General Language Research Articles

AI-Generated Object: Is it a ‘Work’ and Who Owns it?

The evolution of generative AI and the popularity of diffusion models have opened up new opportunities for creativity, but have also raised several legal challenges in the field of copyright. This concerns both the use of copyrighted works in databases for training algorithms and the legal protection of the generated results. The generation of new text, video, audio objects, software codes, and synthetic data is directly connected to the legal terminology, such as ‘author’, ‘work’, ‘originality’ and ‘copyright protection’. In practice, questions arise as to whether AI-generated creations meet the requirements of originality, who should be considered as the author of such an object, and so on.Therefore, the aim of this article is to assess the impact of the development of generative AI on today’s copyright law and to answer whether objects created by generative AI should be granted copyright protection. The paper assesses whether the objects created by generative AI are considered ‘work’ in the sense of the Lithuanian (EU) copyright law and discusses the current issues of the authorship and ownership of such objects.

Enhancing Persona Creation Through the Hermeneutic Fusion of Horizons: Business Applications in AI Natural Language Generation

Enhancing Persona Creation Through the Hermeneutic Fusion of Horizons: Business Applications in AI Natural Language Generation

Region-guided transformer for remote sensing image captioning

ABSTRACT Remote sensing image acquisition is an essential way to obtain information. However, research on remote sensing images mainly focuses on object detection or image classification. The emergence of remote sensing image captioning (RSIC) has enabled understanding and inference of remote sensing images, thus attracting considerable attention. There are still challenges in RSIC: the features used in RSIC are mostly based on grid features, and this form of features makes it difficult for the model to determine the main description targets. Hence, a more effective cross-modal matching method is needed for better text generation. Thus, we propose a region-guided transformer in response to the aforementioned issues. We extracted region features to enhance the ability of the model to focus on the main targets. To address the issue of information loss caused by region feature extraction, we proposed environment features to supplement background information. To improve the matching between text and image features, we propose a region-guided decoder that enhances the model's perception of different features through a weighted cross-attention mechanism. Meanwhile, we introduce region-guided information to guide the text-generation process. The effectiveness and superiority of our model have been demonstrated through extensive experiments.

Research on the role of LLM in multi-agent systems: A survey

In recent years, the rapid development of large language model (LLM) has demonstrated superior performance in language understanding, text generation, planning, reasoning, and knowledge integration. This has led to the emergence of intelligent agents based on LLM. By leveraging the capabilities of LLM, these agents can effectively make decisions based on given objectives and possess certain learning and adaptation abilities. However, single-agent systems are generally suited to solving relatively simple problems and are limited in handling complex tasks that require coordination. For instance, in fields such as power grid management or traffic control systems, relying solely on a single agent is often insufficient for effective decision-making. In this context, adopting multi-agent systems proves to be more effective: through collaboration among multiple agents, each undertaking specific tasks, complex problems can be efficiently managed through interaction and coordination. This survey will analyze the role of LLM in multi-agent collaboration, discuss and analyze the current research challenges and key issues, and explore potential directions for future development.

Research on a New Model of Cultivating Creative Thinking in the Language Discipline Based on ChatGPT--Taking Project-based Learning of Senior Secondary Language as an Example

This paper This paper discusses a new model of cultivating creative thinking in the language discipline based on ChatGPT, aiming to provide new ideas and methods for language teaching reform. ChatGPT, an advanced natural language processing technology with powerful language comprehension and generation capabilities, provides new means and methods for language teaching. The study presents a case study of the teaching practice of the new model of creative thinking cultivation in language discipline based on ChatGPT through the casestudy method, which provides guidance for the further improvement and optimization of the new model. The model takes ChatGPT as the core, combines the characteristics of language teaching and students' needs, and designs teaching processes and activities suitable for cultivating students' creative thinking. The study emphasizes the importance of changing teaching concepts when using the Humanistic Version of the senior secondary language textbook, as well as incorporating practical language activities to improve students' core literacy through personalized participation and practice.

