Search

Gerhard Richter's books Text - a collection of painter's verbal statements about his artistic method - and Atlas - 783 sheets with images, mainly photographs and visual notations - are two archives that complement the understanding of his diverse artistic practice. The paper presents a textual model that experimentally simulates a possible ordering principle for archives. Richter's statements in the book Text are cut up and used as short quotations. Those that relate to multiple aspects of the painter's oeuvre are identified as hubs in the semantic network. The hubs are organized paratactically, as an array of different themes. The paper presents a methodological hypothesis and an experimental model that aim to connect the research of real networks with the paradigms of humanistic interpretation. We have to bear in mind that the network is a result of the researcher's interpretative approach, which is added to the initial archive included in the book Text. The breaking up of Richter's poetics into atoms of quotations is an experimental proposal of a new textuality in art history and humanities, which has its own history. In comparison to digital archives with complex interfaces that often tend to obscure the content, the elements in our experiment appear as specific configurations of the semantic network and are presented in a limited number of linear texts. The method of listing of quotations gathers the fragments into a potential "whole", i.e. a narrativized gateway to an archive according to the researcher's interpretation. [ABSTRACT FROM AUTHOR]

The next generation of Internet services is driven by users and user-generated content. The complex nature of user behavior makes it highly challenging to manage and secure online services. On one hand, service providers cannot effectively prevent attackers from creating large numbers of fake identities to disseminate unwanted content (e.g., spam). On the other hand, abusive behavior from real users also poses significant threats (e.g., cyberbullying). In this article, we propose clickstream models to characterize user behavior in large online services. By analyzing clickstream traces (i.e., sequences of click events from users), we seek to achieve two goals: (1) detection: to capture distinct user groups for the detection ofmalicious accounts, and (2) understanding: to extract semantic information from user groups to understand the captured behavior. To achieve these goals, we build two related systems. The first one is a semisupervised system to detect malicious user accounts (Sybils). The core idea is to build a clickstream similarity graph where each node is a user and an edge captures the similarity of two users' clickstreams. Based on this graph, we propose a coloring scheme to identify groups of malicious accounts without relying on a large labeled dataset. We validate the system using groundtruth clickstream traces of 16,000 real and Sybil users from Renren, a large Chinese social network. The second system is an unsupervised system that aims to capture and understand the fine-grained user behavior. Instead of binary classification (malicious or benign), this model identifies the natural groups of user behavior and automatically extracts features to interpret their semantic meanings. Applying this system to Renren and another online social network,Whisper (100K users), we help service providers identify unexpected user behaviors and even predict users' future actions. Both systems received positive feedback from our industrial collaborators including Renren, LinkedIn, and Whisper after testing on their internal clickstream data. [ABSTRACT FROM AUTHOR]

Background: Biological databases store data about laboratory experiments, together with semantic annotations, in order to support data aggregation and retrieval. The exact meaning of such annotations in the context of a database record is often ambiguous. We address this problem by grounding implicit and explicit database content in a formal-ontological framework. Methods: By using a typical extract from the databases UniProt and Ensembl, annotated with content from GO, PR, ChEBI and NCBI Taxonomy, we created four ontological models (in OWL), which generate explicit, distinct interpretations under the BioTopLite2 (BTL2) upper-level ontology. The first three models interpret database entries as individuals (IND), defined classes (SUBC), and classes with dispositions (DISP), respectively; the fourth model (HYBR) is a combination of SUBC and DISP. For the evaluation of these four models, we consider (i) database content retrieval, using ontologies as query vocabulary; (ii) information completeness; and, (iii) DL complexity and decidability. The models were tested under these criteria against four competency questions (CQs). Results: IND does not raise any ontological claim, besides asserting the existence of sample individuals and relations among them. Modelling patterns have to be created for each type of annotation referent. SUBC is interpreted regarding maximally fine-grained defined subclasses under the classes referred to by the data. DISP attempts to extract truly ontological statements from the database records, claiming the existence of dispositions. HYBR is a hybrid of SUBC and DISP and is more parsimonious regarding expressiveness and query answering complexity. For each of the four models, the four CQs were submitted as DL queries. This shows the ability to retrieve individuals with IND, and classes in SUBC and HYBR. DISP does not retrieve anything because the axioms with disposition are embedded in General Class Inclusion (GCI) statements. Conclusion: Ambiguity of biological database content is addressed by a method that identifies implicit knowledge behind semantic annotations in biological databases and grounds it in an expressive upper-level ontology. The result is a seamless representation of database structure, content and annotations as OWL models. [ABSTRACT FROM AUTHOR]

Experimental evaluation carried out in international large-scale campaigns is a fundamental pillar of the scientific and technological advancement of information retrieval ( IR) systems. Such evaluation activities produce a large quantity of scientific and experimental data, which are the foundation for all the subsequent scientific production and development of new systems. In this work, we discuss how to semantically annotate and interlink this data, with the goal of enhancing their interpretation, sharing, and reuse. We discuss the underlying evaluation workflow and propose a resource description framework model for those workflow parts. We use expertise retrieval as a case study to demonstrate the benefits of our semantic representation approach. We employ this model as a means for exposing experimental data as linked open data ( LOD) on the Web and as a basis for enriching and automatically connecting this data with expertise topics and expert profiles. In this context, a topic-centric approach for expert search is proposed, addressing the extraction of expertise topics, their semantic grounding with the LOD cloud, and their connection to IR experimental data. Several methods for expert profiling and expert finding are analysed and evaluated. Our results show that it is possible to construct expert profiles starting from automatically extracted expertise topics and that topic-centric approaches outperform state-of-the-art language modelling approaches for expert finding. [ABSTRACT FROM AUTHOR]

