High-level Concepts Research Articles

A cognitive science perspective on knowledge acquisition

Looking back over 25yrs of R&D in knowledge acquisition – particularly mediated by workshops and a journal founded by Brian Gaines – it is remarkable how topics and perspectives have changed. We started from the assumption that the knowledge acquisition bottleneck was due to the problems of experts expressing their expert knowledge. However, explicitly available knowledge, as, e.g. in handbooks, appeared often sufficient to construct an adequate ‘artificial problem solver’. Expert systems became knowledge systems. The bottleneck was rather experienced in modeling the domain knowledge and the reasoning control – the problem solving method (PSM) – for automated execution. By developing libraries of reusable PSMs, or implementing shells with built-in PSMs, this bottleneck could be eased. Specifications of domain knowledge lead to the development of ontologies: another source for reuse. Despite the fact that this brought sufficient technology for a mature engineering methodology, little use is made of it in current practice, and developing ‘intelligent’ software is not much different today from what it was at the beginning of the 1980s. A more impressive heritage of knowledge acquisition R&D is the introduction of technology for building ontologies. These found their way via knowledge management to the architecture of the Semantic Web. In research apparent solutions also bring new problems. Two of these problems are suggested by empirical cognitive science research. The first one is that knowledge as represented in currently available (top-level) ontologies are too simple, because high level concepts may come in design ‘patterns’: a view that has recently also been taken up in the knowledge acquisition community. The second problem is the fact that despite much empirical research in cognitive psychology, we have insufficient insight in how we acquire new conceptualisations from text. This is a serious bottleneck for an early dream in knowledge acquisition: automated knowledge acquisition.

Learning a hierarchical image manifold for Web image classification

Image classification is an essential task in content-based image retrieval. However, due to the semantic gap between low-level visual features and high-level semantic concepts, and the diversification of Web images, the performance of traditional classification approaches is far from users’ expectations. In an attempt to reduce the semantic gap and satisfy the urgent requirements for dimensionality reduction, high-quality retrieval results, and batch-based processing, we propose a hierarchical image manifold with novel distance measures for calculation. Assuming that the images in an image set describe the same or similar object but have various scenes, we formulate two kinds of manifolds, object manifold and scene manifold, at different levels of semantic granularity. Object manifold is developed for object-level classification using an algorithm named extended locally linear embedding (ELLE) based on intra- and inter-object difference measures. Scene manifold is built for scene-level classification using an algorithm named locally linear submanifold extraction (LLSE) by combining linear perturbation and region growing. Experimental results show that our method is effective in improving the performance of classifying Web images.

Cross-Category Adaptation: Objects Produce Gender Adaptation in the Perception of Faces

Adaptation aftereffects have been found for low-level visual features such as colour, motion and shape perception, as well as higher-level features such as gender, race and identity in domains such as faces and biological motion. It is not yet clear if adaptation effects in humans extend beyond this set of higher order features. The aim of this study was to investigate whether objects highly associated with one gender, e.g. high heels for females or electric shavers for males can modulate gender perception of a face. In two separate experiments, we adapted subjects to a series of objects highly associated with one gender and subsequently asked participants to judge the gender of an ambiguous face. Results showed that participants are more likely to perceive an ambiguous face as male after being exposed to objects highly associated to females and vice versa. A gender adaptation aftereffect was obtained despite the adaptor and test stimuli being from different global categories (objects and faces respectively). These findings show that our perception of gender from faces is highly affected by our environment and recent experience. This suggests two possible mechanisms: (a) that perception of the gender associated with an object shares at least some brain areas with those responsible for gender perception of faces and (b) adaptation to gender, which is a high-level concept, can modulate brain areas that are involved in facial gender perception through top-down processes.

Exemplar: A Source Code Search Engine for Finding Highly Relevant Applications

A fundamental problem of finding software applications that are highly relevant to development tasks is the mismatch between the high-level intent reflected in the descriptions of these tasks and low-level implementation details of applications. To reduce this mismatch we created an approach called EXEcutable exaMPLes ARchive (Exemplar) for finding highly relevant software projects from large archives of applications. After a programmer enters a natural-language query that contains high-level concepts (e.g., MIME, datasets), Exemplar retrieves applications that implement these concepts. Exemplar ranks applications in three ways. First, we consider the descriptions of applications. Second, we examine the Application Programming Interface (API) calls used by applications. Third, we analyze the dataflow among those API calls. We performed two case studies (with professional and student developers) to evaluate how these three rankings contribute to the quality of the search results from Exemplar. The results of our studies show that the combined ranking of application descriptions and API documents yields the most-relevant search results. We released Exemplar and our case study data to the public.

