Semantic Coverage Research Articles

CAN: Concept‐Aligned Neurons for Visual Comparison of Deep Neural Network Models

AbstractWe present concept‐aligned neurons, or CAN, a visualization design for comparing deep neural networks. The goal of CAN is to support users in understanding the similarities and differences between neural networks, with an emphasis on comparing neuron functionality across different models. To make this comparison intuitive, CAN uses concept‐based representations of neurons to visually align models in an interpretable manner. A key feature of CAN is the hierarchical organization of concepts, which permits users to relate sets of neurons at different levels of detail. CAN's visualization is designed to help compare the semantic coverage of neurons, as well as assess the distinctiveness, redundancy, and multi‐semantic alignment of neurons or groups of neurons, all at different concept granularity. We demonstrate the generality and effectiveness of CAN by comparing models trained on different datasets, neural networks with different architectures, and models trained for different objectives, e.g. adversarial robustness, and robustness to out‐of‐distribution data.

Evaluating the effectiveness of functional decomposition in early-stage design: development and application of problem space exploration metrics

This paper aims to explore metrics for evaluating the effectiveness of functional decomposition methods regarding problem space exploration at the early design stage. Functional decomposition involves breaking down the main purpose of a complex problem or system into a set of more manageable sub-functions, leading to a clearer understanding of the problem space and its various aspects. While various metrics have been used to evaluate functional decomposition outcomes, little literature has focused on assessing its effectiveness in problem space exploration. To address the gap, this research introduces three metrics for problem space evaluation defined by functional models: quantity of unique functions (M1), breadth and depth of the hierarchical structure (M2), and relative semantic coverage ratio of the problem space (M3). An example study is conducted to illustrate the evaluation process, comparing functional analysis with and without explicit human-centric considerations using a power screwdriver as a case product. The analysis in the example study reveals that the breadth of the hierarchical structure (part of M2) is marginally larger in the condition with explicit human-centric considerations (Condition A) compared to the condition without such considerations (Condition B). However, no significant differences are observed in terms of other metrics. The qualitative analysis based on semantic comparisons suggests that Condition A facilitates participants in generating a diverse set of functions supporting user safety requirements more effectively than Condition B. Overall, the example study demonstrates the evaluation process for each metric and discusses their nuances and limitations. By proposing these metrics, this research contributes to benchmarking and evaluating the effectiveness of different methods in promoting functional analysis in engineering design. The metrics provide valuable insights into problem space exploration, offering designers a better understanding of the efficacy of their functional decomposition methods in early design stages. This, in turn, fosters more informed decision-making and contributes to the advancement of functional analysis methodologies in engineering design practices.

Learning semantic categories of L2 verbs: The case of cutting and breaking verbs.

To attain native-like proficiency in second-language word usage, learners have to discover intricate semantic categories in the target language. We investigated the factors influencing the development of two aspects of second-language learners' semantic categories: the category center and category boundary of word meanings. In the experiment, second-language learners of Japanese, whose first language is Mandarin, were asked to produce the best verb for 28 videos depicting various cutting and breaking events. Descriptive analyses were conducted to compare the verb patterns used by second-language learners with those of native speakers. The second-language learners' verb use pattern suggested their struggle in delineating the semantic ranges of breaking verbs in a native-like manner. Model analyses further revealed that different factors contribute to learning two different aspects of word meanings. The learning category center of word meaning depended on the similarity between the lexical domains in the first and second languages. On the contrary, the success of learning the semantic boundaries of verbs required a large input frequency and smaller semantic coverage, and smaller category ambiguity. The results suggest that constructing a semantic domain in the second language should be evaluated from at least two different aspects of semantic representation.