Most of current person re-identification works identify a person by matching his/her probe image against a galley images set. One feasible way to improve the identification accuracy is the multi-shot re-identification, where the probe includes a small set of images rather than a single image. In this paper, we study the tracklet-to-tracklet identification, where both the probe and the target dataset are composed of small sets of sequential images, i.e., tracklets. To solve this problem and make our algorithm robust under multi-camera setting, we take full advantage of low-level features, attributes and inter-attribute correlations at the same time. Attributes are expected to offer semantic information complementary to low-level features. In order to discover the correlations among attributes, a novel discriminative model is proposed to exploit low-level features and map attributes to a discriminative latent prototypes space. An alternating optimization procedure is designed to perform the learning process. We also devise a number of voting schemes to total up matching scores from images to tracklets. Experiments on four public datasets show that our approach achieves a consistently better performance than existing person re-identification methods. [ABSTRACT FROM AUTHOR]

We introduce a novel measure of abstractness based on the amount of information of a concept computed from its position in a semantic taxonomy. We refer to this measure as precision. We propose two alternative ways to measure precision, one based on the path length from a concept to the root of the taxonomic tree, and another one based on the number of direct and indirect descendants. Since more information implies greater processing load, we hypothesize that nouns higher in precision will have a processing disadvantage in a lexical decision task. We contrast precision to concreteness, a common measure of abstractness based on the proportion of sensory-based information associated with a concept. Since concreteness facilitates cognitive processing, we predict that while both concreteness and precision are measures of abstractness, they will have opposite effects on performance. In two studies we found empirical support for our hypothesis. Precision and concreteness had opposite effects on latency and accuracy in a lexical decision task, and these opposite effects were observable while controlling for word length, word frequency, affective content and semantic diversity. Our results support the view that concepts organization includes amodal semantic structures which are independent of sensory information. They also suggest that we should distinguish between sensory-based and amount-of-information-based abstractness. [ABSTRACT FROM AUTHOR]

Neuronal morphology is extremely diverse across and within animal species, developmental stages, brain regions, and cell types. This diversity is functionally important because neuronal structure strongly affects synaptic integration, spiking dynamics, and network connectivity. Digital reconstructions of axonal and dendritic arbors are thus essential to quantify and model information processing in the nervous system. NeuroMorpho.Org is an established repository containing tens of thousands of digitally reconstructed neurons shared by several hundred laboratories worldwide. Each neuron is annotated with specific metadata based on the published references and additional details provided by data owners. The number of represented metadata concepts has grown over the years in parallel with the increase of available data. Until now, however, the lack of standardized terminologies and of an adequately structured metadata schema limited the effectiveness of user searches. Here we present a new organization of NeuroMorpho.Org metadata grounded on a set of interconnected hierarchies focusing on the main dimensions of animal species, anatomical regions, and cell types. We have comprehensively mapped each metadata term in NeuroMorpho.Org to this formal ontology, explicitly resolving all ambiguities caused by synonymy and homonymy. Leveraging this consistent framework, we introduce OntoSearch, a powerful functionality that seamlessly enables retrieval of morphological data based on expert knowledge and logical inferences through an intuitive string-based user interface with auto-complete capability. In addition to returning the data directly matching the search criteria, OntoSearch also identifies a pool of possible hits by taking into consideration incomplete metadata annotation. [ABSTRACT FROM AUTHOR]

A long-standing but implicit assumption is that words strongly associated with a presented cue are automatically activated in the memory through rapid spread of activation within brain semantic networks. The current study was aimed to provide direct evidence of such rapid access to words' semantic representations and to investigate its neural sources using magnetoencephalography (MEG) and distributed source localization technique. Thirty-three neurotypical subjects underwent the MEG recording during verb generation task, which was to produce verbs related to the presented noun cues. Brain responses evoked by the noun cues were examined while manipulating the strength of association between the noun and the potential verb responses. The strong vs. weak noun-verb association led to a greater noun-related neural response at 250-400 ms after cue onset, and faster verb production. The cortical sources of the differential response were localized in left temporal pole, previously implicated in semantic access, and left ventrolateral prefrontal cortex (VLPFC), thought to subserve controlled semantic retrieval. The strength of the left VLPFC's response to the nouns with strong verb associates was positively correlated to the speed of verbs production. Our findings empirically validate the theoretical expectation that in case of a strongly connected noun-verb pair, successful access to target verb representation may occur already at the stage of lexico-semantic analysis of the presented noun. Moreover, the MEG results suggest that contrary to the previous conclusion derived from fMRI studies left VLPFC supports selection of the target verb representations, even if they were retrieved from semantic memory rapidly and effortlessly. The discordance between MEG and fMRI findings in verb generation task may stem from different modes of neural activation captured by phase-locked activity in MEG and slow changes of blood-oxygen-level-dependent (BOLD) signal in fMRI. [ABSTRACT FROM AUTHOR]

