Textual Information Research Articles

Indoor Spatial Cognition for the Hearing/Visually Impaired: Google ARCore Augmented Interaction Using WiFi Map

<p dir="ltr"><span>Over 430 million people worldwide have disabling hearing loss, and 2.2 billion suffer from vision impairment. Assisting these populations with enhanced indoor navigation tools is critical for improving their independence and quality of life. This research advances existing indoor spatial cognition systems for individuals with hearing or visual impairments by developing two Java-based Android applications on the Android 10 operating system. The first application gathers the WiFi BSSIDs data via the Java Android class WifiManager to design a WiFi map inside the multistage building. In the second application, the WiFi map is employed to localize individuals with impairments and provide audio feedback for the visually impaired; for individuals with disabling hearing loss, the feedback is provided in a form of 3D GLB colored virtual object with textual information via the Google ARCore augmented reality library. Experimental results show a 100 % positioning accuracy in indoor localization at the multistage academic building at the University of Central Asia (Naryn campus, Kyrgyzstan).</span></p><div><span><br /></span></div>

Multimodal Contrastive Learning for Remote Sensing Image Feature Extraction Based on Relaxed Positive Samples.

Traditional multimodal contrastive learning brings text and its corresponding image closer together as a positive pair, where the text typically consists of fixed sentence structures or specific descriptive statements, and the image features are generally global features (with some fine-grained work using local features). Similar to unimodal self-supervised contrastive learning, this approach can be seen as enforcing a strict identity constraint in a multimodal context. However, due to the inherent complexity of remote sensing images, which cannot be easily described in a single sentence, and the fact that remote sensing images contain rich ancillary information beyond just object features, this strict identity constraint may be insufficient. To fully leverage the characteristics of remote sensing images, we propose a multimodal contrastive learning method for remote sensing image feature extraction, based on positive sample tripartite relaxation, where the model is relaxed in three aspects. The first aspect of relaxation involves both the text and image inputs. By introducing learnable parameters in the language and image branches, instead of relying on fixed sentence structures and fixed image features, the network can achieve a more flexible description of remote sensing images in text and extract ancillary information from the image features, thereby relaxing the input constraints. Second relaxation is achieved through multimodal alignment of various features. By aligning semantic information with the corresponding semantic regions in the images, the method allows for the relaxation of local image features under semantic constraints. This approach addresses the issue of selecting image patches in unimodal settings, where there is no semantic constraint. The proposed method for remote sensing image feature extraction has been validated on four datasets. On the PatternNet dataset, it achieved a 91.1% accuracy with just one-shot.

DRCmpVis: Visual Comparison of Physical Targets in Mobile Diminished and Mixed Reality.

Numerous physical objects in our daily lives are grouped or ranked according to a stereotyped presentation style. For example, in a library, books are typically grouped and ranked based on classification numbers. However, for better comparison, we often need to re-group or re-rank the books using additional attributes such as ratings, publishers, comments, publication years, keywords, prices, etc., or a combination of these factors. In this article, we propose a novel mobile DR/MR-based application framework named DRCmpVis to achieve in-context multi-attribute comparisons of physical objects with text labels or textual information. The physical objects are scanned in the real world using mobile cameras. All scanned objects are then segmented and labeled by a convolutional neural network and replaced (diminished) by their virtual avatars in a DR environment. We formulate three visual comparison strategies, including filtering, re-grouping, and re-ranking, which can be intuitively, flexibly, and seamlessly performed on their avatars. This approach avoids breaking the original layouts of the physical objects. The computation resources in virtual space can be fully utilized to support efficient object searching and multi-attribute visual comparisons. We demonstrate the usability, expressiveness, and efficiency of DRCmpVis through a user study, NASA TLX assessment, quantitative evaluation, and case studies involving different scenarios.

Визуализация точек многомерного информационного текстового массива на упругой карте для оценки кластерной структуры данных

The article presents the results of computational experiments on displaying the points of the original multidimensional information array on the elastic map scan to assess the relative positions of semantic proximity areas in order to improve the processing of text information. Elastic maps are considered as a tool for providing analytical work with text information. As previous works show, in order to obtain the required distances corresponding to the cluster picture of the studied multidimensional volume, it is necessary to use the distances on the elastic map, which reflects the cluster portrait of the studied multidimensional data volume. The paper presents the cluster structures of points of the studied multidimensional volume obtained in this way on the elastic map scan in the plane of the first two principal components. An analysis of the relative positions of clusters of different configurations at different points in time is presented.

