Browse Items (16481 total)
Sort by:
-
CRISPR plants against heavy metal and metalloid stresses: Methods and applications
Metal and metalloid stresses present substantial obstacles for plants, exerting an impact on agricultural productivity as well as environmental well-being. The assimilation and accumulation of heavy metals, for example, cadmium and lead, along with metalloids such as arsenic, has the potential to affect the growth and development of plants detrimentally. To address these stresses, innovative biotechnological approaches involving genetic engineering, clustered regularly interspaced short palindromic repeat (CRISPR) technology, phytoremediation, nanobiotechnology, and panomics are required to augment plant-metal tolerance. The dawn of CRISPR technology has revolutionized the field of plant biotechnology with its unparalleled precision in enhancing a plants capability to tolerate heavy metal and metalloid stresses. This present review aims to discuss the various applications and techniques of CRISPR technology in the enhancement of metal stress-tolerant capability in plants. It places special attention on the technologys critical role in abbreviating the detrimental effects of metal stress on plant growth and productivity. CRISPR technology enables researchers to engage in precise gene editing, thereby allowing for the deliberate targeting of specific genes associated with metal transport, detoxification, and stress responses in plants. By manipulating these key genes, CRISPR facilitates the creation of plant varieties that exhibit enhanced resilience to the challenges posed by heavy metal and metalloid stresses. In addition, this approach contributes to the advancement of agricultural sustainability and environmental stewardship by enhancing a plants ability to resist and overcome heavy metals contamination. Moreover, in the face of a changing climate landscape influenced by metal pollution challenges, the precise gene editing capabilities of CRISPR can be harnessed to engineer plants that possess heightened resilience to metal stress, thereby playing a vital role in ensuring food security and promoting sustainable agriculture. The review also underscores the pivotal role of CRISPR technology in shaping the future of research on plant stress tolerance and highlights its immense potential in addressing the evolving challenges associated with metal stress in plant systems. 2025 Elsevier Inc. All rights are reserved. -
CRISPR plants against heavy metal and metalloid stresses: Methods and applications
Metal and metalloid stresses present substantial obstacles for plants, exerting an impact on agricultural productivity as well as environmental well-being. The assimilation and accumulation of heavy metals, for example, cadmium and lead, along with metalloids such as arsenic, has the potential to affect the growth and development of plants detrimentally. To address these stresses, innovative biotechnological approaches involving genetic engineering, clustered regularly interspaced short palindromic repeat (CRISPR) technology, phytoremediation, nanobiotechnology, and panomics are required to augment plant-metal tolerance. The dawn of CRISPR technology has revolutionized the field of plant biotechnology with its unparalleled precision in enhancing a plants capability to tolerate heavy metal and metalloid stresses. This present review aims to discuss the various applications and techniques of CRISPR technology in the enhancement of metal stress-tolerant capability in plants. It places special attention on the technologys critical role in abbreviating the detrimental effects of metal stress on plant growth and productivity. CRISPR technology enables researchers to engage in precise gene editing, thereby allowing for the deliberate targeting of specific genes associated with metal transport, detoxification, and stress responses in plants. By manipulating these key genes, CRISPR facilitates the creation of plant varieties that exhibit enhanced resilience to the challenges posed by heavy metal and metalloid stresses. In addition, this approach contributes to the advancement of agricultural sustainability and environmental stewardship by enhancing a plants ability to resist and overcome heavy metals contamination. Moreover, in the face of a changing climate landscape influenced by metal pollution challenges, the precise gene editing capabilities of CRISPR can be harnessed to engineer plants that possess heightened resilience to metal stress, thereby playing a vital role in ensuring food security and promoting sustainable agriculture. The review also underscores the pivotal role of CRISPR technology in shaping the future of research on plant stress tolerance and highlights its immense potential in addressing the evolving challenges associated with metal stress in plant systems. 2025 Elsevier Inc. All rights are reserved. -
Upconversion nanoparticles for detection of small biomolecules and ions