The enormous exploitation of connected objects for patient monitoring becomes a source of a huge quantity of health data. This data can be badly expressed, understood, and exploited by other systems and devices. Additionally, the huge amount of connected objects embeds with resource-constrained devices (sensors, actuators, RFIDS, etc.) are also challenging. Consequently, semantic representation of both medical connected objects and their data becomes important. Furthermore, the semantic reasoning of this data requires the definition of diverse rules in order to provide adequate and efficient results. This paper suggests two main contributions. The first one focuses on the semantic representation of both the medical connected objects and their data by proposing a HealthIoT ontology. The second contribution is proposed to provide practical backup to the use of this ontology. In fact, an IoT Medicare system is proposed to evaluate the HealthIoT ontology. [ABSTRACT FROM AUTHOR]

Information retrieval technologies behind web search engines in the field of computer science were bought up in the year of 1950s. It is a process of retrieving the relevant documents based on the queries raised by the user. It deals with the representation, storage and access of information items. In this system, the generated outputs are ranked according to their relevance. The information retrieval (IR) uses data models that make a retrieval process easier when compared to the traditional IR database model. In this work, we analyse the most popular information retrieval models such as boolean, vector space, probabilistic and latent semantic analysis and evaluate the performance of the models by using the underlying parameters like concept, representation, word occurrence, information type, output, pros and cons of the models. This study aims to determine the appropriate model for different situations and additionally describes the indexing methods for decrementing search space and different probing (searching) techniques to retrieve the information. [ABSTRACT FROM AUTHOR]

This article reports on an empirical study examining English speakers’ L2 processing of Chinese base-generated-topic (BGT) sentences. Forty-four highly proficient English-speaking L2 learners of Chinese and 23 native Chinese speakers were involved in the study. Results of a self-paced reading task reveal that both native Chinese speakers’ and L2 Chinese learners’ processing of Chinese BGT sentences is syntactically induced in a top-down manner. English speakers are sensitive to and are able to make use of syntactic cues as well as semantic information in their processing of Chinese BGT sentences. The study provides disconfirming evidence against the Shallow Structure Hypothesis ( Clahsen and Felser, 2006a,b ), which predicts that unlike native speakers, L2 learners do not rely on structure-based processing strategies when solving ambiguities in L2 sentence processing. [ABSTRACT FROM AUTHOR]

Reference to space, especially with prepositions, plays a central role in natural language and is receiving more and more attention over the past decades. One line of research uses formal semantic modeling, using topological and other geometrical concepts such as regions, vectors, and paths. Another line of research has drawn attention to the role of function, force-dynamics, polysemy, prototypes, and crosslinguistic variation in this domain. This paper gives an overview of both lines of research and argues that a synthesis is possible, based on a proper division of labor between semantics and pragmatics, richer ontologies, and a perspective on categorization that uses conceptual spaces and semantic maps. [ABSTRACT FROM AUTHOR]

Social media is a rich source of up-to-date information about events such as incidents. The sheer amount of available information makes machine learning approaches a necessity to process this information further. This learning problem is often concerned with regionally restricted datasets such as data from only one city. Because social media data such as tweets varies considerably across different cities, the training of efficient models requires labeling data from each city of interest, which is costly and time consuming. To avoid such an expensive labeling procedure, a generalizable model can be trained on data from one city and then applied to data from different cities. In this paper, we present Semantic Abstraction to improve the generalization of tweet classification. In particular, we derive features from Linked Open Data and include location and temporal mentions. A comprehensive evaluation on twenty datasets from ten different cities shows that Semantic Abstraction is indeed a valuable means for improving generalization. We show that this not only holds for a two-class problem where incident-related tweets are separated from non-related ones but also for a four-class problem where three different incident types and a neutral class are distinguished. To get a thorough understanding of the generalization problem itself, we closely examined rule-based models from our evaluation. We conclude that on the one hand, the quality of the model strongly depends on the class distribution. On the other hand, the rules learned on cities with an equal class distribution are in most cases much more intuitive than those induced from skewed distributions. We also found that most of the learned rules rely on the novel semantically abstracted features. [ABSTRACT FROM AUTHOR]

In this paper, we extend some usual techniques of classification resulting from a large-scale data-mining and network approach. This new technology, which in particular is designed to be suitable to big data, is used to construct an open consolidated database from raw data on 4 million patents taken from the US patent office from 1976 onward. To build the pattern network, not only do we look at each patent title, but we also examine their full abstract and extract the relevant keywords accordingly. We refer to this classification as semantic approach in contrast with the more common technological approach which consists in taking the topology when considering US Patent office technological classes. Moreover, we document that both approaches have highly different topological measures and strong statistical evidence that they feature a different model. This suggests that our method is a useful tool to extract endogenous information. [ABSTRACT FROM AUTHOR]

Here, we present a study on how the structure of semantic networks based on cliques (specifically, article titles) behaves when vertex removal strategies (i.e., random and uniform vertex removal — RUR, highest degree vertex removal — HDR, and highest intermediation centrality vertex removal — HICR) are applied to this type of network. We propose a method for calculation of the average size of the small components and we identify the existence of a fraction ( f p ) where the topological structure of the network changes. Semantic networks based on cliques maintain the small-world phenomenon when subjected to RUR, HDR and HICR for fractions of removed vertices less than or equal to f p . [ABSTRACT FROM AUTHOR]