Prompt template construction by Average Gradient Search with External Knowledge for aspect sentimental analysis

Aspect-based Sentiment Analysis (ABSA) aims to predict the sentiment polarity towards a particular aspect in a sentence. Most of the existing methods construct neural networks or fine-tune the pre-trained language models. However, it is difficult to fully utilize the knowledge of language model learned during pretraining process, and it usually performs worse in few-shot experiment. Prompt learning can alleviate the above problems effectively. Manual prompt template construction methods are adopted usually but it leads to high costs and low efficiency, and there are seldom prompt templates applicable to ABSA task. We propose a method for ABSA prompt template construction, named Average Gradient Search with External Knowledge Based on KNN. The vocabulary list of the pre-trained language model is built as a KD-Tree, and the KNN algorithm is used to search the best prompt template on KD-Tree. For the sentiment polarity prediction, the best prompt template is used to wrap the ABSA data, converting the classification task to generative task, which enables the language model to fully leverage the learned knowledge from the pre-training stage. Moreover, the constructed verbalizer incorporates external knowledge to provide label words of each class with extensive semantic coverage and lower subjective bias. The comparison experimental results on SemEval2014 Restaurant and Laptop datasets demonstrate that the proposed method outperforms the existing SOTA models. Further, the few-shot experimental results indicate that, with only a small number of training samples, the proposed method achieves comparable or even better performance than the baseline model trained on the full dataset.

A Corpus-based Study on Shell Nouns in “N + that” Construction in Popular Science Discourse: Rachel Carson’s Works as Example

By adopting the corpus-based approach, the present study compares the frequency of use, semantic distribution, and stance expression of shell nouns in the “N + that” clause in popular science discourse compared with academic writing. The results show that, firstly, the frequency of use of shell nouns in the “N + that” clause in popular science is lower than that in academic discourse. Secondly, the semantic coverage of shell nouns in popular science discourse is smaller than that in academic discourse, but its semantic proportion distribution is roughly the same, mainly including mental and linguistic shell nouns. Thirdly, in popular science discourse and academic discourse, the proportion of shell nouns with epistemic stances occupies an absolute advantage, while the proportion of shell nouns with attitudinal stances is less. The proportion of shell nouns expressing attitudinal stance in popular science discourse is greater than that in academic discourse.

Improving semantic coverage of data-to-text generation model using dynamic memory networks

AbstractThis paper proposes a sequence-to-sequence model for data-to-text generation, called DM-NLG, to generate a natural language text from structured nonlinguistic input. Specifically, by adding a dynamic memory module to the attention-based sequence-to-sequence model, it can store the information that leads to generate previous output words and use it to generate the next word. In this way, the decoder part of the model is aware of all previous decisions, and as a result, the generation of duplicate words or incomplete semantic concepts is prevented. To improve the generated sentences quality by the DM-NLG decoder, a postprocessing step is performed using the pretrained language models. To prove the effectiveness of the DM-NLG model, we performed experiments on five different datasets and observed that our proposed model is able to reduce the slot error rate rate by 50% and improve the BLEU by 10%, compared to the state-of-the-art models.

K‐Go: An executable formal semantics of Go language in K framework

AbstractGo is a relatively new programming language and has become one of the most important programming languages used widely in blockchain. A formal semantics plays an important role in program synthesis, analysis and verification. The existed formal semantics of Go are not executable and do not cover the core features, hence, this paper presents an executable formal semantics for Go, called K‐Go, using rewriting logic in framework to make sure the semantics is both executable and applicable. K‐Go includes syntax, configuration and rules and covers the declaration and definition of basic types, basic expression, concurrency, main statement data reading and writing based on channel. K‐Go is validated by test set with 100% semantic coverage based on the official test suite. The evaluation experiment shows that K‐Go is correct. Using formal semantics of Go language, building Go program synthesis, analysis and verification tools is easy for analyzing more characteristics of Go programs.

Cultural memory as a concept and phenomenon: the foundations of conceptualization