In recent years, upconversion nanoparticles (UCNPs) have resulted in substantial advances in the area of sensitive and selective detection of small biomolecules and ions. UCNPs possess a unique optical property known as lanthanide upconversion luminescence. This phenomenon enables them to absorb low-energy light from the near-infrared region and subsequently emit higher-energy light in the visible or ultraviolet part of the spectrum. This process, often referred to as the anti-Stokes shift, resists the conventional fluorescence behavior by absorbing lower-energy photons, followed by emission of higher-energy photons. This chapter provides a comprehensive overview of the mechanism of upconversion fluorescence and explores the properties of UCNPs. It then delves into the applications of UCNPs in detection of biomolecules like proteins and amino acids, nucleic acids, and tumor biomarkers, thus facilitating early diagnosis and patient care. Additionally, UCNPs are useful in the detection of ions by altering their surface chemistry to bind selectively to target ions, expanding their utility in environmental monitoring and chemical analysis. 2026 Elsevier Ltd. All rights reserved. -
Pangenomics for developing salinity stress-tolerant plants
Soil salinity is a critical agricultural challenge that significantly reduces crop productivity and threatens global food security. With approximately 20% of irrigated land affected by salinity, innovative strategies are essential to develop salinity stress-tolerant crops. The field of pangenomics, a comprehensive approach to studying the genetic diversity within species, has immense potential to address this issue. Pangenomics includes core genomes, spanning the entire genus, and accessory genomes, which are species-specific, thus capturing the full spectrum of genetic variation. This approach enables the identification of novel genes and alleles associated with salinity tolerance, providing a robust foundation for genetic improvement programs. Salinity stress has a profound molecular and physiological impact on plants with multiple phenotypic manifestations, such as stunted growth, lesser crop yield, and reduced reproductive success. To solve these issues, advanced sequencing technologies and bioinformatics tools used in constructing and analyzing pangenomes play a crucial role. This chapter goes into detail about techniques such as comparative genomics and genome-wide association studies (GWAS), which are important for their effectiveness in identifying salinity tolerance genes. Functional validation methods, including CRISPR/Cas9 and RNA interference (RNAi), have also been discussed. This chapter highlights case studies on crops like rice and wheat to demonstrate the practical applications of pangenomics in developing salinity-tolerant varieties. Furthermore, by addressing the challenges and future directions in the field, one can emphasize the need for integrating multiomics data and refining analytical methods. Such an approach can help guide future research and breeding efforts toward sustainable agricultural practices and enhanced global food security. 2025 Elsevier Inc. All rights reserved. -
Pangenomics for developing viral disease-resistant plants: Harnessing synergenetic pangenomics for advanced viral disease resistance in crops
Agriculture plays a pivotal role in the global economy, but it also faces several significant challenges. One major challenge is the necessity to sustainably provide food for a rapidly expanding population. Additionally, agriculture must withstand the threat of different disease-causing pathogens, including various microorganisms like bacteria, viruses, and fungi. Viral diseases significantly impact crop growth, vitality, and yields, posing substantial threats to global agriculture. Despite their compact genomes, the viruses still lead to major losses in crop production. The study of plant viruses began over a century ago with the tobacco mosaic virus discovery, marking the inception of plant virology. Since then, hundreds of viruses have been identified, many of which induce severe damage to various crops. Viruses exhibit extensive genetic diversity, with RNA serving as the predominant genetic material. This genetic diversity profoundly influences their reproductive cycles and lifestyles, necessitating innovative strategies for disease management. This chapter discusses different pathogenic viruses along with various approaches to pest management, including the emergence of pangenomics as one of the advanced tools for understanding the diversity of genes among plant populations. Pangenomics encompasses both function-based and structure-based approaches, elucidating the core and dispensable genomes that contribute to discovering many agronomic traits related to disease resistance mechanisms, crop yield, flowering time, etc. Integrating the information from pangenomics into plant virology can revolutionize agriculture by enhancing crop production and crop yield, and also engineering viral disease-resistant plants to ensure global food security. 2025 Elsevier Inc. All rights reserved. -
AI- and ML-driven intelligent design of digital twins