Geographic Object-Based Image Analysis (GEOBIA) techniques have become increasingly popular in remote sensing. GEOBIA has been claimed to represent a paradigm shift in remote sensing interpretation. Still, GEOBIA--similar to other emerging paradigms--lacks formal expressions and objective modelling structures and in particular semantic classification methods using ontologies. This study has put forward an object-based semantic classification method for high resolution satellite imagery using an ontology that aims to fully exploit the advantages of ontology to GEOBIA. A three-step workflow has been introduced: ontology modelling, initial classification based on a data-driven machine learning method, and semantic classification based on knowledge-driven semantic rules. The classification part is based on data-driven machine learning, segmentation, feature selection, sample collection and an initial classification. Then, image objects are re-classified based on the ontological model whereby the semantic relations are expressed in the formal languages OWL and SWRL. The results show that the method with ontology--as compared to the decision tree classification without using the ontology--yielded minor statistical improvements in terms of accuracy for this particular image. However, this framework enhances existing GEOBIA methodologies: ontologies express and organize the whole structure of GEOBIA and allow establishing relations, particularly spatially explicit relations between objects as well as multi-scale/hierarchical relations. [ABSTRACT FROM AUTHOR]

We present a denotational semantics designed in continuation-passing style (CPS) for an abstract language providing operators for nondeterministic choice, sequential and parallel composition, and a general mechanism of interaction between multisets of distributed actions. We show that the basic laws of concurrent systems are satisfied in this semantics. Next, by customizing the behavior of continuations we obtain denotational semantics for a couple of concurrent languages and a nature-inspired formalism. The languages discussed include Hoare's communicating sequential processes (CSP), and two formalisms based on multiparty interactions: a version of CSP extended with communication and synchronization on multiple channels, and a language similar to a process algebra for DNA computing introduced by Cardelli. We accomplish the semantic investigation in the mathematical framework of complete metric spaces. [ABSTRACT FROM AUTHOR]

Polarizationless P systems with active membranes are non-cooperative systems, that is, the left-hand side of their rules have a single object. Usually, these systems make use of division rules as a mechanism to produce an exponential workspace in linear time. Division rules are inspired by cell division, a process of nuclear division that occurs when a parent cell divides to produce two identical daughter cells. On the other hand, separation rules are inspired by the membrane fission process, a mechanism by which a biological membrane is split into two new ones in such a manner that the contents of the initial membrane is distributed between the new membranes. In this paper, separation rules are used instead of division rules. The computational efficiency of these models is studied and the role of the (minimal) cooperation in object evolution rules is explored from a computational complexity point of view. [ABSTRACT FROM AUTHOR]

Attack trees profile the sub-goals of the proponent of an attack. Attack trees have a variety of semantics depending on the kind of question posed about the attack, where questions are captured by an attribute domain. We observe that one of the most general semantics for attack trees, the multiset semantics, coincides with a semantics expressed using linear logic propositions. The semantics can be used to compare attack trees to determine whether one attack tree is a specialisation of another attack tree. Building on these observations, we propose two new semantics for an extension of attack trees named causal attack trees. Such attack trees are extended with an operator capturing the causal order of sub-goals in an attack. These two semantics extend the multiset semantics to sets of series-parallel graphs closed under certain graph homomorphisms, where each semantics respects a class of attribute domains. We define a sound logical system with respect to each of these semantics, by using a recently introduced extension of linear logic, called MAV, featuring a non-commutative operator. The non-commutative operator models causal dependencies in causal attack trees. Similarly to linear logic for attack trees, implication defines a decidable preorder for specialising causal attack trees that soundly respects a class of attribute domains. [ABSTRACT FROM AUTHOR]

The Abstract Conceptual Feature ( ACF) framework predicts that word meaning is represented within a high-dimensional semantic space bounded by weighted contributions of perceptual, affective, and encyclopedic information. The ACF, like latent semantic analysis, is amenable to distance metrics between any two words. We applied predictions of the ACF framework to abstract words using eyetracking via an adaptation of the classical 'visual word paradigm' (VWP). Healthy adults ( n = 20) selected the lexical item most related to a probe word in a 4-item written word array comprising the target and three distractors. The relation between the probe and each of the four words was determined using the semantic distance metrics derived from ACF ratings. Eye movement data indicated that the word that was most semantically related to the probe received more and longer fixations relative to distractors. Importantly, in sets where participants did not provide an overt behavioral response, the fixation rates were nonetheless significantly higher for targets than distractors, closely resembling trials where an expected response was given. Furthermore, ACF ratings which are based on individual words predicted eye fixation metrics of probe-target similarity at least as well as latent semantic analysis ratings which are based on word co-occurrence. The results provide further validation of Euclidean distance metrics derived from ACF ratings as a measure of one facet of the semantic relatedness of abstract words and suggest that they represent a reasonable approximation of the organization of abstract conceptual space. The data are also compatible with the broad notion that multiple sources of information (not restricted to sensorimotor and emotion information) shape the organization of abstract concepts. While the adapted 'VWP' is potentially a more metacognitive task than the classical visual world paradigm, we argue that it offers potential utility for studying abstract word comprehension. [ABSTRACT FROM AUTHOR]