Improving prediction performance of general protein language model by domain-adaptive pretraining on DNA-binding protein

DNA-protein interactions exert the fundamental structure of many pivotal biological processes, such as DNA replication, transcription, and gene regulation. However, accurate and efficient computational methods for identifying these interactions are still lacking. In this study, we propose a method ESM-DBP through refining the DNA-binding protein sequence repertory and domain-adaptive pretraining based the general protein language model. Our method considers the lacking exploration of general language model for DNA-binding protein domain-specific knowledge, so we screen out 170,264 DNA-binding protein sequences to construct the domain-adaptive language model. Experimental results on four downstream tasks show that ESM-DBP provides a better feature representation of DNA-binding protein compared to the original language model, resulting in improved prediction performance and outperforming the state-of-the-art methods. Moreover, ESM-DBP can still perform well even for those sequences with only a few homologous sequences. ChIP-seq on two predicted cases further support the validity of the proposed method.

Efficient Headline Generation with Hybrid Attention for Long Texts

Headline generation aims to condense key information from an article or a document into a concise one-sentence summary. The Transformer structure is in general effective for such tasks, yet it suffers from a dramatic increase in training time and GPU consumption as the input text length grows. To address this problem, a hybrid attention mechanism is proposed. Both local and global semantic information among words are modeled in a way that significantly improves training efficiency, especially for long text. Effectiveness is not sacrificed; in fact, fluency and semantic coherence of the generated headlines are enhanced. Experimental results on an open benchmark dataset show that, compared to the baseline model’s best performance, the proposed model obtains a 14.7%, 16.7%, 14.4% and 9.1% increase in the F1 values of the ROUGE-1, the ROUGE-2, the ROUGE-L and the ROUGE-WE metrics, respectively. The semantic coherence of the generated text is also improved, as shown by a 2.8% improvement in the BERTScore’s F1 value. These results show that the effectiveness of the proposed headline generation model with the hybrid attention mechanism is also improved. The hybrid attention mechanism could provide references for relevant text generation tasks.

Design and Implementation of an Interactive Question-Answering System with Retrieval-Augmented Generation for Personalized Databases

This study introduces a novel approach to personalized information retrieval by integrating retrieval augmentation generation (RAG) with a personalized database system. Recent advancements in large language models (LLMs) have shown impressive text generation capabilities but face limitations in knowledge accuracy and hallucinations. Our research addresses these challenges by combining LLMs with structured, personalized data to enhance search precision and relevance. By tagging keywords within personal documents and organizing information into context-based categories, users can conduct efficient searches within their data repositories. We conducted experiments using the GPT-3.5 and text-embedding-ada-002 models and evaluated the RAG assessment framework with five different language models and two embedding models. Our results indicate that the combination of GPT-3.5 and text-embedding-ada-002 is effective for a personalized database question-answering system, with potential for various language models depending on the application. Our approach offers improved accuracy, real-time data updates, and enhanced user experience, making a significant contribution to information retrieval by LLMs and impacting various artificial intelligence applications.

The art of prompts' formulation: limitations, potential, and practical examples in large language models

Introduction: "prompt engineering" is crucial in the use of AI models like GPT-3 and GPT-4, as it helps obtain effective responses in areas such as text generation and programming. A well-crafted prompt improves the quality of the responses. The study analyzed how LLMs function and gathered advice for prompt engineering, also examining technological limitations and the impact of user language. Methods: the evolution of large language models, from recurrent neural networks (RNN) to the introduction of the Transformer architecture in 2017, is explained. Responses from ChatGPT 3.5 and 4.0 were evaluated in two case studies to analyze the complexity and personalization of the prompts. Results: in the case studies, it was found that adding context and specificity improved the models' responses. Detailed and personalized responses resulted in greater accuracy and relevance. Conclusion: the quality of LLM responses depends on the precision and specificity of the prompts. Personalization and appropriate technical language enhance interaction with Artificial Intelligence (AI), increasing user satisfaction. Future studies should analyze semantic fields and metrics to evaluate the quality of AI-generated responses.