Digital twins (DTs), or virtual copies of real-world systems, have changed and improved many industries in terms of monitoring, analysis, and optimization in real time. Artificial intelligence (AI) and machine learning (ML) together have significantly enhanced the functionalities of DTs so that they become more efficient and versatile decision-making and process improvement tools. The production and application of DTs most importantly rely on AI and ML. Such technologies allow integration and analysis of very large amounts of data from various sources and provide an overview of the physical system. The personnel involved in the company may gain deeper insights into overall business processes and identify changes that would remain unknown when applying the traditional methods, based on the employment of the capabilities of AI-based integration and data analysis. An essential example of ML use cases in the framework of DTs is predictive maintenance. Any ML algorithm can resort to historical data and immediate sensor data to predict potential failures of application equipment and propose a repair schedule, significantly reducing operational downtime and refining the distribution of resources. The AI-powered optimization and simulation methods can give organizations the possibility to consider numerous scenarios and identify the most effective ways to resolve complex issues. The DTs are AI-enabled and can detect and decide on the fly, which allows them to react to changing conditions instantly and prevent some of the issues before they happen. In addition, AI-powered predictive analysis and risk management allow the firms to go a step ahead and address the potential problems in advance by developing effective risk reduction strategies. DTs are mainly constructed with AI and ML in various industries. In the context of manufacturing and Industry 4.0, DTs play an important role in optimizing production and increasing the quality control standards. Urban planners use the DTs to strategize building smart cities, while healthcare professionals use them for medical diagnosis and planning. In the aerospace and auto industries, DTs are beneficial in improving the product development, testing, and other maintenance processes. This chapter focuses on the smart creation of DTs with the help of AI and ML technology. The discussion will also dive into the complex mechanism behind building advanced, digital replicas of physical systems, particularly the support of the AI and ML in the advancement of their usefulness and precision. The chapter begins with the discussion of the role of data integration and analysis in the creation of a DT. This section shows how AI and ML algorithms facilitate the seamless combination of different sources of data into one and reach a much more dynamic and detailed similitude of the physical member. The chapter illustrates how these technologies can convert raw information into valuable information, which makes the DT capable of replicating the real-world situations and behaviors quite dramatically. Moreover, the chapter addresses the profound role of AI and ML in the optimization and simulation of DTs. We observe how these advanced technologies are able to give more precise predictions and process the decision-making and testing of even complex scenarios. The chapter focuses on how AI-enabled optimization methodologies and AI-based simulations driven by ML are broadening the opportunities of DTs, thus driving innovation in a number of verticals. 2026 Elsevier Inc. All rights reserved. -
Introduction to multimodal learning and heterogenous data
With the rising advancements in computation, technology, and many innovative evolutions coming into play, multimodal learning is one of the most rapidly growing fields within the domain of artificial intelligence and Machine Learning. It mainly focuses on integrating information from multiple sources called "modalities," allowing the systems to utilize the varieties of data to furtherenhance their understanding and performance. These so-called modalities make use of various types of data in the form of text, images, audio, and sensor readings. They are able to process complex information due to these modalities and thus provide more insightful results for the tasks that they are assigned. Another important aspect of multimodal learning is heterogeneous datadata that differs significantly in structure, format, and origin. This type of data falls mainly into three categories, comprising structured data, which is quite organized and therefore easy to locate or search, as in the case of the database records. Then comes unstructured data characterized by the free form, which comprises mainly social media posts, videos, and images. In addition, it has been possible to separate semistructured data. It incorporates some features of being ordered, like the metadata included in XML or JSON files; however, a fixed schema does not apply. The understanding of the kind is important because each type calls for a different problem, and each type poses new opportunities in analysis. Handling the heterogeneous data effectively is all the more important because the said system will be fed heterogeneous data, and if its combination and analysis go reasonably logically, it is expected to be a source for multimodal systems. The ability to merge structured, unstructured, and semistructured data improves performance across a wide range of tasks, including but not limited to common applications