Introduction Interest in the function of the inferior parietal lobule (IPL) has resulted in increased understanding of its involvement in visuospatial and cognitive functioning, and its role in semantic networks. A basic understanding of the nuanced white-matter anatomy in this region may be useful in improving outcomes when operating in this region of the brain. We sought to derive the surgical relationship between the IPL and underlying major white-matter bundles by characterizing macroscopic connectivity. Methods Data of 10 healthy adult controls from the Human Connectome Project were used for tractography analysis. All IPL connections were mapped in both hemispheres, and distances were recorded between cortical landmarks and major tracts. Ten postmortem dissections were then performed using a modified Klingler technique to serve as ground truth. Results We identified three major types of connections of the IPL. (1) Short association fibers connect the supramarginal and angular gyri, and connect both of these gyri to the superior parietal lobule. (2) Fiber bundles from the IPL connect to the frontal lobe by joining the superior longitudinal fasciculus near the termination of the Sylvian fissure. (3) Fiber bundles from the IPL connect to the temporal lobe by joining the middle longitudinal fasciculus just inferior to the margin of the superior temporal sulcus. Conclusions We present a summary of the relevant anatomy of the IPL as part of a larger effort to understand the anatomic connections of related networks. This study highlights the principle white-matter pathways and highlights key underlying connections. [ABSTRACT FROM AUTHOR]

Sleep is known to affect learning and memory, but the extent to which it influences behavioural processing in the left and right hemispheres of the brain is as yet unknown. We tested two hypotheses about lateralised effects of sleep on recognition memory for words: whether sleep reactivated recent experiences of words promoting access to the long-term store in the left hemisphere (LH), and whether sleep enhanced spreading activation differentially in semantic networks in the hemispheres. In Experiment 1, participants viewed lists of semantically related words, then slept or stayed awake for 12h before being tested on seen, unseen but related, or unrelated words presented to the left or the right hemisphere. Sleep was found to promote word recognition in the LH, and to spread activation equally within semantic networks in both hemispheres. Experiment 2 ensured that the results were not due to time of day effects influencing cognitive performance. [ABSTRACT FROM AUTHOR]

The purpose of this study is to investigate Iran’s eLearning research status in comparison with the world. We propose a method based on a text mining approach for extracting knowledge from Iranian published articles and generating the corresponding semantic network automatically. eLearning concepts are extracted from papers published in 6 years’ proceedings of ICeLeT, an International Conference on eLearning and eTeaching, in Iran. After extracting the domain-specific concepts, each pair of concepts get the possibility to be linked together based on co-occurrence in the articles. A weight is assigned to each edge according to the pointwise mutual information value of the pair of concepts. To identify gaps between the latest local and global research, the obtained semantic network is compared with another semantic network extracted from 6 years’ proceedings of ICALT, an International Conference on Advanced Learning Technologies. By applying a hybrid clustering algorithm on two networks based on the combination of label propagation and Markov clustering, and identifying the differences between node memberships and hubs, strengths and weaknesses of each network are demonstrated. [ABSTRACT FROM PUBLISHER]

Ontology integration is an important work when several heterogeneous ontologies share or exchange their knowledge. It takes the two heterogeneous ontologies as input, and carries out three important steps: (1) creating mappings for the similar concepts from different ontologies based on the similarity calculation; (2) merging the mapped concepts into a new one, and generating a network-based model, called integrated model; (3) reconstructing the integrated model in order to obtain an integrated ontology (a tree-based model). However, the methods of merging concepts and generating integrated ontology have not been well discussed. To this end, this paper proposes an Ontology Integration Method based on Semantic Mapping, called OIM-SM. Firstly, OIM-SM calculates similarities for the concept-pairs, in which concepts are from different ontologies, and creates semantic equivalent mappings for the concept-pairs based on their similarities. For the mapped concepts, OIM-SM merges them to generate an integrated model. Then, the merged concepts are used to split the integrated model into blocks. Finally, OIM-SM re-aligns the concepts in the same block by analyzing the semantic relationships among them, in order to generate an integrated ontology. Some ontologies from the real world are used to test OIM-SM, and the results indicate that OIM-SM performs encouragingly well in the aspect of ontology integration. [ABSTRACT FROM AUTHOR]

A weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these ‘free’ tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel $L_1$ -optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the $L_1$ -optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy. [ABSTRACT FROM PUBLISHER]

The purpose of this study is to determine the psychological meaning of coauthorship for a group of scientists, based on the assumption that the meaning of a concept is related to experience on “how a person behaves in a situation, depending on what the situation signifies to him”. The semantic meaning provides for an interpretation of action in beliefs, goals and intentions, following the idea that semantic meaning is a basis for inferring intentions to perform action. We used the Natural Semantic Networks Method. We interviewed 162 scientists in the physical and natural sciences. Results show a cluster of the semantic meaning of coauthorship in which collaboration, teamwork, work, common interest, discussion and others, are retrieved from memory and represented as a natural semantic network of the concept of coauthorship. Collaboration and teamwork are the two concepts that have the highest semantic value and this holds true for various disciplines although teamwork is the most elicited word associated to coauthorship. We propose a model in which coauthorship represents the result of the experience, attitudes and behavior of a group process concerning collaboration and teamwork in science.ABSTRACT FROM PUBLISHERCopyright of Social Epistemology is the property of Routledge and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. [ABSTRACT FROM AUTHOR]

Truly universal helper robots capable of coping with unknown, unstructured environments must be capable of spatial reasoning, i.e. establishing geometric relations between objects and locations, expressing those in terms understandable by humans. It is therefore desirable that spatial and semantic environment representations are tightly interlinked. 3D robotic mapping and the generation of consistent metric representations of space are highly useful for navigation and exploration, but they do not capture symbol-level information about the environment. This is, however, essential for reasoning, and enables interaction via natural language, which is arguably the most common and natural communication channel used and understood by humans. This article presents a review of research in three major fields relevant for this discussion of spatial reasoning and interaction. Firstly, dialogue systems are an integral part of modern approaches to situated human–robot interaction. Secondly, interactive robots must be equipped with environment representations and reasoning methods that are suitable for both navigation and task fulfillment, as well as for interaction with human partners. Thirdly, at the interface between these domains are systems that ground language in systemic environment representation and which allow the integration of information from natural language descriptions into robotic maps. For each of these areas, important approaches are outlined and relations between the fields are highlighted, and challenging applications as well as open problems are discussed. [ABSTRACT FROM AUTHOR]