Contextual Semantic Embeddings Based on Transformer Models for Arabic Biomedical Questions Classification

Arabic biomedical question classification (ABQC) is a challenging task due to various reasons including, the specialized jargon expressed in Arabic language, complex semantics of Arabic vocabulary and the lack of specific datasets and corpora. When representing questions, only a few studies deal with ABQC by taking into account the word context. In this work, we propose a classification model designed for Arabic biomedical questions. We build vector representations capturing the contextual and semantic information of Arabic biomedical text, which presents numerous challenges, such as the derivational morphology of Arabic language, the specialized terminology of biomedical terms and the lack of capitalization in text. Our representation adapts the extensive knowledge encoded in BERT (Bidirectional Encoder Representations from Transformers) and other transformer models, to address the aforementioned challenges. Several experiments have been conducted on a dedicated Arabic biomedical dataset namely: MAQA, with well-known transformer models including BERT, AraBERT, BioBERT, RoBERTa, and DistilBERT fine-tuned for the classification task. Obtained results show that our method achieves remarkable performance with an accuracy of 93.31% and an F1-score of 93.35%. Doi: 10.28991/HIJ-2024-05-04-011 Full Text: PDF

Utilizing Deep Learning for Field-Level Information Extraction from German Real Estate Tax Notices

ABSTRACT Document processing and related tasks such as information extraction represent a large portion of business workloads and therefore offer high potential for efficiency improvements and process automation. Although semistructured documents such as invoices and receipts have been extensively studied, tax notices represent a largely unexplored area due to their information richness and layout complexity. This article presents a study investigating the potential of deep learning-based information extraction for German real estate tax notices. We propose a hybrid approach based on the LayoutLM transformer model that uses textual, positional, and visual information as well as a rule-based information extraction. We show how such a system can be developed and integrated into the business workflow of organizations on the example of a German tax consulting firm. We also discuss the implications and key challenges that organizations face when planning to adopt such technologies. Data Availability: Custom data and code are not publicly available. Relevant public source code is cited in the text. JEL Classifications: M15; M40; M41.

Enhanced scene text recognition using deep learning based hybrid attention recognition network

<span lang="EN-US">The technique of automatically recognizing and transforming text that is present in pictures or scenes into machine-readable text is known as scene text recognition. It facilitates applications like content extraction, translation, and text analysis in real-world visual data by enabling computers to comprehend and extract textual information from images, videos, or documents. Scene text recognition is essential for many applications, such as language translation and content extraction from photographs. The hybrid attention recognition network (HARN), unique technology presented in this research, is intended to greatly improve efficiency and accuracy of text recognition in complicated scene situations. HARN makes use of cutting-edge elements including alignment-free sequence-to-sequence (AFS) module, creative attention mechanisms, and hybrid architecture that blends attention models with convolutional neural networks (CNNs). Thanks to its novel attention processes, HARN is capable of comprehending wide range of scene text components by capturing both local and global context information. Through faster network convergence, shorter training times, and better utilization of computing resources, the suggested technique raises bar for state-of-the-art. HARN’s versatility makes it a good choice for range of scene text recognition applications, including multilingual text analysis and data extraction. Extensive tests are conducted to assess the effectiveness of HARN approach and demonstrate it is ability to greatly influence real-world applications where accurate and efficient text recognition is essential.</span>