like image recognition, sentiment analysis, and decision-making processes in autonomous systems. For example, in the multimodal learning case, it would be beneficial for the system that learns customer feedback to merge textual reviews, audio recordings of customer interactions, and visual data from product images. It has been known to yield a much clearer picture of what customers really want and how they actually behave. This chapter introduces notions of multimodal learning as well as heterogeneous datatheir characteristics, types, sources, and practical usage. It will attempt to establish a basic understanding of these two concepts in relation to each other in order to support more advanced applications through machine learning. In a review of the possible compositions between multimodal learning as well as heterogeneous data, the chapter will introduce their importance regarding the creation of intelligent systems that can address complex, intricate tasks across differing fields. As we enter the data age, with multiple sources churning out data at unprecedented rates that appear to have no bounds, integration of multimodal learning with heterogeneous data cannot be ignored. This will be vital for coming up with flexible yet useful applications to real-world problems. This region is promising for systems that perceive, interpret, and respond to the variability of information in a fashion similar to human reasoning and decision-making. Future application of artificial intelligence in the life of man will result from continuous research in the areas of multimodal learning and heterogeneous data. 2026 Elsevier Inc. All rights reserved. -
An introduction to multimodal data representation
The contemporary digital epoch is characterized by a radical transformation of data representation methodologies that imply increased intricacy as well as an enlarged bulk of data. An unimodal approach focusing on judicious data types, considered in isolation, was the earlier norm. The emphasis was on structured data, which had the advantage of being arranged systematically within relational databases and entity-relationship frameworks. This facilitated efficient data management. With the introduction of the internet and digital communication, such unstructured data as textual content, images, and audio began to be placed up front. But unimodal techniques were not adequately equipped to manage the intricate and interconnected nature of real-world phenomena. The welcome result was the development of multimodal data representation methodologies, which constitute a sophisticated paradigm that integrates data from such varied sources as text, images, audio, video, and sensor data. This results in a more holistic comprehension of complex scenarios. Distinct attributes and inherent challenges characterize each modality. To exemplify, text data need advanced natural language processing strategies to comprehend context and semantics; Image data necessitate methodologies well versed in managing spatial features and elevated dimensionality; audio data requires concentration on temporal patterns and noise; video data, on the contrary, integrates these complexities, leading to efficient processing techniques to accommodate its substantial volume and dynamic characteristics. The unsynchronous and heterogeneous sensor data complicate the integration of diverse data streams. Sophisticated fusion techniques, that is, early fusion, late fusion, and hybrid fusion, capable of integrating features from various modalities, are employed to mitigate the challenges faced by multimodal data representation. It increases interpretative insights and precision. The deep learning technologies, such as convolutional neural networks for image analysis, recurrent neural networks for sequential data processing, and attention mechanisms, have led to advancements in this domain. These models have become competent in recognizing complex patterns across modalities. Naturally, they bring about significant progress in domains such as health care, autonomous systems, multimedia processing, and natural language comprehension. This chapter explores the historical background of data representation, right from the beginnings in unimodal to its advancement in multimodal. The unique characteristics and challenges associated with each modality are scrutinized; Fusion techniques alongside contemporary deep learning models are examined; and underscore real-world applications, which are effective examples of the transformative potential of multimodal data representation. The chapter also emphasizes the necessity of escalating these methodologies in an increasingly data-centric world. It lays the foundation for advancements in the future with the goal of overcoming existing limitations and enlarging the scope of multimodal applications. 2026 Elsevier Inc. All rights reserved. -
Deep learning architectures for multimodal fusion
