Concept Retrieval Research Articles

Discovery and recognition of formula concepts using machine learning

Citation-based Information Retrieval (IR) methods for scientific documents have proven effective for IR applications, such as Plagiarism Detection or Literature Recommender Systems in academic disciplines that use many references. In science, technology, engineering, and mathematics, researchers often employ mathematical concepts through formula notation to refer to prior knowledge. Our long-term goal is to generalize citation-based IR methods and apply this generalized method to both classical references and mathematical concepts. In this paper, we suggest how mathematical formulas could be cited and define a Formula Concept Retrieval task with two subtasks: Formula Concept Discovery (FCD) and Formula Concept Recognition (FCR). While FCD aims at the definition and exploration of a ‘Formula Concept’ that names bundled equivalent representations of a formula, FCR is designed to match a given formula to a prior assigned unique mathematical concept identifier. We present machine learning-based approaches to address the FCD and FCR tasks. We then evaluate these approaches on a standardized test collection (NTCIR arXiv dataset). Our FCD approach yields a precision of 68% for retrieving equivalent representations of frequent formulas and a recall of 72% for extracting the formula name from the surrounding text. FCD and FCR enable the citation of formulas within mathematical documents and facilitate semantic search and question answering, as well as document similarity assessments for plagiarism detection or recommender systems.

Pilot trial comparing COVID-19 publication database to conventional online search methods

Background and objectivesLiterature review using search engines results in a list of manuscripts but does not provide the content contained in the manuscripts. Our goal was to evaluate user performance-based...

Image Retrieval Technology of Economic Regulations Based on Semantic Segmentation

Content-based image retrieval (CBIR) is an important part of pattern recognition and artificial intelligence. It has broad application prospects in many important fields, such as digital library, medical image analysis, petroleum geological survey, and public security information retrieval. In this study, statistical modeling and discriminant learning methods are used to analyze and study some key problems in image retrieval, including image concept retrieval, image example retrieval, and relevance feedback. The main research results obtained are as follows: an image classification method based on the Gaussian mixture model (GMM) and max-min posterior pseudo-probability (MMP) discriminant learning is proposed, which is called GMM-MMP method for short; a concept retrieval method based on GMM-MMP is proposed. According to the image concept, the image is divided into two categories: the concept-related image and the concept-unrelated image. The Gaussian mixture model is used to establish the mapping from the image low-level features to the image concept, and the image is classified according to the posterior pseudo-probability classifier to realize the image concept retrieval; an example retrieval method based on GMM-MMP is proposed. According to the image similarity semantics, the image is divided into two categories: the related image and the uncorrelated image of the example image. The Gaussian mixture model is used to establish the mapping from the low-level features of the image to the image similarity semantics. The image is classified according to the posterior pseudo-probability classifier to realize the image case retrieval. Based on the above work, this study implements a content-based image retrieval system.

A Distributed Network for Multimodal Experiential Representation of Concepts.

Neuroimaging, neuropsychological, and psychophysical evidence indicate that concept retrieval selectively engages specific sensory and motor brain systems involved in the acquisition of the retrieved concept. However, it remains unclear which supramodal cortical regions contribute to this process and what kind of information they represent. Here, we used representational similarity analysis of two large fMRI datasets with a searchlight approach to generate a detailed map of human brain regions where the semantic similarity structure across individual lexical concepts can be reliably detected. We hypothesized that heteromodal cortical areas typically associated with the default mode network encode multimodal experiential information about concepts, consistent with their proposed role as cortical integration hubs. In two studies involving different sets of concepts and different participants (both sexes), we found a distributed, bihemispheric network engaged in concept representation, composed of high-level association areas in the anterior, lateral, and ventral temporal lobe; inferior parietal lobule; posterior cingulate gyrus and precuneus; and medial, dorsal, ventrolateral, and orbital prefrontal cortex. In both studies, a multimodal model combining sensory, motor, affective, and other types of experiential information explained significant variance in the neural similarity structure observed in these regions that was not explained by unimodal experiential models or by distributional semantics (i.e., word2vec similarity). These results indicate that during concept retrieval, lexical concepts are represented across a vast expanse of high-level cortical regions, especially in the areas that make up the default mode network, and that these regions encode multimodal experiential information.SIGNIFICANCE STATEMENT Conceptual knowledge includes information acquired through various modalities of experience, such as visual, auditory, tactile, and emotional information. We investigated which brain regions encode mental representations that combine information from multiple modalities when participants think about the meaning of a word. We found that such representations are encoded across a widely distributed network of cortical areas in both hemispheres, including temporal, parietal, limbic, and prefrontal association areas. Several areas not traditionally associated with semantic cognition were also implicated. Our results indicate that the retrieval of conceptual knowledge during word comprehension relies on a much larger portion of the cerebral cortex than previously thought and that multimodal experiential information is represented throughout the entire network.