Image retrieval based on high level concept detection and semantic labelling

This paper presents a novel approach to high-level concept detection and retrieval in images based on a combination of visual thesaurus and multi-class supervised learning. The visual thesaurus includes both conceptual and spatial location information of semantic concepts that are key to image labelling. Our image annotation (or labelling) process includes segmenting and building an image signature. The visual thesaurus is then built using a multi-class supervised SYM classifier. Algorithm for spatial location matching is included. Similarity matching during retrieval is performed on both the content as well as the location information using the standard Euclidean distance. Corel data set was used for experimentation and results were compared with two related approaches to visual thesaurus and image retrieval.

Social Q&A and virtual reference-comparing apples and oranges with the help of experts and users

Online question-answering (Q&A) services are becoming increasingly popular among information seekers. We divide them into two categories, social Q&A (SQA) and virtual reference (VR), and examine how experts (librarians) and end users (students) evaluate information within both categories. To accomplish this, we first performed an extensive literature review and compiled a list of the aspects found to contribute to a “good” answer. These aspects were divided among three high-level concepts: relevance, quality, and satisfaction. We then interviewed both experts and users, asking them first to reflect on their online Q&A experiences and then comment on our list of aspects. These interviews uncovered two main disparities. One disparity was found between users’ expectations with these services and how information was actually delivered among them, and the other disparity between the perceptions of users and experts with regard to the aforementioned three characteristics of relevance, quality, and satisfaction. Using qualitative analyses of both the interviews and relevant literature, we suggest ways to create better hybrid solutions for online Q&A and to bridge the gap between experts’ and users’ understandings of relevance, quality, and satisfaction, as well as the perceived importance of each in contributing to a good answer. © 2012 Wiley Periodicals, Inc.

Non-collaborative content detecting on video sharing social networks

In this work we are concerned with detecting non-collaborative videos in video sharing social networks. Specifically, we investigate how much visual content-based analysis can aid in detecting ballot stuffing and spam videos in threads of video responses. That is a very challenging task, because of the high-level semantic concepts involved; of the assorted nature of social networks, preventing the use of constrained a priori information; and, which is paramount, of the context-dependent nature of non-collaborative videos. Content filtering for social networks is an increasingly demanded task: due to their popularity, the number of abuses also tends to increase, annoying the user and disrupting their services. We propose two approaches, each one better adapted to a specific non-collaborative action: ballot stuffing, which tries to inflate the popularity of a given video by giving "fake" responses to it, and spamming, which tries to insert a non-related video as a response in popular videos. We endorse the use of low-level features combined into higher-level features representation, like bag-of-visual-features and latent semantic analysis. Our experiments show the feasibility of the proposed approaches.

Reviving organisational memetics through Cultural Linnæanism

PurposeThe initial purpose of this paper is to review the explanatory power that memetics promised for socio‐cultural evolutionary theory, for organisational adaptation, and emergent patterns of traits. Second, to argue that philosophical accusations and premature demands have retarded a science of memetics; regardless, isolated demonstrations of empirical research feasibility suggest a pragmatic resolution. Third, to speculate about practical applications, future advances, and prompt consideration about resuming methodological research initiatives that draw extensively from biology into organisational and managements science.Design/methodology/approachOwing to present methodological immaturity of cultural science then a high conceptual level of meta‐methodology is required. This scope necessarily overlooks specific technical details. Life‐science principles are well known in comparison to the embryonic memetic and cultural sciences. The meme‐gene analogy builds a bridge across which we can draw candidate hypotheses and established methods. However, memetics has inherited the expectations of genetics but without its developmental history. Memetics therefore would benefit from recapitulating the ontogenesis of the more senior science by drawing upon foundational methods.FindingsLinnæan Systematics was elemental to evolutionary theory and genetics; a cultural analogue is proposed. Retreating to description would support emerging objective organisational taxonomies that are laying the methodological foundations for a potential synthesis between organisational replicator and evolutionary theories.Research limitations/implicationsAt the moment, the number of organisational examples are few, which further suggests the fundamental nature of this area of research. They serve to illustrate that a large array of hypotheses and methods can be adapted from the biological domain, opening up a bloom of research implications for the organisational domain.Originality/valueDiscourse about memetics is commonplace, but empirical research has been undermined. Originality stems from reapplying established biological methods to the new organisational domain. The value is in conferring the rigour of natural science to socio‐cultural study.