The advancement in technology during the recent years has provided deep learning technology as an emerging and powerful paradigm which can be used for processing and understanding complex data across various domains. Multimodal fusion is integrating the information that is collected from various sources or modalities, which requires a comprehensive understanding of data like autonomous driving, medical diagnosis, etc. In this chapter, we will explore the various advanced deep learning architectures that have been specially designed based on the multimodal fusion. The various challenges that are being faced in multimodal data, which include heterogeneity, noise, reliability of data, etc. Various deep learning architectures that are built to address the various challenges, like convolutional neural networks, recurrent neural networks, are reviewed, and the suitability of the fusion strategies is highlighted. The various techniques that are used for combining the information from disparate modalities, like early fusion, late fusion, and hybrid approaches, are also discussed with their pros and cons. Various real-time applications in the field of healthcare, multimedia, robotics, etc., are demonstrated based on the impact of the architectures. Finally, the potential of deep learning architecture based on the revolutionary multimodal fusion will be discussed. 2026 Elsevier Inc. All rights reserved. -
Multimodal learning for autonomous systems and robotics
The realm of autonomous systems and robotics is experiencing a paradigm shift driven by the integration of advanced artificial intelligence (AI) techniques and multimodal learning approaches. This abstract explores the latest advancements and research topics that are propelling the field toward more intelligent, efficient, and versatile autonomous systems. Multimodal learning leverages multiple sensory inputs to enhance the perception and decision-making capabilities of autonomous systems. This involves the integration of visual, auditory, tactile, and other sensory data to form a coherent understanding of the environment. Deep learning techniques, such as multimodal neural networks and crossmodal embeddings, play a pivotal role in this integration, enabling the system to learn joint representations and improve robustness in perception under varying conditions. Computer vision remains a cornerstone of autonomous systems, with advancements in techniques such as real-time object detection, tracking, and high-resolution image synthesis through generative adversarial networks. Vision-based reinforcement learning is also gaining traction, enabling systems to learn from visual inputs and improve their decision-making processes in dynamic environments. The integration of advanced sensors, including high-resolution light detection and ranging, radio detection and ranging, and event-based cameras, enhances the capability of autonomous systems to perceive their surroundings accurately. Multisensor data fusion, using methods like Kalman and particle filters, ensures robust perception even in adverse conditions, providing a comprehensive view of the environment. Innovations in actuation and control systems are fundamental for the development of responsive and adaptive robots. Soft robotics, inspired by biological systems, offers new possibilities in design, modeling, and control. Hybrid control systems facilitate the coordination of multimodal actuation, enhancing the robots versatility and performance. The deployment of high-performance embedded systems, incorporating heterogeneous computing architectures (CPU-GPU-FPGA integration), is vital for real-time data processing and decision-making. Neuromorphic computing and AI hardware accelerators provide low-power solutions that are crucial for the efficiency of autonomous systems. Techniques for uncertainty estimation, outlier detection, and anomaly detection are essential for maintaining system reliability. Advanced robotic perception and cognition, combined with cognitive architectures for autonomous reasoning, enable systems to operate safely in complex and dynamic environments. The interface between humans and robots is evolving, with a focus on multimodal human-robot interaction. Learning from human demonstrations and ensuring safety and trust in human-robot teams are critical areas of research, promoting effective collaboration between humans and robots. Advanced simulation techniques, including high-fidelity physics-based simulations and domain randomization, are employed to test and validate autonomous systems. Virtual reality and augmented reality provide immersive environments for training and testing. Real-time simulation and hardware-in-the-loop testing ensure the robustness and reliability of autonomous systems before deployment. Ethical AI and autonomous decision-making frameworks are being developed to address these issues. Privacy-preserving machine learning techniques and cybersecurity measures are essential for protecting sensitive data and ensuring the security of autonomous systems. This comprehensive overview underscores the rapid advancements and multifaceted nature of multimodal learning and autonomous systems, heralding a new era of intelligent and adaptive robotics capable of transforming numerous industries and improving the quality of human life. 2026 Elsevier Inc. All rights reserved. -
Cognitive computing: merging modalities for human like artificial intelligence