Character-Level Text Generation for Shakespearean Style with LSTMs

This paper presents a pioneering approach to text generation employing Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM) architecture, inspired by the rich and timeless prose of William Shakespeare. The motivation stems from the enduring allure of Shakespearean language, which has captivated audiences across centuries, and the challenge of replicating itsintricate style using modern computational techniques. Our research contributes a novel methodology that leverages the capabilities of RNN LSTM networks to emulate the linguistic nuances of Shakespeare with remarkable fidelity. The paper begins by providing a comprehensive overview of RNN LSTM networks, highlighting their suitability for sequential data processing tasks and their ability to capture long-rangedependencies. A review of related work in the field sets the stage for our proposed approach, shedding light on recent advancements and methodologies employed in text generation using similar techniques. We formulate the problem by defining the mathematical framework, optimization objectives, and evaluation metrics for our proposed model. The architecture consists of three layers: the data layer for preprocessing input text data, the intelligence layer comprising multiple LSTM units for capturing different aspects of Shakespearean language, and the application layer for generating output text based on learned representations. Experimental results demonstrate the effectiveness of our approach, with evaluations conducted on a corpus of Shakespearean texts.In conclusion, our research presents a significant advancement in the field of natural language generation, opening new avenues for exploring the intersection of literature and artificial intelligence.

Teaching in the Era of Artificial Intelligence: Reimagining Activities and Assignments in Preservice Special Education Teacher Education Programs

Artificial intelligence (AI) text generators, such as ChatGPT and Copilot, have been used for various purposes such as creating written content, writing or debugging computer code, answering questions, providing information, and improving written communication. This robust functionality of AI text generators along with the high rate of use reported by college students highlights the need for instructors in special education teacher preparation programs to contend with the impact of AI on teaching and learning. The purpose of this article is to provide examples of how teacher educators might reimagine assignments and activities in the era of AI text generators. Specifically, the authors explore ways that AI applications might be intentionally incorporated or, conversely, intentionally restricted in the design of course assignments and activities.

Advanced Computational Methods for News Classification: A Study in Neural Networks and CNN integrated with GPT

In an era inundated with vast amounts of information, the imperative for efficient news classification is paramount. This research explores the sophisticated integration of neural networks and convolutional neural networks (CNN) with Generative Pre-trained Transformers (GPT) to enhance the precision and efficacy of news categorization. The rapid digital dissemination of news necessitates advanced computational methodologies capable of accurate classification and event prediction that include finance and economic events. Leveraging recent advancements in machine learning and natural language processing (NLP), this study utilizes large language models (LLMs) such as GPT and BERT, known for their exceptional comprehension and generation of human-like text. Over 232 days, our methodology classified 33,979 news articles into Education & Learning, Health & Medicine, and Science & Technology, with further subcategorization into 32 distinct subcategories. For evaluation, a sample of 5,000 articles was assessed using metrics such as True Positive (TP), True Negative (TN), False Positive (FP), False Negative (FN), Precision, Recall, and F1-Score. In comparison with the existing studies, the proposed method achieving significantly higher with average scores of 0.986 (Precision), 0.987 (Recall), and 0.987 (F1-Score). This research offers substantial practical contributions, providing detailed insights into news source contributions, effective anomaly detection, and predictive trend analysis using neural networks. The theoretical contributions are profound, demonstrating the mathematical integration of GPT with CNNs and recurrent neural networks. This integration advances computational news classification and exemplifies how sophisticated mathematical frameworks enhance large-scale text data analysis, marking a pivotal advancement in applying advanced computational methods in real-world scenarios.

Application of text mining in analysing notes to financial statements: A Hungarian case

Company stakeholders must have reliable and accurate information about the companies falling into their sphere of interest. In Hungary, one of the key sources of information for company stakeholders is the financial statements and related explanations, which are included in the notes of the financial statements (notes). This study used text mining to analyse the Hungarian annual financial statements notes for 2017, 2019 and 2021. The selection of the notes was based on the proportions of each sector in the national economy. The research analysed 28,700 company notes annually, totalling 86,100 documents for the three years. The text mining and generation of the Term Frequency Matrix have performed 'quanteda' packages of the R statistical system, which incorporate the results of artificial intelligence research to enhance the efficiency of text mining. Based on the results, the contents of the notes to the financial statements appear to be a rather mixed picture in Hungary. Analysing the term frequency matrix for the 67 most common terms has revealed no significant difference between the years. However, considerable differences have been caused by size categories and sectors. The notes are statistically significant using Jaccard similarity analysis, considering the year, corporate size, and sector.