Machine Learning Algorithms for Emotion Recognition Using Audio and Text Data

Over the years, emotion recognition has become one of the cornerstones of affective computing, enabling machines to recognize and be responsive to human emotions. The current study will present approaches to accurately classify emotions by exploiting multimodal data, namely audio and text. Challenges faced in this domain are the noisy speech signals and inherently ambiguous textual expressions that generally reduce the accuracy of unimodal systems. Classic approaches cannot make good use of the complementary nature of these modalities and, therefore, require a robust and combined framework. This study proposes a method called SVM-ERATI, Support Vector Machine (SVM) based emotion recognition (ER) approach that inputs audio and text information (ATI). Extracted audio features in this regard will include Mel-frequency cepstral coefficients (MFCCs) and prosody-like pitch and energy related to the acoustic properties of emotions. Meanwhile, semantic embeddings obtained from transformer models like BERT serve to analyze text data. A feature-level fusion scheme is then followed, whereby the feature vectors from both audio and text are combined into an integrated representation. Then, features after fusion will be classified by the multi-class SVM with a proper radial basis function (RBF) kernel function that is most appropriate to capture the non-linear relationships inherent in the multimodal emotional data. Experiments on benchmark datasets such as CMU-MOSEI demonstrate that the proposed multimodal approach using SVM significantly outperforms unimodal baselines by about 12%. The findings highlight SVM's effectiveness in combining audio and text data for emotion recognition, which has exciting implications for AI in AI-powered mental health diagnostics and AI-powered intelligent virtual assistants.

Adaptive Graph Convolutional Network for the Document Prediction employing Dollmaker Optimization Algorithm

Document prediction is the application of advanced deep learning models in the analysis and prediction of the content of documents. Thus, it effectively improves text generation, information retrieval, and automatic summarization by changing the ways relating to textual data completely. This field is particularly challenging because maintaining high prediction accuracy requires efficient computing and scaling. Hence, this paper performs the document prediction using advanced deep learning-oriented optimization methodology. The data is first collected from online sources that consist of a group of documents together with their corresponding categories. The pre-processing of this collected data is next accomplished using the stop word removal, invalid character removal, and sentence segmentation approaches. The features are then extracted from these pre-processed data employing the Difference of Gaussian (DoG) method. The final prediction stage is done by the novel Adaptive Graph Convolutional Network (AGCN), in which the parameters are tuned by the well performing optimization algorithm known as Dollmaker Optimization Algorithm (DOA) with the consideration of error minimization as the fitness function. The findings demonstrated the superiority of the proposed model when it is compared with distinct conventional models.The proposed AGCN-DOA for the document prediction model in terms of prediction accuracy, MSE, MAE, MAPE, and RMSE is 11.18%, 94.73%, 31.71%, 47.73%, and 77.03% better than the considered existing methods respectively..

Development of a Voice Information Presentation Guidance System (VIPGS) Using Coanda‐Drone‐V2

AbstractIn this study, an aerial tour guide drone was developed based on the existing Coanda‐drone. The research integrated a Voice Information Presentation System with ROS2 using the open‐source PX4 flight control system to update the drone's control module. This system can recognize and read QR codes on displayed items and convert the text information from QR codes into speech for playback. It also utilizes left and right laser sensors for obstacle avoidance. Furthermore, the study conducted both simulation experiments and actual flight tests to validate the feasibility of equipping the Coanda‐drone with the voice information presentation system and to assess the flight capabilities of the Coanda‐drone. This system offers a novel solution for guidance and obstacle avoidance, demonstrating the potential of drones as flexible, highly secure, and efficient mobile service robots in the future. © 2024 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Photo‐Induced Fluorescence Discoloration of Carbon Dots for Advanced Dynamic Information Encryption

AbstractThe fluorescence discoloration of carbon dots (CDs) induced by UV irradiation without chemical modification has proved to be difficult to achieve. Herein, the synthesis of UV‐sensitive CDs with photo‐induced discoloration is reported using a heteroatom co‐doped hydrothermal method and the fluorescence of the red‐emitting CDs (R‐CDs) can be tuned under UV irradiation from red to yellow to green, showing a new strategy for simpler and easier modulating the luminescence of CDs. The UV‐induced fluorescence color change of CDs involves the changes of molecular fluorophores and oxygen defects on the surface of CDs under UV irradiation. Results show that the photo‐induced fluorescence discoloration of CDs results in the transformation of part of the internal groups, such as C─O and CO. Based on the unique UV‐induced fluorescence color‐changing property of the R‐CDs, a UV‐sensing film is fabricated, which is expected to be further applied in the field of UV detection. Furthermore, by combining CDs fluorescent ink with binary code and Morse code, it can encrypt and decrypt the information of specific text. This new concept of photo‐induced fluorescence discoloration in CDs provides strong technical support for the next generation of advanced information encryption.