This paper proposes an image retrieval using fused deep convolutional features to solve the semantic gap between low-level features and high-level semantic features of traditional contend-based image retrieval method. Firstly, the improved network architecture LeNet-L is obtained by improving convolutional neural network LeNet-5. Then, fusing two different deep convolutional features which are extracted by LeNet-5 and AlexNet. Finally, after the fusion, the similar image is obtained through comparing the similarity between the image being retrieved and the image in database by distance function. In Corel dataset, this method is compared with the single convolutional neural network extracted features for image retrieval method, it has a higher precision and recall. The results show that this method has a better retrieval accuracy. [ABSTRACT FROM AUTHOR]

An approach to performing linguistic summaries of graph datasets, with particular focus on usage of ontologies is presented in this paper. This well-known mining technique is based on fuzzy set theory, which is used to model natural language words (e.g. 'many', 'tall'), and in result - generates natural-like sentences describing the data. Although intensely developed, before our work this method has been applied only to relational databases, while more and more data is available in graph model. A special case of such graph datasets is the Semantic Web, in which ontologies provide meaning, therefore enabling advanced machine learning. In our paper we analyze the problem of generating linguistic summaries for a graph data case (for which the method cannot be directly applied), with associated ontologies. The key element of ontologies are concept hierarchies, which are the core of our work. Firstly, due to heterogeneity and lack of schema we propose to use an ontological concept (including all sub-concepts in hierarchy) as a subject for summaries, and extract their attributes (neighboring vertexes). Then we show that by ascending these ontological concept hierarchies (so by attribute-based induction) we obtain additional, generalized summaries. We showthis process for both summarizers and qualifiers, and propose an extension to their respective imprecision measures - T2 and T9. We perform two experiments on DBPedia - one for summary subject 'Artist', and second for 'Musical Album'. For the latter, we show the optimized process of obtaining the truth values using bottom-up approach. [ABSTRACT FROM AUTHOR]

Cross modal (e.g., text-to-image or image-to-text) retrieval has received great attention with the flushed multi-modal social media data. It is of considerable challenge to stride across the heterogeneous gap between modalities. Existing methods project different modalities into a common space by minimizing the distance within the heterogeneous pairs (intra-pair) of the new latent space. However, the relationship among these multi-modal pairs (inter-pair) are neglected, which are beneficial to eliminate the heterogeneity. In this paper, we propose a novel algorithm based on canonical correlation analysis by considering the high-order relationship among pairs (HCCA) for cross-modal retrieval. Supervised with additional semantic labels and unsupervised without semantic labels are simultaneously considered by treating the intra- and inter-pair correlation discriminatively. Moreover, kernel tricks are also performed on HCCA to learn a non-linear projection, termed HKCCA. Extensive experiments conducted on three public datasets demonstrate the superiority of the proposed methods compared with the state-of-the-art approaches in cross modal retrieval. [ABSTRACT FROM AUTHOR]

A recurring theoretical focus in risk and safety research is the analysis of certain concepts of special interest (e.g. the concepts of risk, safety, and hazard) and how they are related. Together, related concepts form networks sometimes referred to as ‘semantic fields’. In this paper, the semantic field of risk is identified and analysed. Based on dictionaries, thesauri, and other lexicological resources, 244 words related to the concept of risk are identified. These words are classified into 25 specific categories based on their type of semantic relationship to the risk concept. Based on this semantic field analysis, three implications for risk and safety research are discussed: (i) the merit of an extensive and general overview of a conceptual field (previously subject to only partial and scattered treatment) for future analysis, (ii) methodological consequences for the study of social processes addressing risk and safety issues, and (iii) the opportunity for cross-cultural comparison. [ABSTRACT FROM AUTHOR]

Human exposure to and contamination by environmental toxic compounds generates discourses and practices that merit greater attention. In this article, we assess internal chemical contamination and the risk of toxic effects as an experience related to the production of meaning in everyday life. Drawing on the analysis of semantic networks of narratives from semi-structured interviews conducted with 43 informants in Catalonia, Spain, we consider participants’ perceptions of the health risks of toxic compounds, including social discourses on exposure, toxicity, and internal chemical contamination, and on responsibilities, consequences, and proposed strategies for controlling toxic compounds. Informants’ narratives on the relationships between nature and nurture suggest that they no longer perceive rigid boundaries separating the human body from the external environment and its chemical pollutants. [ABSTRACT FROM PUBLISHER]

Today's highly competitive business world requires that managers be able to make fast and accurate strategic decisions, as well as learn to adapt to new strategic challenges. This necessity calls for a deep experience and a dynamic understanding of strategic management. The trait of dynamic understanding is mainly the skill of generating additional knowledge and innovative solutions under the new environmental conditions. Building on the concepts of information processing, this paper aims to support managers in constructing new strategic management knowledge, through representing and mining existing knowledge through graph visualization. To this end, a three-stage framework is proposed and described. The framework can enable managers to develop a deeper understanding of the strategic management domain, and expand on existing knowledge through visual analysis. The model further supports a case study that involves unstructured knowledge of profit patterns and the related strategies to succeed using these patterns. The applicability of the framework is shown in the case study, where the unstructured knowledge in a strategic management book is first represented as a semantic network, and then visually mined for revealing new knowledge. [ABSTRACT FROM AUTHOR]