The article is devoted to cultural memory, on the one hand, as the most optimal concept denoting a socially determined form of supra–individual memory, and, on the other hand, as a theoretical construct (structural and functional model) denoted by this concept. Cultural memory is understood as a type of supra-individual memory that accumulates collective, value-significant memories, intentionally preserved and transmitted in mytho-symbolic forms. The main features of cultural memory include: conservatism, collectivism, presentism, symbolism, anti-historicity, written character, artificiality, institutionality, expert character, sacredness. The functional potential of cultural memory is realized in the following functions: accumulative, integrative, identification, interpretive, stabilization.The scientific novelty of the presented article is reduced to several positions. Firstly, to substantiate the concept of "cultural memory" as the most optimal (in comparison with the concepts of "historical memory", "social memory", "collective memory", etc.) to designate one of the types of supra-individual memory. The advantages of this concept are: broad semantic coverage, fixation of the collective, social and constructed nature of memories. Secondly, to conceptualize the phenomenon of "cultural memory", to determine its essential features and functional potential. As the main conclusion of the article, the following can be noted. In the context of the terminological and conceptual diversity of memory studies, it is extremely important to determine the categorical foundations of memorial research, fixing the content, structural, functional boundaries of the phenomena studied. In relation to supra-individual memory, this is especially important, mainly due to the presence of a very wide range of its conceptual and meaningful interpretation. In the article we offer our reasoned vision of the phenomenon of supra–individual memory, in particular one of its forms - cultural memory as a special reservoir of collective, social, value-significant memories, intentionally preserved and transmitted in mytho-symbolic forms.

Search and Learn: Improving Semantic Coverage for Data-to-Text Generation

Data-to-text generation systems aim to generate text descriptions based on input data (often represented in the tabular form). A typical system uses huge training samples for learning the correspondence between tables and texts. However, large training sets are expensive to obtain, limiting the applicability of these approaches in real-world scenarios. In this work, we focus on few-shot data-to-text generation. We observe that, while fine-tuned pretrained language models may generate plausible sentences, they suffer from the low semantic coverage problem in the few-shot setting. In other words, important input slots tend to be missing in the generated text. To this end, we propose a search-and-learning approach that leverages pretrained language models but inserts the missing slots to improve the semantic coverage. We further finetune our system based on the search results to smooth out the search noise, yielding better-quality text and improving inference efficiency to a large extent. Experiments show that our model achieves high performance on E2E and WikiBio datasets. Especially, we cover 98.35% of input slots on E2E, largely alleviating the low coverage problem.

An orthogonal-space-learning-based method for selecting semantically helpful reviews

• Selecting a helpful review set fits customers’ heterogeneous information demands. • We conceptualize review set helpfulness into credibility, relevance, diversity, and coverage. • We propose an orthogonal-semantic-space-based method for helpful review set selection. • Experiment results on designed metrics validate the helpfulness of the selected review sets. • Explanatory analysis offer insightful findings regarding the review set helpfulness. Customers in the online purchase context always bear the burden of enormous reviews. In this study, we endeavor to effectively identify and select a helpful review subset from original reviews to fit customers’ heterogeneous information demands for product evaluation. Rooted in theories about information quality, we first conceptualize review set helpfulness into four key factors, i.e., content credibility, semantic relevance, semantic diversity, and semantic coverage. We then propose a novel orthogonal-semantic-space-learning-based method that adaptively learns orthogonal semantic spaces based on original reviews and enables the generation of semantically credible, relevant, diverse, and high-coverage review subsets for varying review selection scales. Experiment results based on more than 1.7 million online reviews show that our proposed method outperformed benchmark review selection methods with improvements up to 2.07%, 7.9%, 5.52%, and 50% under semantic relevance, semantic diversity, semantic coverage, and review set helpfulness prediction, respectively. Explanatory analysis results offer insightful findings regarding helpful reviews at the review-set level.

Self-supervised Natural Image Reconstruction and Large-scale Semantic Classification from Brain Activity

Reconstructing natural images and decoding their semantic category from fMRI brain recordings is challenging. Acquiring sufficient pairs of images and their corresponding fMRI responses, which span the huge space of natural images, is prohibitive. We present a novel self-supervised approach that goes well beyond the scarce paired data, for achieving both: (i) state-of-the art fMRI-to-image reconstruction, and (ii) first-ever large-scale semantic classification from fMRI responses. By imposing cycle consistency between a pair of deep neural networks (from image-to-fMRI & from fMRI-to-image), we train our image reconstruction network on a large number of “unpaired” natural images (images without fMRI recordings) from many novel semantic categories. This enables to adapt our reconstruction network to a very rich semantic coverage without requiring any explicit semantic supervision. Specifically, we find that combining our self-supervised training with high-level perceptual losses, gives rise to new reconstruction & classification capabilities. In particular, this perceptual training enables to classify well fMRIs of never-before-seen semantic classes, without requiring any class labels during training. This gives rise to: (i) Unprecedented image-reconstruction from fMRI of never-before-seen images (evaluated by image metrics and human testing), and (ii) Large-scale semantic classification of categories that were never-before-seen during network training. Such large-scale (1000-way) semantic classification from fMRI recordings has never been demonstrated before. Finally, we provide evidence for the biological consistency of our learned model.