EEG theta responses induced by emoji semantic violations

This study investigated emoji semantic processing by measuring changes in event-related electroencephalogram (EEG) power. The last segment of experimental sentences was designed as either words or emojis consistent or inconsistent with the sentential context. The results showed that incongruent emojis led to a conspicuous increase of theta power (4–7 Hz), while incongruent words induced a decrease. Furthermore, the theta power increase was observed at midfrontal, occipital and bilateral temporal lobes with emojis. This suggests a higher working memory load for monitoring errors, difficulty of form recognition and concept retrieval in emoji semantic processing. It implies different neuro-cognitive processes involved in the semantic processing of emojis and words.

Psychometric properties of a simple measure of conceptual knowledge: The concept retrieval technique

The objective of the present study was to establish reliability and validity evidence of a simple procedure to measuring conceptual knowledge, the Concept Retrieval Technique (CRT). The CRT procedure requires test takers to freely retrieve concepts from memory they feel are relevant to a given topic, with raters awarding one mark for each correctly retrieved concept when matched against a target word list. Four studies were conducted. In Study 1 (N = 73), inter-rater agreement for the marking procedure of the CRT was determined by means of Cohen’s kappa. The kappa was κ = .85. Study 2 (N = 164) explored how consistent inter-rater agreement was across different subjects, age groups, and raters. Mean kappa was κ = .92 suggesting that the CRT can reliably be marked. Study 3 was concerned with establishing the convergent and discriminant validity of the CRT (N = 55). The correlation between the CRT scores and scores on essay-type items was r = .69 and between assessments that measured other ability domains r = .04 to .21 (n.s.). In Study 4, an experiment was conducted to determine the construct validity of the CRT (N = 45). Participants either acquired, or did not acquire, new knowledge of a particular topic and this manipulation was consistently reflected in their CRT scores. The results of the studies suggest that the CRT is a reliable and valid procedure to measure conceptual knowledge. Implications for education are discussed.

Social Network Big Data Hierarchical High‐Quality Node Mining

Compared with the conventional network data analysis, the data analysis based on social network has a very clear object of analysis, various forms of analysis, and more methods and contents of analysis. If the conventional analysis methods are applied to social network data analysis, we will find that the analysis results do not reach our expected results. The results of the above studies are usually based on statistical methods and machine learning methods, but some systems use other methods, such as self‐organizing self‐learning mechanisms and concept retrieval. With regard to the current data analysis methods, data models, and social network data, this paper conducts a series of researches from data acquisition, data cleaning and processing, data model application and optimization of the model in the process of application, and how the formed data analysis results can be used for managers to make decisions. In this paper, the number of customer evaluations, the time of evaluation, the frequency of evaluation, and the score of evaluation are clustered and analyzed, and finally, the results obtained by the two clustering methods applied in the analysis process are compared to build a customer grading system. The analysis results can be used to maintain the current Amazon purchase customers in a hierarchical manner, and the most valuable customers need to be given key attention, combining social network big data with micro marketing to improve Amazon’s sales performance and influence, developing from the original single shopping mall model to a comprehensive e‐commerce platform, and cultivating their own customer base.

Collective dynamics of repeated inference in variational autoencoder rapidly find cluster structure

Deep neural networks are good at extracting low-dimensional subspaces (latent spaces) that represent the essential features inside a high-dimensional dataset. Deep generative models represented by variational autoencoders (VAEs) can generate and infer high-quality datasets, such as images. In particular, VAEs can eliminate the noise contained in an image by repeating the mapping between latent and data space. To clarify the mechanism of such denoising, we numerically analyzed how the activity pattern of trained networks changes in the latent space during inference. We considered the time development of the activity pattern for specific data as one trajectory in the latent space and investigated the collective behavior of these inference trajectories for many data. Our study revealed that when a cluster structure exists in the dataset, the trajectory rapidly approaches the center of the cluster. This behavior was qualitatively consistent with the concept retrieval reported in associative memory models. Additionally, the larger the noise contained in the data, the closer the trajectory was to a more global cluster. It was demonstrated that by increasing the number of the latent variables, the trend of the approach a cluster center can be enhanced, and the generalization ability of the VAE can be improved.