Technology opportunity analysis (TOA) has been the subject of many prior studies, most of which have focused on deconstructing and restructuring the original knowledge structure in a single domain. This study suggests a method by extending technology elements with BERT and TRIZ that endeavors to address these issues. First, patents collected from the Derwent database were used as data sources. Second, BERT was employed to construct a technology landscape as a vector space model where similar technology elements are classified into the same technology topic. Meanwhile, TEMPEST was employed to cluster technology topics and elements according to different functions and other dimensions. Third, technology elements were extended by function-oriented search (FOS), which is a useful method of TRIZ. It includes extracting new technology elements from newly retrieved patents about implementing a specific function in other domains. Fourth, technology opportunities were identified by recombining original and new technology elements and then verifying their feasibility. Finally, the proposed approach was employed in empirical analysis for unmanned ships and 10 technology opportunities generated through knowledge migration. The process designed in this study combines quantitative modeling and qualitative analysis, which realizes accurate search and efficient innovation among different domains. • TOA is completed by building a technology landscape across multiple technology domains. • FOS, a useful method of TRIZ is employed to search existing methods precisely to guide innovative activities. • BERT is introduced to cluster technology topics based on semantic relationships for enhance clustering effects. • The proposed framework is validated through the empirical analysis of unmanned ships. [ABSTRACT FROM AUTHOR]

With the emergence of COVID-19, the food-delivery market has grown and now employs large numbers of people as drivers. The independent contractors enjoy flexibility, but their schedules are at the mercy of algorithmic management. This study explores the experiences of food-delivery-platform workers with algorithmic management and their perceptions in response to algorithmic decisions. A collection of 1046 posts containing "AI" (for artificial intelligence) and "algorithm" made to online communities frequently used by platform workers is subjected to keyword analysis, semantic network analysis, and topic modeling using latent Dirichlet allocation. Fifteen latent topics are classified into AI matching, driver behavior, and platform system, and discussed using representative posts from online communities. The results highlight the advancement of AI technology and collaboration strategies used by platform operators to generate trust in algorithmic decisions among food-delivery-platform workers. In addition, this study uses using topic modeling to comprehensively and objectively explore how food-delivery-platform workers with experience working under AI algorithmic management perceive that management. • Food-delivery platforms have begun to apply algorithmic management driven by AI. • We describe how food-delivery-platform workers perceive algorithmic decisions. • Text-mining techniques are used to derive collective perceptions of platform workers. • Latent topics are classified into AI matching, driver behavior, and platform systems. • We highlight the advances in AI technology and the need for a collaborative strategy. [ABSTRACT FROM AUTHOR]

Nowadays, such socio-technical systems as virtual research environments are increasingly employed in the studies of textual heritage. The paper presents a modular platform based on a combination of Web services that provide tools for historical records publishing and research. A promising approach is the creation of semantic networks that encompass marked up texts of records (in this case, medieval documents preserved at Latvian State Historical Archives) and corresponding meta-information. The authors describe the prototype of a software system, which maps XML markup into ACE (Attempto Controlled English) statements; these statements can be automatically translated into OWL (Ontological Web Language), visualized as semantic networks, or queried by means of Web services of Attempto Project. This system is designed as an intelligent environment for collaborative research. [ABSTRACT FROM AUTHOR]

In this paper the issues of upgrading study courses at the Faculty of Computer Science and Information Technology at Riga Technical University and means of their description and evaluation are described. In connection with changes of objectives and methods of education, introduction of new requirements necessity of better structuring and classification concepts in the process of design of study courses arises. The aim of this work was to implement in education such teaching technology that would provide informational and methodological support of the qualitative educational process. For several years, the authors have been researching the way changing and supplementing study courses affects academic performance of students and the quality of training. In this paper, the authors attempt to find a solution for the evaluation of study course quality and success of it modification when changing course content and teaching methodology using quantitative metrics of describing study course semantic network graph. [ABSTRACT FROM AUTHOR]

In this paper , the concept of collaborative intelligent educational system is presented. Different knowledge representation models are compared in the context of their use in collaborative intelligent educational systems. Advantages of semantic networks for knowledge representation in such systems are described. Main advantages of extended semantic networks are shown and a set of basic operations regarding them is drawn up. [ABSTRACT FROM AUTHOR]

Discovery of (strong) association rules, or implications, is an important task in data management, and it finds application in artificial intelligence, data mining and the semantic web. We introduce a novel approach for the discovery of a specific set of implications, called the D -basis, that provides a representation for a reduced binary table, based on the structure of its Galois lattice. At the core of the method are the D -relation defined in a lattice theoretic framework and the hypergraph dualization algorithm that allows us to effectively produce the set of transversals for a given Sperner hypergraph. The latter algorithm, first developed by specialists from Rutgers Center for Operations Research, has already found numerous applications in solving optimization problems in data base theory, artificial intelligence and game theory. One application of the method is for analysis of gene expression data related to a particular phenotypic variable, and some initial testing is done for data provided by the University of Hawaii Cancer Center. [ABSTRACT FROM AUTHOR]