A declarative model specification system allowing NeuroML to be extended with user-defined component types

Methods for storing and sharing biological models tend to focus on directly encoding a model and its equations. However, many models share the same equations, subject only to differences in parameter values or the number of instances of particular processes. We have therefore developed a mechanism within NeuroML[1] to express the common structural and mathematical features of a family of models separately from the parameter values that define a particular member of the family. Such specifications are important because they correspond closely to the mathematical understanding that computational scientists have for the modeled system. Supporting this conceptual level separately from the models themselves allows particular instances of models to be expressed more concisely and makes the commonalities among models explicit. It also facilitates a wide range of model related processing including systematic comparison, model composition, and sensitivity analysis. The resulting system, known as LEMS (Low Entropy Model Specification), makes it possible to retro-fit most of the existing NeuroML version 1 definitions with domain independent definitions constructed from LEMS elements. This preserves the benefits of the existing semantically rich high-level concepts in NeuroML while providing the ability to add new concepts at the user level rather than requiring changes to the language. Simulator developers have a choice between directly supporting the library of core types, or supporting the underlying LEMS definitions. These options are not exclusive: a tool could provide hard-coded support for certain key components such as ion channels, and fall back to a generic (and probably less efficient) implementation for new component definitions. Component type definitions in LEMS consist of three main parts: a declaration of the static properties needed to define a particular type of model; specification of the dynamics of the component; and specification of the component’s structure. The static definition includes parameters types, component nesting, and references between components. The dynamics contains the equations that govern the behavior of an instance of the model. The structure definition specifies how the static properties should be interpreted when a component is used. This may be as simple as creating a single instance of each of the state variables, but may also involve instantiating other components or connecting them together as in, for example, component types that express network structure or connectivity. As well as an extensive library of component types, and example models, NeuroML2/LEMS contains a reference interpreter that can run any LEMS model. Although initial development has been driven by the need to support the requirements of the NeuroML community, the resulting language only includes mathematical and structural elements and should be applicable to other model specification tasks.

The next 50years: A personal view

I review history, starting with Turing’s seminal paper, reaching back ultimately to when our species started to outperform other primates, searching for the questions that will help us develop a computational account of human intelligence. I answer that the right questions are: What’s different between us and the other primates and what’s the same. I answer the what’s different question by saying that we became symbolic in a way that enabled story understanding, directed perception, and easy communication, and other species did not. I argue against Turing’s reasoning-centered suggestions, offering that reasoning is just a special case of story understanding. I answer the what’s the same question by noting that our brains are largely engineered in the same exotic way, with information flowing in all directions at once. By way of example, I illustrate how these answers can influence a research program, describing the Genesis system, a system that works with short summaries of stories, provided in English, together with low-level common-sense rules and higher-level concept patterns, likewise expressed in English. Genesis answers questions, notes abstract concepts such as revenge, tells stories in a listener-aware way, and fills in story gaps using precedents. I conclude by suggesting, optimistically, that a genuine computational theory of human intelligence will emerge in the next 50 years if we stick to the right, biologically inspired questions, and work toward biologically informed models.