Visual Emotion Analysis via Affective Semantic Concept Discovery

With the development of social media, people prefer to express views and share daily life online via visual content, which has led to widespread attention in automatic emotion analysis from images. Capturing the emotions embedded in these social images has always been important yet challenging. In this paper, we propose a visual emotion prediction method that utilizes the affective semantic concepts of an image to predict its emotion. To solve the problems of narrow semantic coverage and low discriminative power of emotions in current semantic concept sets used for visual emotion analysis, we develop a concept selection model to mine emotion-related concepts from social media. Specifically, we propose several selection strategies to build an affective semantic concept set that contains various visual concepts related to emotion conveyance. And they are discovered from affective image datasets and associated tags crawled from websites. To further leverage the discovered affective semantic concepts, we train concept classifiers to predict the concept score of each concept, which are used as the intermediate features to tackle the semantic gap problem for image emotion recognition. Extensive experimental results confirm the validity of the affective semantic concepts and show the improved performance of our method.

AUTOMATIC ENRICHMENT OF INDOOR 3D MODELS USING A DEEP LEARNING APPROACH BASED ON SINGLE IMAGES WITH UNKNOWN CAMERA POSES

Abstract. 3D building modeling is a diverse field of research with a multitude of challenges, where data integration is an inherent component. The intensively growing market of BIM-related consumer applications requires methods and algorithms that enable efficient updates of existing 3D models without the need for cost-intensive data capturing and repetitive reconstruction processes. We propose a novel approach for semantic enrichment of existing indoor models by window objects, based on amateur camera RGB images with unknown exterior orientation parameters. The core idea of the approach is the parallel estimation of image camera poses with semantic recognition of target objects and their automatic mapping onto a 3D vector model. The presented solution goes beyond pure texture matching and links deep learning detection techniques with camera pose estimation and 3D reconstruction. To evaluate the performance of our procedure, we compare the estimated camera parameters with reference data, obtaining median values of 13.8 cm for the camera position and 1.1° for its orientation. Furthermore, a quality of 3D mapping is assessed based on the comparison to the reference 3D point cloud. All the windows presented in the data source were detected successfully, with a mean distance between both point sets equal to 3.6 cm. The experimental results prove that the presented approach achieves accurate integration of objects extracted from single images with an input 3D model, allowing for an effective increase of its semantic coverage.

ЛОГИКО-ПОНЯТИЙНАЯ КЛАССИФИКАЦИЯ ТЕРМИНОЛОГИИ СФЕРЫ «МЕДИЦИНА» В АНГЛИЙСКОМ, РУССКОМ И ТУРЕЦКОМ ЯЗЫКАХ

The original part of terminological units in comparison with the words of the national language lies in their internal semantic organization. In the case of terms, information about the logical generalizations of certain aspects and features inherent in classes of objects that are homogeneous in terms of these features comes to the fore. The article is devoted to revealing the semantics, in particular the logical and conceptual characteristics of the medical terminology in the English, Russian and Turkish languages. The purpose of the semantic analysis is to identify the conceptual structure characteristics of texts in a particular subject area. There was assumed the differentiation of terms by conceptual classes and the modeling of the term system by subclasses. As the analysis, medical articles with data from clinical trials were taken, terms of the “medicine” sphere in the presented languages ​​were selected by the continuous sampling method. The analysis of the given terminological sphere showed that terms related to the category of objects predominate in semantic coverage (112, which is 40% of all terminological units), terms related to the categories of states and processes are rare.

