range of dynamic plants.

Towards a semantics for metanormative constructivism

The status of constructivism as a metaethical or metanormative theory is unclear partly due to the lack of a clear semantics for central normative terms such as ‘reason’ and ‘ought’. In a series of recent papers, Sharon Street has attempted to clarify the central commitments of constructivism by focusing on the idea of a practical point of view and what follows from it. We improve upon the informal understanding provided by Street and attempt to provide a semantics for ‘ought’. Our semantics respects the core intuition of the constructivist that normative claims are made true because of our practical commitments as agents and also reflects the constructivist’s commitment to the centrality of practical deliberation to normative truth. On our view, a normative claim of the form ⌜S ought to ϕ⌝ is true if ϕ is entailed from S’s set of evaluative attitudes. We argue that a virtue of our definition is that it allows us to see precisely what is distinctive about constructivism as opposed to realism and expressivism.