Sentic Album: Content-, Concept-, and Context-Based Online Personal Photo Management System

The world of online personal photo management has come a long way in the past few years, but today, there are still huge gaps in annotating, organizing, and retrieving online pictures in such a way that they can be easily queried and visualized. Existing content-based image retrieval systems apply statistics, pattern recognition, signal processing, and computer vision techniques but these are still too weak to ‘bridge the semantic gap’ between the low-level data representation and the high-level concepts the user associates with images. Image meta search engines, on the other hand, rely on tags associated with online pictures but results are often too inaccurate since they mainly depend on keyword-based rather than concept-based algorithms. Sentic Album is a novel content-, concept-, and context-based online personal photo management system that exploits both data and metadata of online personal pictures to intelligently annotate, organize, and retrieve them. Many salient features of pictures, in fact, are only noticeable in the viewer’s mind, and the cognitive ability to grasp such features is a key aspect for accordingly analyzing and classifying personal photos. To this end, Sentic Album exploits not just colors and texture of online images (content), but also the cognitive and affective information associated with their metadata (concept), and their relative timestamp, geolocation, and user interaction metadata (context).

A Two-View Concept Correlation Based Video Annotation Refinement

Recently, concept correlation defining the relationship between concepts has been playing an important role in video annotation (or concept detection). To improve the annotation performance, this paper presents a two-view concept correlation based video annotation refinement, using data-specific spatial and temporal concept correlations. Specifically, instead of generic concept correlation within shots, the spatial view estimates a data-specific concept correlation for each shot, via introducing concept correlation bases to map low-level features to high-level concept distribution under the framework of sparse representation. On the other hand, beyond the temporal consistency of one concept, a richer temporal correlation between different concepts respectively locating in the current shot and its neighbors is utilized to adjust the detection scores. In the end, these two types of concept correlations are integrated into a probability calculation based framework to refine the initial results derived from multiple concept detectors. And the experiments conducted on TRECVID 2006-2008 datasets and comparison with existing works demonstrate its effectiveness.

A Parametric, Temporal Model of Musical Tension

tension in music is a high-level concept that is difficult to formalize due to its complex, multidimensional nature. This paper proposes a quantitative model of musical tension that takes into account the dynamic, temporal aspects of listening. The model is based on data from two experiments. The first was a web-based study that was designed to examine how individual musical parameters contribute directly to a listener's overall perception of tension and how those parameters interact. The second study was an in-lab experiment in which listeners were asked to provide continuous responses to longer, more complex musical stimuli. Both studies took into account a number of musical parameters including harmony, pitch height, melodic expectation, dynamics, onset frequency, tempo, meter, rhythmic regularity, and syncopation. As an initial step, linear and nonlinear models were explored for predicting tension given analytical descriptions of various musical parameters. These models were tested on the continuous-response data from Experiment 2 and shown to be insufficient. An alternate model was proposed based on the notion of a moving perceptual window in time and the concept of trend salience. High correlation with empirical data indicates that this parametric, temporal model accurately predicts tension judgments for complex musical stimuli.

Remotely sensed image retrieval based on region-level semantic mining

AbstractAs satellite images are widely used in a large number of applications in recent years, content-based image retrieval technique has become important tools for image exploration and information mining; however, their performances are limited by the semantic gap between low-level features and high-level concepts. To narrow this semantic gap, a region-level semantic mining approach is proposed in this article. Because it is easier for users to understand image content by region, images are segmented into several parts using an improved segmentation algorithm, each with homogeneous spectral and textural characteristics, and then a uniform region-based representation for each image is built. Once the probabilistic relationship among image, region, and hidden semantic is constructed, the Expectation Maximization method can be applied to mine the hidden semantic. We implement this approach on a dataset consisting of thousands of satellite images and obtain a high retrieval precision, as demonstrated through experiments.

The Spiral Discovery Method: An Interpretable Tuning Model for CogInfoCom Channels

Cognitive Infocommunications (CogInfoCom) messages that are used to carry information on the state of the same high-level concept can be regarded as belonging to a CogInfoCom channel. Such channels can be generated using any kind of parametric model. By changing the values of the parameters, it is possible to arrive at a large variety of CogInfoCom messages, a subset of which can belong to a CogInfoCom channel – provided they are perceptually well-suited to the purpose of conveying information on the same highlevel concept. Thus, for any CogInfoCom channel, we may speak of a parameter space and a perceptual space that is created by the totality of messages in the CogInfoCom channel. In this paper, we argue that in general, the relationship between the parameter space and the perceptual space is highly non-linear. For this reason, it is extremely difficult for the designer of a CogInfoCom channel to tune the parameters in such a way that the resulting CogInfoCom messages are perceptually continuous, and suitable to carry information on a single high-level concept. To address this problem, we propose a cognitive artifact that uses a rank concept available in tensor algebra to provide the designer of CogInfoCom channels with practical tradeoffs between complexity and interpretability. We refer to the artifact as the Spiral Discovery Method (SDM).