IMPLEMENTASI PROGRAM ONE PESANTREN ONE PRODUCT ( OPOP) SEBAGAI UPAYA PENINGKATAN PEMBERDAYAAN EKONOMI PESANTREN

Islamic boarding schools, known as Pondok Pesantren, no longer solely focus on Islamic religious education but have also expanded into various fields such as general knowledge, languages, technology, and agribusiness. Economic empowerment has become a crucial aspect of the development of modern pesantren, enhancing not only financial independence but also empowering the community through various Shariah-compliant economic initiatives. One program that supports these efforts is One Pesantren One Product (OPOP). This research aims to understand the implementation, processes, and outcomes of economic empowerment at Pondok Pesantren Al-Hidayah through the implementation of the One Pesantren One Product (OPOP) program. This program has become a significant initiative in developing the economic potential of pesantren and the independence of the students in Garut. The study focuses on the economic empowerment implementation at Pondok Pesantren Al-Hidayah Garut through the OPOP program. The research method used is descriptive study with a qualitative approach. Data were collected through observation, interviews, and documentation studies. The research findings indicate that the One Pesantren One Product program has successfully enhanced the economic empowerment at Al-Hidayah Islamic Boarding School.

ICT and the impact of digital textbooks in foreign language teaching and learning

Abstract As part of the effort to find more effective methods for teaching foreign languages, this study focuses on the role and impact of digital books within the Albanian educational context, with particular emphasis on pedagogical practices related to the teaching of foreign languages in general and French language in particular. Using a mixed methodological approach, both quantitative and qualitative, including surveys of teachers, interviews, and analysis of data collected during classroom observations, this study assesses how the use of digital technologies, including digital books, influences student engagement and their achievements in language learning. The data analysis shows a positive correlation between the use of digital books and the improvement of students’ performance in various language skills. The integration of digital technologies not only increases engagement and performance but also contributes to the development of a more personalized and autonomous approach to teaching.

Real-Time Extraction of News Events Based on BERT Model

Abstract For the large number of news reports generated every day, in order to obtain effective information from these unstructured news text data more efficiently. In this paper, we study the real-time crawling of news data from news websites through crawling techniques and propose a BERT model-based approach to extract events from news long text. In this study, NetEase news website is selected as an example for real-time extraction to crawl the news data of this website. BERT model as a pre-trained model based on two-way encoded representation of transformer performs well on natural language understanding and natural language generation tasks. In this study, we will fine-tune the training based on BERT model on news corpus related dataset and perform sequence annotation through CRF layer to finally complete the event extraction task. In this paper, the DUEE dataset is chosen to train the model, and the experiments show that the overall performance of the BERT model is better than other network models. Finally, the model of this paper is further optimised, using the ALBERT and RoBERTa models improved on the basis of the BERT model, experiments were conducted, the results show that both models are improved compared to the BERT model, the ALBERT model algorithm performs the best, the model algorithm's F1 value is 1% higher than that of BERT. The results show that the performance is optimised.

Harmonizing Fine-tuned Llama 2 for Content Generation with Stable Diffusion for Image Synthesis in Article Creation

The research explores the integration of generative AI in multimedia content production using a fine-tuned Llama 2 model for text generation and the Stable Diffusion algorithm for image synthesis. The research analyses the fine-tuned Llama 2-7b-chat model's adaptability to specific content generation contexts, enhanced by a unique dataset and QLoRa, a Quantized Low-Rank Adaptation for parameter-efficient fine-tuning, achieving significant reductions in training loss and nuanced quality in the generated content. Notably, the model's evaluation yielded an impressive perplexity score of 1.49, indicating advanced predictive performance. Additionally, stable diffusion's ability to transform textual descriptions into intricate images, highlighting its potential in AI-mediated content creation is demonstrated. The experiments and qualitative analyses reveal improvements in efficiency and creativity, emphasizing the collaborative potential of these models to revolutionize multidisciplinary content generation. The research underscores the transformative impact of fine-tuned generative models on content creation and offers insights into the broader implications for future AI research, while acknowledging the critical need for ethical considerations in the deployment of such technologies.

CGFTrans: Cross-Modal Global Feature Fusion Transformer for Medical Report Generation.