English Career Terms Synonymy

Synonymy is a linguistic phenomenon which reflects the complexity of reality representation in a linguistic worldview. Terminological synonymy poses certain difficulties for specialists who perform practical tasks in application areas such as translation and education. Career development , career advancement , career growth , career progression , career progress , career enhancement are among the most frequently used terms with similar semantics which indicates a certain degree of indefiniteness in the terminology in the correspondent professional field. The aim of this paper is to conduct comparative linguistic analysis of these terms. The material is presented by research article titles from eight scientific databases. Methods include quantitative and qualitative study of frequency, semantic fields by semantically related concepts, keywords, syntagmatic collocability and componential analysis. The results show that career development can be currently considered a hyperonym as it is the most frequent and time-proved term with the widest semantic coverage and indication to professional spheres in the dictionaries. Despite its obvious prevalence, career advancement , career progression , career growth and career progress are gaining momentum as alternative terms highlighting a lack of important semantic nuances in career development . Though the general number of career advancement and career progression across research papers is considerably lower, their presence in article titles is almost the same as that of career development . Title is arguably the most demonstrative part of a research paper; therefore, it can be presumed that the career terms might acquire lexical meanings which will clearly differentiate vertical career growth from horizontal organizational movement.

Unsupervised Derivation of Keyword Summary for Short Texts

Automatically summarizing a group of short texts that mainly share one topic is a fundamental task in many applications, e.g., summarizing the main symptoms for a disease based on a group of medical texts that are usually short, i.e., tens of words. Conventional unsupervised short text summarization techniques tend to find the most representative short text document. However, they may cause privacy issues, e.g., personal information in the medical texts may be exposed. Moreover, compared with the complete short text where some unimportant words may exist, a summary consisting of only a few keywords is more preferable by the user due to its clear and concise form. Due to the above reasons, in this article, we aim to solve the problem of unsupervised derivation of keyword summary for short texts. Existing keyword extraction methods such as Latent Dirichlet Allocation cannot be applied to solve this problem, since (1) the ordering relations among the extracted keywords are ignored, which causes troubles for people to capture the main idea of the event, and (2) short texts contain limited context, which makes it hard to find the optimal words for semantic coverage. Hence, we propose a simple but yet effective method named Frequent Closed Wordsets Ranking (FCWRank) to derive the keyword summary from a short text cluster. FCWRank is an unsupervised method that builds on the idea of frequent closed itemset mining in transaction database. FCWRank first mines all frequent closed wordsets from a cluster of short texts and then selects the most important wordset based on an importance model where the similarity between closed wordsets and the relation between the closed wordset and the short text document are considered simultaneously. To make the keywords within the wordset more understandable, FCWRank further unfolds the semantics behind them by sorting them. Experiments on real-world short text collections show that FCWRank outperforms the state-of-the-art baselines in terms of Recall-Oriented Understudy for Gisting Evaluation-Longest common subsequence F1, precision and recall scores.

Automated segmentation of RGB-D images into a comprehensive set of building components using deep learning

Building information modeling (BIM) has a semantic scope that encompasses all building systems, e.g. architectural, structural, mechanical, electrical, and plumbing. Automated, comprehensive digital modeling of buildings will require methods for semantic segmentation of images and 3D reconstructions capable of recognizing all building component classes. However, prior building component recognition methods have had limited semantic coverage and are not easily combined or scaled. Here we show that a deep neural network can semantically segment RGB-D (i.e. color and depth) images into 13 building component classes simultaneously despite the use of a small training dataset with only 1490 object instances. For this task, the method achieves an average intersection over union (IoU) of 0.5. The dataset was designed using a common building taxonomy to ensure comprehensive semantic coverage and was collected from a diversity of buildings to ensure intra-class diversity. As a consequence of its semantic scope, it was necessary to perform pre-segmentation and 3D to 2D projection as leverage for dataset annotation. In creating our deep learning pipeline, we found that transfer learning, class balancing, and prevention of overfitting effectively overcame the dataset’s borderline adequate class representation. Our results demonstrate how the semantic coverage of a building component recognition method can be scaled to include a larger diversity of building systems. We anticipate our method to be a starting point for broadening the scope of the semantic segmentation methods involved in digital modeling of buildings.