The relationship between methods of scoring the alternate uses task and the neural correlates of divergent thinking: Evidence from voxel-based morphometry

Divergent thinking tests have been used extensively in neuroscientific studies of creativity. However, output from tests of divergent thinking can be scored in different ways, and those scores can influence assessments of divergent thinking performance and its relationship with brain activation. Here we sought to investigate the relationship between various methods of scoring the Alternate Uses Task (AUT)—a well-known test of divergent thinking—and regional grey matter volume (GMV) using voxel-based morphometry (VBM). We assessed AUT performance based on (a) traditional approaches that involve scoring participants’ output on fluency, flexibility, originality, and elaboration, (b) a subjective approach that involves scoring output directly on “snapshot” creativity, and (c) the definitional approach that involves scoring output separately on novelty and usefulness—the two criteria deemed necessary and jointly sufficient to categorize an idea as creative. Correcting for age, sex, intracranial volume, verbal IQ and working memory capacity, we found negative correlations between regional GMV in the left inferior temporal gyrus (ITG) and novelty and usefulness scores, but no correlation involving other scoring approaches. As part of the brain's core semantic system, this region is involved in concept retrieval and integration. We discuss the implications of these findings for our understanding of the neural bases of divergent thinking, and how ITG could be related to the generation of novel and useful responses.

Changes in description naming for common and proper nouns after left anterior temporal lobectomy

Numerous studies have shown that surgical resection of the left anterior temporal lobe (ATL) is associated with a decline in object naming ability (Hermann et al., 1999). In contrast, few studies have examined the effects of left ATL surgery on auditory description naming (ADN) or category-specific naming. Compared with object naming, which loads heavily on visual recognition processes, ADN provides a more specific measure of concept retrieval. The present study examined ADN declines in a large group of patients who were tested before and after left ATL surgery, using a 2 × 2 × 2 factorial manipulation of uniqueness (common vs. proper nouns), taxonomic category (living vs. nonliving things), and time (pre- vs. postsurgery). Significant declines occurred across all categories but were substantially larger for proper living (PL) concepts, i.e., famous individuals. The disproportionate decline in PL noun naming relative to other conditions is consistent with the notion that the left ATL is specialized not only for retrieval of unique entity concepts, but also plays a role in processing social concepts and person-specific features.

Medical Students Who Prepare for a Lecture Learn More Than Those Who Do Not: An Experimental Study

Purpose: Medical students tend to attend lectures unprepared before studying the literature for an examination. Would students not profit more from lectures if they study the literature first? Method: One hundred and two students from a Saudi-Arabian medical school were included in the study. Half of them received a four-page text to be studied before a 15-minutes lecture; the other half received the lecture first and then the text. Knowledge level was tested using a concept retrieval test at three points in time, before the experiment, after reading the text, and after attending the lecture. This enabled studying the effects of the various treatments separately. Results: The mean student-scores were as follows: After lecture-only: 54.7±23.5, after reading-only: 45.7±11.2, lecture-then-reading: 69.9±15.8 and reading-then-lecture: 78.2±17.4. The combination of lecture and reading in both groups was statistically significantly different from lecture-only. Reading-then-lecture was significantly better than lecture-then-reading. Conclusion: In contrast to existing practices of students in medical education, it is better to study first before attending a lecture. Keywords: Lectures; self-study; concept retrieval test. DOI: 10.7176/JEP/11-8-06 Publication date: March 31 st 2020

Automated screening of natural language in electronic health records for the diagnosis septic shock is feasible and outperforms an approach based on explicit administrative codes

PurposeIdentification of patients for epidemiologic research through administrative coding has important limitations. We investigated the feasibility of a search based on natural language processing (NLP) on the text sections of electronic health records for identification of patients with septic shock. Materials and methodsResults of an explicit search strategy (using explicit concept retrieval) and a combined search strategy (using both explicit and implicit concept retrieval) were compared to hospital ICD-9 based administrative coding and to our department's own prospectively compiled infection database. ResultsOf 8911 patients admitted to the medical or surgical ICU, 1023 (11.5%) suffered from septic shock according to the combined search strategy. This was significantly more than those identified by the explicit strategy (518, 5.8%), by hospital administrative coding (549, 5.8%) or by our own prospectively compiled database (609, 6.8%) (p < .001). Sensitivity and specificity of the automated combined search strategy were 72.7% (95%CI 69.0%–76.2%) and 93.0% (95%CI 92.4%–93.6%), compared to 56.0% (95%CI 52.0%–60.0%) and 97.5% (95%CI 97.1%–97.8%) for hospital administrative coding. ConclusionsAn automated search strategy based on a combination of explicit and implicit concept retrieval is feasible to screen electronic health records for septic shock and outperforms an administrative coding based explicit approach.