This paper uses a network analysis approach to better understand the relationships and interdependencies between essential flashtrack practices to identify the most central and core practices for successful execution of flash-track projects. Building on prior studies, this paper describes a practice-to-practice semantic network analysis, which reveals a rank order of the differing measures of essential flash-track practices. The findings indicate that establishing fully integrated teams, personnel selection, contractually aligning project participants and colocation of the project team are the most central and core flash-track enablers. Using a semantic network analysis suggests that the following three practices, which were ranked as Tier II practices based on delphi and analytic hierarchy process (AHP) approach in the earlier research, are found to be core and enabling practices and thus merit to be promoted to the Tier I flash-track practices: (1) colocation of the project team; (2) staffing with cooperative and collaborative personnel; and (3) seeking out suppliers and specialty contractors as a source for time-saving innovations. In addition, this research indicates that the following three Tier I practices (according to delphi and AHP) even though important are not considered core or enabling practices (according to network analysis): (1) emphasizing coordination planning during the design process; (2) providing enough resources to critical path items; and (3) simplifying approval procedures. This research offers practical benefits by revealing why a particular flash-track practice is deemed essential. Although there has been some use of network analysis, particularly social network analysis in construction industry research, the authors believe that this flash-track practice-to-practice semantic network analysis is unique. [ABSTRACT FROM AUTHOR]

Semi-supervised clustering seeks to augment traditional clustering methods by incorporating side information provided via human expertise in order to increase the semantic meaningfulness of the resulting clusters. However, most current methods are passive in the sense that the side information is provided beforehand and selected randomly. This may require a large number of constraints, some of which could be redundant, unnecessary, or even detrimental to the clustering results. Thus in order to scale such semi-supervised algorithms to larger problems it is desirable to pursue an active clustering method—i.e., an algorithm that maximizes the effectiveness of the available human labor by only requesting human input where it will have the greatest impact. Here, we propose a novel online framework for active semi-supervised spectral clustering that selects pairwise constraints as clustering proceeds, based on the principle of uncertainty reduction. Using a first-order Taylor expansion, we decompose the expected uncertainty reduction problem into a gradient and a step-scale, computed via an application of matrix perturbation theory and cluster-assignment entropy, respectively. The resulting model is used to estimate the uncertainty reduction potential of each sample in the dataset. We then present the human user with pairwise queries with respect to only the best candidate sample. We evaluate our method using three different image datasets (faces, leaves and dogs), a set of common UCI machine learning datasets and a gene dataset. The results validate our decomposition formulation and show that our method is consistently superior to existing state-of-the-art techniques, as well as being robust to noise and to unknown numbers of clusters. [ABSTRACT FROM AUTHOR]

In this work, we address the human parsing task with a novel Contextualized Convolutional Neural Network (Co-CNN) architecture, which well integrates the cross-layer context, global image-level context, semantic edge context, within-super-pixel context and cross-super-pixel neighborhood context into a unified network. Given an input human image, Co-CNN produces the pixel-wise categorization in an end-to-end way. First, the cross-layer context is captured by our basic local-to-global-to-local structure, which hierarchically combines the global semantic information and the local fine details across different convolutional layers. Second, the global image-level label prediction is used as an auxiliary objective in the intermediate layer of the Co-CNN, and its outputs are further used for guiding the feature learning in subsequent convolutional layers to leverage the global image-level context. Third, semantic edge context is further incorporated into Co-CNN, where the high-level semantic boundaries are leveraged to guide pixel-wise labeling. Finally, to further utilize the local super-pixel contexts, the within-super-pixel smoothing and cross-super-pixel neighbourhood voting are formulated as natural sub-components of the Co-CNN to achieve the local label consistency in both training and testing process. Comprehensive evaluations on two public datasets well demonstrate the significant superiority of our Co-CNN over other state-of-the-arts for human parsing. In particular, the F-1 score on the large dataset [1] reaches 81.72\,\textpercent by Co-CNN, significantly higher than 62.81\,\textpercent and 64.38\,\textpercent by the state-of-the-art algorithms, M-CNN [2] and ATR [1] , respectively. By utilizing our newly collected large dataset for training, our Co-CNN can achieve 85.36\,\textpercent in F-1 score. [ABSTRACT FROM AUTHOR]

Norms and policies are a fundamental component of nowadays open distributed systems operating on the Internet, and a crucial application is the field of Access Control. In such context, policies have been widely used to express data consumer permissions and prohibitions to get access to certain data, whereas obligations have been modelled and applied more infrequently. In this paper, we present a unified semantic model for specifying data consumer and data provider obligations in the context of an exchange of logically formalized information between agents, inside a Multiagent System. We explain how to specify those obligations, in a formal way; we present how to enforce the data provider obligations on the data released by the agent before returning it to the data consumer; we show how to monitor the fulfilment or violation of the specified data consumer obligations; all of that in the context of a complete reasoning-based obligation framework. [ABSTRACT FROM AUTHOR]

The amount of legal information, available digitally, has increased gradually in the past three decades. We are now approaching a situation in which practically all legal information a lawyer needs on a daily basis can be obtained from digital sources. At the same time, powerful retrieval systems capable of integrating these sources and performing more effective search operations have become available. In this paper, new possibilities are outlined that have emerged now that such a large proportion of legal resources have been combined in unified collections. Also, the need to incorporate more advanced 'legal information skills' in the legal curriculum will be discussed. It will be argued that these skills are required to ensure that all newly educated lawyers will be able to use digital legal information optimally, now and in the future. [ABSTRACT FROM AUTHOR]