Artificial intelligence (AI) has transformed with cognitive computing, which integrates different modalities such as vision, language, and reasoning to mimic human thought processes. Cognitive computing systems can recognize patterns, understand natural language, and make complex decisions by integrating these modalities. As a result of the interaction between machines, neural networks, and natural language processing, large amounts of data are analyzed, learning continuously from their interactions. As a result of this multimodal integration, the system is able to comprehend and interpret human communication with greater accuracy and context knowledge. Cognitive computing has made significant advances by processing and synthesizing information from a variety of sources, enabling a more holistic understanding of complicated situations. For instance, these systems can better understand context when visual data is combined with textual information. Using cognitive computing to analyze medical images, patients' histories, and clinical notes, healthcare professionals can find it especially helpful in diagnosing patients and planning treatment. As a result of cognitive computing, we can create systems that not only perform specific tasks but also mimic human cognitive functions, so that we can make better decisions. Human-like AI has the potential to revolutionize various industries with intelligent, context-aware solutions that increase productivity and decision-making. The advancement of cognitive computing has made it possible for humans and machines to communicate more intuitively and efficiently, bridging the gap between them. 2026 Elsevier Inc. All rights reserved. -
Case studies: multimodal applications in natural language processing
This chapter explores the incorporation of natural language processing (NLP) with multimodal information sources, including text, speech, and visual information, towards the improvement of practical applications. NLP may very significantly enhance tasks such as sentiment analysis, image captioning, and cross-modal retrieval by combining these modalities. Two examples of deep learning approaches are neural networks and transformers, which are examples of critical approaches for developing robots that analyze and understand complex multimodal inputs. The chapter is full of case examples that illustrate how multimodal NLP can revolutionize many industries, including healthcare data analysis and voice-activated assistant development. These illustrations demonstrate how NLP can enhance user interactions and decision-making processes by offering deeper, more contextual insights. In fact, the chapter also covers issues and ways ahead of multimodal NLPas integrating data, handling faulty or missing data, and how to resolve ethical dilemmas. These ongoing changes will define future artificial intelligence systems with increased adaptability, intuitiveness, and applicability. 2026 Elsevier Inc. All rights reserved. -
Visual-audio fusion in multimedia content analysis
The analysis of multimedia material has grown in importance due to the rapid expansion of digital media. Exploiting the combination of auditory and visual modalities for improved comprehension and interpretation of multimedia material has gained popularity in recent years. An extensive review of the methods, strategies, and uses of visual-audio fusion in multimedia content analysis is provided in this chapter. Combining auditory and visual modalities provides a number of benefits over single-modal analysis, such as increased robustness, deeper semantic comprehension, and better user experience. This chapter investigates a variety of fusion approaches, from late combination at the decision near to early synthesis at the feature nearby. Besides, it studies sophisticated fusion methods that allow for efficient integraton of data across modalities, such cross-modal attention processes and multimodal deep learning. Also looks at several other areas, such as multimedia retrieval, event detection, sentiment analysis, and emotional computing, where visual-audio fusion has been used successfully. It dialogs about the difficulties and potential paths ahead for the area, including how to deal with modality inconsistencies, manage massive multimedia information, and create fusion models that are understandable. To sum up, visual-audio fusion presents new possibilities for comprehending and analyzing complicated multimedia data and has the potential to significantly advance multimedia content analysis. 2026 Elsevier Inc. All rights reserved. -
Future trends in multimodal learning: from theory to practical applications