Use of Concept Retrieval Technique as an Assessment Tool of Long-term Knowledge of Medical Students

PurposeMedical students’ knowledge assessment is a challenge. Multiple-choice questions (MCQs) despite their limitations are commonly used for this purpose. The Concept Retrieval Technique (CRT) is a relatively new knowledge assessment method. In this study we investigated the use of CRT to assess a particular long-term knowledge of medical students from different study levels. MethodOne hundred forty-four medical students (50% females) participated in an Endocrinology CRT exam. The students were either in fourth, fifth-or sixth-year of the medical program. Fourth-year students were tested before they received an endocrinology course whereas students from the fifth- and sixth-year respectively took the course ten and twenty-four months before being tested. The CRT was rated by two raters independently. ResultsThere was strong effect of year of study on the CRT score. Fourth-year students scored significantly lower than fifth- and sixth-year students. No significant differences in CRT score emerged between fifth- and sixth-year students. The inter-rater agreement was acceptable indicating that the CRT is a reliable assessment method. The CRT showed construct validity. ConclusionThe CRT is an assessment tool that could assess long-term knowledge retention properly, and has the ability to differentiate students according to their study level.

The left temporal pole is a convergence region mediating the relation between names and semantic knowledge for unique entities: Further evidence from a “recognition-from-name” study in neurological patients

Prior research has implicated the left temporal pole (LTP) as a critical region for naming semantically unique items, including famous faces, landmarks, and musical melodies. Most studies have used a confrontation naming paradigm, where a participant is presented with a stimulus and asked to retrieve its name. We have proposed previously that the LTP functions as a two-way, bidirectional convergence region brokering between conceptual knowledge and proper names for unique entities. Under this hypothesis, damage to the LTP should result in a “two way” impairment: (1) defective proper name retrieval when presented with a unique stimulus (as shown in prior work); and (2) defective concept retrieval when presented with a proper name. Here, we directly tested the second prediction using a “recognition-from-name” paradigm. Participants were patients with LTP damage, brain-damaged comparisons with damage outside the LTP, and healthy comparisons. Participants were presented with names of famous persons (e.g., “Marilyn Monroe”), landmarks (e.g., “Leaning Tower of Pisa”), or melodies (e.g., “Rudolph the Red-Nosed Reindeer”) and were asked to provide conceptual knowledge about each. We found that individuals with damage to the LTP were significantly impaired at conceptual knowledge retrieval when given names of famous people and landmarks (but this finding did not hold for melodies). This outcome supports the theory that the LTP is a bidirectional convergence region for proper naming, but suggests that melody retrieval may rely on processes different from those supported by the LTP.

IF-MCA: Importance Factor-Based Multiple Correspondence Analysis for Multimedia Data Analytics

Multimedia concept detection is a challenging topic due to the well-known class imbalance issue, where the data instances are distributed unevenly across different classes. This problem becomes even more prominent when the minority class that contains an extremely small proportion of the data represents the concept of interest as has occurred in many real-world applications such as frauds in banking transactions and goal events in soccer videos. Traditional data mining approaches often have difficulty handling largely skewed data distributions. To address this issue, in this paper, an importance-factor (IF)-based multiple correspondence analysis (MCA) framework is proposed to deal with the imbalanced datasets. Specifically, a hierarchical information gain analysis method, which is inspired by the decision tree algorithm, is presented for critical feature selection and IF assignment. Then, the derived IF is incorporated with the MCA algorithm for effective concept detection and retrieval. The comparison results in video concept detection using the disaster dataset and the soccer dataset demonstrate the effectiveness of the proposed framework.