Measuring Sentences Similarity Based on Discourse Representation Structure

The problem of measuring similarity between sentences is crucial for many applications in Natural Language Processing (NLP). Most of the proposed approaches depend on similarity of words in sentences. This research considers semantic relations between words in calculating sentence similarity. This paper uses Discourse Representation Structure (DRS) of natural language sentences to measure similarity. DRS captures the structure and semantic information of sentences. Moreover, the estimation of similarity between sentences depends on semantic coverage of relations of the first sentence in the other sentence. Experiments show that exploiting structural information achieves better results than traditional word-to-word approaches. Moreover, the proposed method outperforms similar approaches on a standard benchmark dataset.

Towards an Ontology of Comparative Biogeography: New insights into the semantics of biodiversity conservation

Climate change, habitat destruction, and myriad other ecological stressors will impact us all and have already contributed to what is being labeled the sixth wave of extinction (Ceballos et al. 2015, Régnier et al. 2015). As a countering force, conservation biology strives to identify those areas of the planet most worthy of protecting due to their unique natural value (Dudley and Stolton 2008). Despite their value, criticisms (Camillo and Peter 2011) have been leveled at 1) the social cost of maintaining protected status (Lele et al. 2010) and 2) instances of continued biodiversity decline despite protection regimes (Craigie et al. 2010, Dudley et al. 2014). At present, the selection and delimitation of protected areas is an intuitive and often subjective process, leading to ambiguities in the semantics behind and across their definitions. Thus, we propose that the application of ontological techniques to the ambiguities in protected area semantics is a timely contribution to conservation informatics. We hold that coherent semantic representation of the biogeographic areas which overlap protected areas (often designated empirically) will provide more efficient and standardized informatics, supporting research and decision-making processes. Our approach draws from comparative biogeography, which seeks to classify biogeographic areas based on their natural properties in a process known as bioregionalisation. In particular, we convert a cladogram of biogeographic areas (similar to cladogram of taxa) into a series of ontological classes, each corresponding to a monophyletic clade of areas. In this model, areas of endemism are treated as formal objects related by hierarchical relationships and constrained by the monophyly condition (Ung 2018). This approach unifies a model-theoretic view of endemism with the semantic web and therefore, offering new possibilities to communicate the biogeographic units conservation. We use semantic web standards (RDF and OWL) expressed through interoperable "Open Biological and Biomedical Ontology (OBO) Foundry" and Library resources to model areas of endemism as evolutionary entities for comparative biogeography. This aligns with current efforts in the OBO Foundry to extend their semantic coverage to the domains of Earth and ecosystem science. Due to our work’s heavy reliance on environmental semantics, we base our work on the Environment Ontology (ENVO), extending it with often confounded biogeographic entities including biogeographic areas, such as areas of endemism and endemic areas, as well as their relationships. Hence, we seek to provide a rigorous and simple framework that improves biogeographic analyses and interoperability between systems.

SemIndex+: A semantic indexing scheme for structured, unstructured, and partly structured data

While Information Retrieval (IR) systems have gained success in Web-style search engines in the past two decades, nonetheless, the DataBase (DB) paradigm remains prevalent in handling data in enterprise environments and digital libraries, and is gaining even more importance in the Semantic Web with the increasing need to handle partly structured (NoSQL) data. This paper describes SemIndex+, a semantic-aware indexing and querying framework that allows semantic search, result selection, and result ranking of structured (relational DB-style), unstructured (IR-style), and partly structured (NoSQL) data. Various weighting functions and a parallelized search algorithm have been developed for that purpose and are presented here. We provide a general keyword query model allowing the user to choose the results’ semantic coverage and expressiveness based on her needs. Different from alternative solutions involving query relaxation, query refinement, or query disambiguation, our approach incorporates semantics at the most basic data indexing level: providing more opportunities toward speedups and semantic coverage. An extensive experimental evaluation, comparing SemIndex+ with alternative methods, highlights our approach’s flexibility and effectiveness, which in turn impact efficiency (requiring less or more time following the user specified index and query semantic coverages).