The human ability to seamlessly integrate information from various sensory channelssight, sound, touchhas long inspired researchers in artificial intelligence. This ability to learn from and reason with multimodal datatext, speech, vision, and moreforms the core of multimodal learning. This abstract delves into the theoretical foundations of multimodal learning, explores its cutting-edge advancements, and critically examines the path toward practical applications in diverse fields. At its heart, multimodal learning seeks to exploit the inherent complementarity between different data modalities. Text, for instance, provides rich semantic meaning, while visual data offers valuable context. Speech captures the nuances of emotion and prosody often absent in text. By learning from these combined modalities, models can achieve a more comprehensive understanding of the world around them. Recent years have witnessed significant progress in multimodal learning architectures. Deep learning approaches, particularly convolutional neural networks and recurrent neural networks, have proven adept at capturing complex relationships within individual modalities. New architectures like multimodal transformers further bridge the gap by allowing models to learn joint representations across different modalities. These advancements pave the way for a paradigm shift in areas like computer vision, natural language processing, and robotics. In computer vision, multimodal learning allows models to not only recognize objects in images but also understand the context and actions depicted. By incorporating textual descriptions or speech narratives alongside visual data, models can achieve better scene understanding, image captioning, and action recognition. This has applications in autonomous vehicles, where understanding traffic signs, pedestrians, and road conditions is crucial, and in video surveillance systems, where interpreting visual cues alongside spoken dialogue improves anomaly detection. 2026 Elsevier Inc. All rights reserved. -
Fusion Techniques for medical imaging and clinical data towards precision diagnostics and personalized care
The combination of clinical data with medical imaging has created a revolution in modern healthcare, which provides a clear insight into patient's health sometimes in an earlier stage itself. Magnetic resonance imaging, computed tomography scans, ultrasounds, and X-rays are a few of the medical imaging techniques that offer high-resolution representations of the body's structure and physiological processes. Clinical data, such as physical examinations, test findings, and medical histories, help to contextualize these photographs, which enhances treatment planning and broadens diagnostic discoveries. Fusion approaches combine data from multiple sources to create a unified dataset, making diagnoses more accurate and treatments more personalized. This chapter emphasizes the importance of the fusion of heterogeneous medical data along with various fusion techniques, including deep learning and attention mechanisms, to align medical images with clinical data for meaningful insights. By using fusion techniques, healthcare professionals can make real-time decisions, identify diseases with better accuracy, and derive insights that can lead to actionable treatment or management strategies. Advanced fusion methods enable healthcare providers to obtain a holistic view of a patient's health, allowing advancements in precision medicine and customized treatment plans. 2026 Elsevier Inc. All rights reserved. -
Ethical considerations in multimodal data collection and analysis
Nowadays, research and industry rely more on collecting and analyzing multimodal data that integrates a host of formats like text, images, audio, and video. Such integration increases the decision-making capabilities and builds insightful information, but equally raises serious ethical issues to be followed up with caution. In multifarious and interrelated datasets, questions about ownership of the data, informed consent, and privacy become much more complex. This has further worsened by the advent of social media and big data analytics, exposing participants to possible harms. This paper discusses the ethical issues arising in such scenarios and calls for strong mechanisms that ensure responsible and just conduct in multimodal research. 2026 Elsevier Inc. All rights reserved. -
Modalities in data: understanding text, images, and audio
Data modalities, encompassing diverse forms such as text, audio, image, and video, play a pivotal role in shaping modern data analysis and machine learning applications. Each modality represents information in a unique format, requiring specific processing and interpretation methods. The integration of multiple modalities, known as multimodal data, enhances decision-making and predictive accuracy, particularly in complex systems like sentiment analysis, speech recognition, and medical diagnostics. Deep learning techniques have facilitated the seamless fusion of multimodal data, enabling a more comprehensive understanding across various fields, from healthcare to social media analytics. For example, combining text with images improves sentiment analysis, while integrating audio and video aids in more accurate speech recognition. However, the incorporation of multimodal data presents challenges, including data heterogeneity, synchronization issues, and dimensionality concerns. Data formats differ across modalities, and aligning them for cohesive analysis requires sophisticated algorithms and computational power. Despite these obstacles, multimodal data offers significant benefits, such as enhanced customer experience in business and increased diagnostic accuracy in health care. Furthermore, the rise of large datasets and artificial intelligence (AI) technologies has fueled innovation, enabling the development of more efficient models capable of uncovering intricate relationships within data. This chapter discusses various modalities, their applications, and the technological advancements driving their integration. It also highlights the challenges in multimodal data processing and the solutions being developed to address these complexities, offering valuable insights for businesses, researchers, and AI practitioners. 2026 Elsevier Inc. All rights reserved. -