Correlation-Assisted Imbalance Multimedia Concept Mining and Retrieval

In the past decades, we have witnessed an explosion of multimedia data, especially with the development of social media websites and blooming popularity of smart devices. As a result, multimedia semantic concept mining and retrieval whose objective is to mine useful information from the large amount of multimedia data including texts, images, and videos has become more and more important. The huge amount of multimedia data and the semantic gap between low-level features and high-level semantic concepts have made it even more challenging. To address these challenges, the correlations among the classes can provide important context cues to help bridge the semantic gap. Meanwhile, many real-world datasets do not have uniform class distributions while the minority instances actually represent the concept of interests, like frauds in transactions, intrusions in network security, and unusual events in surveillance. Despite extensive research efforts, imbalanced concept retrieval remains one of the most challenging research problems in multimedia data mining. Different from existing frameworks regarding concept correlations among labels, this paper presents a novel concept correlation analysis model using the correlation between the retrieval scores and labels. Experimental results on the TRECVID benchmark datasets demonstrate that the proposed framework can enhance imbalanced concept mining and retrieval even with trivial scores from the minority class.

Multilingual visual sentiment concept clustering and analysis

Visual content is a rich medium that can be used to communicate not only facts and events, but also emotions and opinions. In some cases, visual content may carry a universal affective bias (e.g., natural disasters or beautiful scenes). Often however, to achieve a parity in the affections a visual media invokes in its recipient compared to the one an author intended requires a deep understanding and even sharing of cultural backgrounds. In this study, we propose a computational framework for the clustering and analysis of multilingual visual affective concepts used in different languages which enable us to pinpoint alignable differences (via similar concepts) and nonalignable differences (via unique concepts) across cultures. To do so, we crowdsource sentiment labels for the MVSO dataset, which contains 16 K multilingual visual sentiment concepts and 7.3M images tagged with these concepts. We then represent these concepts in a distribution-based word vector space via (1) pivotal translation or (2) cross-lingual semantic alignment. We then evaluate these representations on three tasks: affective concept retrieval, concept clustering, and sentiment prediction—all across languages. The proposed clustering framework enables the analysis of the large multilingual dataset both quantitatively and qualitatively. We also show a novel use case consisting of a facial image data subset and explore cultural insights about visual sentiment concepts in such portrait-focused images.

Efficient Imbalanced Multimedia Concept Retrieval by Deep Learning on Spark Clusters

The classification of imbalanced datasets has recently attracted significant attention due to its implications in several real-world use cases. The classifiers developed on datasets with skewed distributions tend to favor the majority classes and are biased against the minority class. Despite extensive research interests, imbalanced data classification remains a challenge in data mining research, especially for multimedia data. Our attempt to overcome this hurdle is to develop a convolutional neural network (CNN) based deep learning solution integrated with a bootstrapping technique. Considering that convolutional neural networks are very computationally expensive coupled with big training datasets, we propose to extract features from pre-trained convolutional neural network models and feed those features to another full connected neutral network. Spark implementation shows promising performance of our model in handling big datasets with respect to feasibility and scalability.

Heteromodal Cortical Areas Encode Sensory-Motor Features of Word Meaning.

The capacity to process information in conceptual form is a fundamental aspect of human cognition, yet little is known about how this type of information is encoded in the brain. Although the role of sensory and motor cortical areas has been a focus of recent debate, neuroimaging studies of concept representation consistently implicate a network of heteromodal areas that seem to support concept retrieval in general rather than knowledge related to any particular sensory-motor content. We used predictive machine learning on fMRI data to investigate the hypothesis that cortical areas in this "general semantic network" (GSN) encode multimodal information derived from basic sensory-motor processes, possibly functioning as convergence-divergence zones for distributed concept representation. An encoding model based on five conceptual attributes directly related to sensory-motor experience (sound, color, shape, manipulability, and visual motion) was used to predict brain activation patterns associated with individual lexical concepts in a semantic decision task. When the analysis was restricted to voxels in the GSN, the model was able to identify the activation patterns corresponding to individual concrete concepts significantly above chance. In contrast, a model based on five perceptual attributes of the word form performed at chance level. This pattern was reversed when the analysis was restricted to areas involved in the perceptual analysis of written word forms. These results indicate that heteromodal areas involved in semantic processing encode information about the relative importance of different sensory-motor attributes of concepts, possibly by storing particular combinations of sensory and motor features. The present study used a predictive encoding model of word semantics to decode conceptual information from neural activity in heteromodal cortical areas. The model is based on five sensory-motor attributes of word meaning (color, shape, sound, visual motion, and manipulability) and encodes the relative importance of each attribute to the meaning of a word. This is the first demonstration that heteromodal areas involved in semantic processing can discriminate between different concepts based on sensory-motor information alone. This finding indicates that the brain represents concepts as multimodal combinations of sensory and motor representations.

Erratum to: Weighted subspace modeling for semantic concept retrieval using gaussian mixture models

Erratum to: Weighted subspace modeling for semantic concept retrieval using gaussian mixture models