The performance of the LSTM-based code generated by Large Language Models (LLMs) in forecasting time series data

Generative AI, and in particular Large Language Models (LLMs), have gained substantial momentum due to their wide applications in various disciplines. While the use of these game changing technologies in generating textual information has already been demonstrated in several application domains, their abilities in generating complex models and executable codes need to be explored. As an intriguing case is the goodness of the machine and deep learning models generated by these LLMs in conducting automated scientific data analysis, where a data analyst may not have enough expertise in manually coding and optimizing complex deep learning models and codes and thus may opt to leverage LLMs to generate the required models. This paper investigates and compares the performance of the mainstream LLMs, such as ChatGPT, PaLM, LLama, and Falcon, in generating deep learning models for analyzing time series data, an important and popular data type with its prevalent applications in many application domains including financial and stock market. This research conducts a set of controlled experiments where the prompts for generating deep learning-based models are controlled with respect to sensitivity levels of four criteria including (1) Clarify and Specificity, (2) Objective and Intent, (3) Contextual Information, and (4) Format and Style. While the results are relatively mix, we observe some distinct patterns. We notice that using LLMs, we are able to generate deep learning-based models with executable codes for each dataset separately whose performance are comparable with the manually crafted and optimized LSTM models for predicting the whole time series dataset. We also noticed that ChatGPT outperforms the other LLMs in generating more accurate models. Furthermore, we observed that the goodness of the generated models vary with respect to the “temperature” parameter used in configuring LLMS. The results can be beneficial for data analysts and practitioners who would like to leverage generative AIs to produce good prediction models with acceptable goodness.

Refreshable Braille Display System Driven by Accordion‐Fold Liquid Crystal Elastomer Actuators

AbstractRefreshable Braille display systems offer dynamic tactile approaches for the blind to obtain text information. However, existing refreshable Braille display systems using soft‐actuator‐driven Braille dots exhibit limitations of small strokes, complex mechanical structures, and high‐driving voltages, which inevitably lead to serious issues in safety, flexibility, and portability. To handle these issues, a novel refreshable Braille display system leveraging accordion‐fold liquid crystal elastomer actuators (AF‐LCEAs) is designed and created, which feature high stroke (1.47 mm), simple mechanical structure, and low driving voltage (20 V). The accordion‐fold structure with a high folding ratio of 59.7% not only enables significantly amplifying actuation strokes, but also brings a compact mechanical architecture for Braille dots. The structured‐LCE‐based refreshable Braille display system can be assembled modularly by multiple display units, which provides the capability to deliver rich and complex tactile information. In addition, a digital electro‐control matrix is designed to enable addressable, programmable electro‐actuation of the multiple LCE Braille dots to enable complicated refreshable tactile patterns. The Braille display system offers a versatile approach to engineering refreshable, portable, small‐sized Braille books useful for the blind.

An Empirical Exploration of the Effectiveness of Word Embedding Techniques for Emotion Recognition

Emotion recognition is an important research direction in the field of natural language processing. It is widely used in social media monitoring, product feedback analysis, and medical diagnosis, and it has significant practical value. The task of emotion recognition aims to recognize the emotional orientation of text through analysis. Word embedding technology is a key component in emotion recognition tasks; it helps the model capture sentiment relationships within the context by converting words into dense vectors, which is crucial for the model's performance. This paper is based on the Internet Movie Database (IMDB) and investigates the performances of different word embedding methods, including Rand, Static, Non-static, and Multi-channel, in sentiment recognition tasks. It also evaluates their effects on various neural network architectures. The experimental results show that random initialization performs best, especially in complex networks, demonstrating strong classification capability and training efficiency. This indicates that deep learning models have strong adaptive capabilities in sentiment recognition tasks. Even without the help of pre-trained word vectors, the model can still effectively capture the sentiment information in text.