Feature extraction and fusion techniques for multimodal data
Integrating multimodal information has become crucial in the big data era for developing a thorough knowledge of complex systems and enhancing decision-making in a variety of fields. The importance of feature extraction and fusion strategies in multimodal learning is examined in this chapter, with particular attention paid to the difficulties and approaches involved in merging several data modalities, including text, pictures, audio, and sensor data. It talks about how conventional feature extraction approaches have evolved into more sophisticated ones like deep learning models for picture and audio data and neural embeddings for text. The chapter also explores several fusion tactics, such as early, late, and intermediate fusion, and focuses on how they are used in domains including sentiment analysis, autonomous cars, healthcare, and multimodal search engines. The chapter highlights future directions, such as lightweight architectures and privacy-preserving techniques, while also addressing contemporary issues, such as managing missing data, scalability, and privacy concerns. The chapter provides a thorough grasp of how feature extraction and fusion aid in the creation of multimodal systems that are more precise, effective, and interpretable by looking at these factors. 2026 Elsevier Inc. All rights reserved. -
Transfer learning in multimodal settings
A powerful machine learning technique in the multimodal environment allows the transmission learning model to adapt to information from one domain to another, which promotes more effective learning in different types of data, including lessons, images, speeches, speeches, and sensor data. This method increases the model's adaptability, reduces the requirement for large marked datasets, and increases performance across domains. It has been used in several domains where multimodal integration is important, such as healthcare, autonomous systems, and natural language processing. Despite the benefits, transmission learning has disadvantages, including high data costs, data shortages, and domain changes. To meet these challenges, model architecture, adaptation strategy, and improvement in dataset growth techniques are necessary. This study examines basic ideas, procedures, and transfer of transfer to multimodal references, and provides insight. Practical use and new development. We show the developing role to learn transfer in improving artificial intelligence (AI) applications by looking at current studies and case studies. As the area develops, a combination of knowledge from many methods will be necessary to create scalable, reliable, and effective AI systems that can handle the problems in the real world. 2026 Elsevier Inc. All rights reserved. -
Challenges in preprocessing and normalization of heterogenous data
In todays information-driven landscape, the exposure to information from a wide range of sources, such as social media, financial transactions, healthcare records, and Internet of Things (IoT) sensors. This variety, while providing valuable insights, also brings significant issues. Heterogeneous information, which varies in formats, structures, and scales, requires careful management to ensure it can be effectively used in analytics and machine learning. Key steps in this process include prehandling and standardization. Prehandling involves cleaning and preparing the information by tackling issues like missing values, identifying and eliminating noise (inaccurate or irrelevant information), and addressing inconsistencies. Normalization, in contrast, converts the information into a uniform format and scale, facilitating easier comparison and analysis. However, managing diverse information effectively comes with several issues. Missing information is a frequent problem, and accurately filling in these gaps can be complicated, especially for complex information types. Noise and inconsistencies can greatly affect the accuracy and reliability of any analysis that follows. Additionally, merging information from various sources with differing formats and structures can be a challenging and time-consuming task. This chapter explores the specific issues faced when prehandling and normalizing different information types, including numerical, categorical, textual, and image information. Real-world examples from India, such as the Aadhaar information base and IoT-enabled smart cities, highlight the practical implications of these issues. By grasping best practices and emerging AI-driven trends, organizations can improve information reliability and enhance decision-making. 2026 Elsevier Inc. All rights reserved.