Medical report generation, as a cross-modal automatic text generation task, can be highly significant both in research and clinical fields. The core is to generate diagnosis reports in clinical language from medical images. However, several limitations persist, including a lack of global information, inadequate cross-modal fusion capabilities, and high computational demands. To address these issues, we propose cross-modal global feature fusion Transformer (CGFTrans) to extract global information meanwhile reduce computational strain. Firstly, we introduce mesh recurrent network to capture inter-layer information at different levels to address the absence of global features. Then, we design feature fusion decoder and define 'mid-fusion' strategy to separately fuse visual and global features with medical report embeddings, which enhances the ability of the cross-modal joint learning. Finally, we integrate shifted window attention into Transformer encoder to alleviate computational pressure and capture pathological information at multiple scales. Extensive experiments conducted on three datasets demonstrate that the proposed method achieves average increments of 2.9%, 1.5%, and 0.7% in terms of the BLEU-1, METEOR and ROUGE-L metrics, respectively. Besides, it achieves average increments -22.4% and 17.3% training time and images throughput, respectively.

Methodology of Using Large Language Models to Solve Tasks of State and Municipal Government for Intelligent Abstracting and Automatic Generation of Text Content

Large language models (LLM) are finding new areas of application in practice, including the sphere of public and municipal administration. To increase the efficiency of the practical application of large language models rules and methods of interaction with them are developed, taking into account the specifics, a wide range of their possible use and increasing accessibility. The article examines the issues of improving the efficiency of large language models with various types of content using prompt engineering techniques. An analysis of a significant number of prompts for large language models and methods for their formation is presented. The article discusses the possibilities of using large language models, trained (customizable) using creative prompting for intelligent abstracting of various content with the subsequent generation of original texts and text documents for the sphere of state and municipal administration. The proposed methodology makes it possible to effectively integrate knowledge from various sources into LLM training and turn it into a truly intelligent tool that expands the possibilities of its work. When applying this approach, the LLM acts as a powerful intelligent assistant that allows you to generate a document authored by the user of the system. The use of large language models opens up wide opportunities for employees in the field of state and municipal administration to automate the process of creating thematic texts, text reports, qualification papers, reviews and analytical notes. It also allows users to see possible new meanings, previously unnoticed associations, and even generate new ideas in the field of management in the process of analyzing the texts received during the abstract. The authors have shown that in order to improve the quality of intellectual abstracting, it is necessary to carry out the iterative use of different methods of teaching (tuning) LLM. At the same time, the initial selection of texts for training, which is made by the user based on his/her own knowledge of the subject area, is important.

Postsurgical motor function and processing speed as predictors of quality of life in patients with chronic-phase glioblastoma

PurposePatients with glioblastomas (GBMs) have poor prognosis despite various treatments; therefore, attention should be paid to maintaining the quality of survival. Neurocognitive deficits can affect the quality of life (QOL) in patients with GBM. Most studies concerning QOL and neurocognitive functions have demonstrated a relationship between QOL and self-reported neurocognitive decline, although this method does not accurately reflect damaged functional domains. Therefore, this study aimed to clarify the neurocognitive functions that influence the QOL in patients with GBMs using an objective assessment of neurocognitive functions.MethodsData from 40 patients newly diagnosed with GBMs were analyzed. All patients completed the assessment of QOL and various neurological and neurocognitive functions including general cognitive function, processing speed, attention, memory, emotion recognition, social cognition, visuospatial cognition, verbal fluency, language, motor function, sensation, and visual field at 6 months postoperatively. QOL was assessed using the 36-Item Short Form Survey (SF-36). In the SF-36, the physical, mental, and role and social component summary (PCS, MCS, and RCS, respectively) scores were calculated. Multiple logistic regression analyses and chi-square tests were used to evaluate the association between SF-36 scores and neurocognitive functions.ResultsThe MCS was maintained, while the PCS and RCS scores were significantly lower in patients with GBMs than in healthy controls (p = 0.0040 and p < 0.0001, respectively). Among several neurocognitive functions, motor function and processing speed were significantly correlated with PCS and RCS scores, respectively (p = 0.0048 and p = 0.030, respectively). Patients who maintained their RCS or PCS scores had a higher probability of preserving motor function or processing speed than those with low RCS or PCS scores (p = 0.0026).ConclusionsMotor function and processing speed may be predictors of QOL in patients with GBMs.