Vision language model for interpretable and fine-grained detection of safety compliance in diverse workplaces

Workplace accidents due to personal protective equipment (PPE) non-compliance raise serious safety concerns and lead to legal liabilities, financial penalties, and reputational damage. While object detection models have shown the capability to address this issue by identifying safety gear, most existing models, such as YOLO, Faster R-CNN, and SSD, are limited in verifying the fine-grained attributes of PPE across diverse workplace scenarios. Vision language models (VLMs) are gaining traction for detection tasks by leveraging the synergy between visual and textual information, offering a promising solution to traditional object detection limitations in PPE recognition. Nonetheless, VLMs face challenges in consistently verifying PPE attributes due to the complexity and variability of workplace environments, requiring them to interpret context-specific language and visual cues simultaneously. We introduce Clip2Safety, an interpretable detection framework for diverse workplace safety compliance, which comprises four main modules: scene recognition, visual prompt, safety gear detection, and fine-grained verification. Scene recognition identifies the current scenario to determine the necessary safety gear. Visual prompt formulates specific visual cues needed for the detection process. Safety gear detection identifies whether the required safety gear is being worn according to the specified scenario. Lastly, fine-grained verification assesses whether the worn safety equipment meets the fine-grained attribute requirements. We conduct real-world case studies across six different scenarios. The results show that Clip2Safety not only demonstrates an accuracy improvement over state-of-the-art question–answering based VLMs but also achieves inference times that are 21× faster.

Analysis of Translation Strategies for Culturally Charged Terms in the French Version of Red Sorghum Clan

Mo Yan's Red Sorghum Clan is known for its unique style of language, in which a large number of words are loaded with cultural meaning. During the translation process, how to accurately convey the unique cultural information of the original text to the reader is a difficult task that translators often encounter. Based on Eugene Nida's method, this article will classify culturally loaded terms and then discover the translation strategies that two translators (Sylvie Gentil and Pascale Guinot) have used, with the aim of providing a reference for translators in translating Chinese literature and better promoting the dissemination of Chinese literary works.

A Brief Review of Multi-modal Knowledge Graph Completion

Knowledge graphs have become the infrastructure of artificial intelligence. However, most current knowledge graphs are incomplete. Consequently, knowledge graph completion (KGC) has become a hot research topic. Researchers primarily focus on unimodal knowledge graph completion, which consists solely of textual information. With the rapid progress of AI, the demand for multi-modal knowledge graphs (MMKGs) is increasing. However, research on multi-modal knowledge graph completion (MMKGC) is still in its initial stages. There is no clear recognition of its status and trends. First, we summarize the multi-modal knowledge graph and its significance. Second, we classify the comparisons between unimodal knowledge graph completion and MMKGC. Finally, we discuss different methods of MMKGC. This paper may provide guidance for future research.

COCONUT 2.0: a comprehensive overhaul and curation of the collection of open natural products database.

The COCONUT (COlleCtion of Open Natural prodUcTs) database was launched in 2021 as an aggregation of openly available natural product datasets and has been one of the biggest open natural product databases since. Apart from the chemical structures of natural products, COCONUT contains information about names and synonyms, species and organism parts in which the natural product has been found, geographic information about where the respective sample has been collected and literature references, where available. COCONUT is openly accessible at https://coconut.naturalproducts.net. Users can search textual information and perform structure, substructure, and similarity searches. The data in COCONUT areavailable for bulk download as SDF, CSV and a database dump. The web application for accessing the data is open-source. Here, we describe COCONUT 2.0, for which the web application has been completely rewritten, and the data have been newly assembled and extensively curated. New features include data submissions by users and community curation facilitated in various ways.

Is Innovation-Related Textual Information True? Insight from Equity Pledging Behavior in Chinese A-Share Listed Firms

Is Innovation-Related Textual Information True? Insight from Equity Pledging Behavior in Chinese A-Share Listed Firms

Feeling better when working with creative tools? Exploring the effects of a sketchnote intervention on students’ achievement emotions

AbstractSketchnoting is a visual note taking strategy that combines written notations with drawings. The potential is seen, in particular, in providing space for enjoyable learning. Since empirical evidence on how sketchnoting effects emotions is currently limited, the present study examined how students’ achievement emotions change after a sketchnoting intervention and how students visualize text information in sketchnotes. In a mixed methods intervention study with fifth grade students (N = 99 students, Mage = 10.04 years, SD = 0.39, 55.91% female) the sketchnote group used sketchnoting in five German grammar lessons, whereas the control group learned the grammar content through writing. Repeated ANOVA measures revealed that after the intervention, negative emotions significantly decreased within the sketchnote group, but not within the control group. In addition, a qualitative content analysis of the sketchnotes showed that students created visual analogies and examples, that connect abstract terms with daily life experiences. The results support the idea that sketchnoting can decrease students’ negative achievement emotions and provides space for linking new learning content to real-world experiences.