Semantic context based refinement for news video annotation

Automatic video annotation is to bridge the semantic gap and facilitate concept based video retrieval by detecting high level concepts from video data. Recently, utilizing context information has emerged as an important direction in such domain. In this paper, we present a novel video annotation refinement approach by utilizing extrinsic semantic context extracted from video subtitles and intrinsic context among candidate annotation concepts. The extrinsic semantic context is formed by identifying a set of key terms from video subtitles. The semantic similarity between those key terms and the candidate annotation concepts is then exploited to refine initial annotation results, while most existing approaches utilize textual information heuristically. Similarity measurements including Google distance and WordNet distance have been investigated for such a refinement purpose, which is different with approaches deriving semantic relationship among concepts from given training datasets. Visualness is also utilized to discriminate individual terms for further refinement. In addition, Random Walk with Restarts (RWR) technique is employed to perform final refinement of the annotation results by exploring the inter-relationship among annotation concepts. Comprehensive experiments on TRECVID 2005 dataset have been conducted to demonstrate the effectiveness of the proposed annotation approach and to investigate the impact of various factors.

Fuzzy cognitive maps and cellular automata: An evolutionary approach for social systems modelling

One of the first decisions to be made when modelling a phenomenon is that of scale: at which level is the phenomenon most appropriately modelled? For some phenomena the answer may seem too obvious to warrant even asking the question, but other phenomena cover the gamut, from high to low levels of abstraction. This paper explores how two modelling approaches that are ‘at home’ at opposite ends of the abstraction spectrum can be combined to yield an evolutionary modelling approach that is especially apt for phenomena that cover a wide range in this spectrum.We employ fuzzy cognitive maps (FCMs) to model the interplay between high-level concepts, and cellular automata (CA) to model the low-level interactions between individual actors. The combination of these models carries both beyond their respective limitations: the FCM concept is extended beyond the derivation of equilibrium outcomes from static initial conditions, to time-evolving systems where conditions may vary; CA are extended beyond the emergence of patterns from local interactions, to systems where global patterns have local repercussions.The applicability of the methodology is demonstrated by modelling the spread of human immunodeficiency virus (HIV) in an environment in which injection drug users share paraphernalia.

Strategic reasoning in p-beauty contests

This paper analyzes strategic choice in p-beauty contests. First, I show that it is not generally a best reply to guess the expected target value, even in games with n>2 players, and that iterated best response strictly applied does not induce a choice sequence approximating pk⋅0.5. Second, I argue that the beliefs and actions of players typically considered to be level 2–4 are intuitively captured also by high-level concepts such as quantal response equilibrium and noisy introspection. Third, I analyze this hypothesis econometrically. The results concur. In six different data sets, the choices are described more adequately as mixtures of quantal response equilibrium and noisy introspection than as level-k mixtures.

Rational Approaches to Improving Selectivity in Drug Design

Rational Approaches to Improving Selectivity in Drug Design

The Impact of Arithmetic Skills on Mastery of Quantitative Analysis

Over the past several years math education has moved from a period where all math calculations were done by hand to an era where most calculations are done using a calculator or computer. There are certainly benefits to this approach, but when one concomitantly recognizes the declining scores on national standardized mathematics exams, it raises the question, “Could the lack of technology-assisted arithmetic manipulation skills have a carryover to understanding higher-level mathematical concepts or is it just a spurious correlation?” Eighty-seven students were tested for their ability to do simple arithmetic and algebra by hand. These scores were then regressed on three important areas of quantitative analysis: recognizing the appropriate tool to use in an analysis, creating a model to carry out the analysis, and interpreting the results of the analysis. The study revealed a significant relationship between the ability to accurately do arithmetic calculations and the ability to recognize the appropriate tool and creating a model. It found no significant relationship between results interpretation and arithmetic skills.