Search

There has been a growing effort to replace manual extraction of data from research papers with automated data extraction based on natural language processing, language models, and recently, large language models (LLMs). Although these methods enable efficient extraction of data from large sets of research papers, they require a significant amount of up-front effort, expertise, and coding. In this work, we propose the ChatExtract method that can fully automate very accurate data extraction with minimal initial effort and background, using an advanced conversational LLM. ChatExtract consists of a set of engineered prompts applied to a conversational LLM that both identify sentences with data, extract that data, and assure the data's correctness through a series of follow-up questions. These follow-up questions largely overcome known issues with LLMs providing factually inaccurate responses. ChatExtract can be applied with any conversational LLMs and yields very high quality data extraction. In tests on materials data, we find precision and recall both close to 90% from the best conversational LLMs, like GPT-4. We demonstrate that the exceptional performance is enabled by the information retention in a conversational model combined with purposeful redundancy and introducing uncertainty through follow-up prompts. These results suggest that approaches similar to ChatExtract, due to their simplicity, transferability, and accuracy are likely to become powerful tools for data extraction in the near future. Finally, databases for critical cooling rates of metallic glasses and yield strengths of high entropy alloys are developed using ChatExtract. Efficient data extraction from research papers accelerates science and engineering. Here, the authors develop an automated approach which uses conversational large language models to achieve high precision and recall in extracting materials data. [ABSTRACT FROM AUTHOR]

The text discusses the intersection of natural language understanding (NLU) and reasoning in the context of large language models (LLMs) and traditional logic-based approaches. It highlights the strengths and weaknesses of both approaches and explores the potential for hybrid models that combine symbolic and distributional representations. The text also mentions specific applications of hybrid approaches in natural language inference, question-answering, sentiment analysis, and dialog. The document concludes by introducing a special issue that features selected contributions from the NALOMA workshop series, which focuses on hybrid methods in NLU. [Extracted from the article]

Large Language Models (LLMs) are a type of machine learning that handles a wide range of Natural Language Processing (NLP) scenarios. Recently, in December 2022, a company called OpenAI released ChatGPT, a tool that, within a few months, became the most representative example of LLMs, automatically generating unique and coherent text on many topics, summarising and rewriting it, or even translating it to other languages. ChatGPT originated some controversy in academia since students can generate unique text for writing assessments being sometimes extremely difficult to distinguish whether it comes from ChatGPT or a person. In research, some journals specifically banned ChatGPT in scientific papers. However, when used correctly, it becomes a powerful tool to rewrite, for instance, scientific papers and, thus, deliver researchers' messages in a better way. In this paper, we conduct an empirical study of the impact of ChatGPT in research. We downloaded the abstract of over 45,000 papers from over 300 journals from Dec 2022 and Feb 2023 belonging to different research editorials. We use four of the most known ChatGPT detection tools and conclude that ChatGPT played a role in around 10% of the papers published in every editorial, showing that authors from different fields have rapidly adopted such a tool in their research. [ABSTRACT FROM AUTHOR]

Due to the growing number of scholarly publications, finding relevant articles becomes increasingly difficult. Scholarly knowledge graphs can be used to organize the scholarly knowledge presented within those publications and represent them in machine-readable formats. Natural language processing (NLP) provides scalable methods to automatically extract knowledge from articles and populate scholarly knowledge graphs. However, NLP extraction is generally not sufficiently accurate and, thus, fails to generate high granularity quality data. In this work, we present TinyGenius, a methodology to validate NLP-extracted scholarly knowledge statements using microtasks performed with crowdsourcing. TinyGenius is employed to populate a paper-centric knowledge graph, using five distinct NLP methods. We extend our previous work of the TinyGenius methodology in various ways. Specifically, we discuss the NLP tasks in more detail and include an explanation of the data model. Moreover, we present a user evaluation where participants validate the generated NLP statements. The results indicate that employing microtasks for statement validation is a promising approach despite the varying participant agreement for different microtasks. [ABSTRACT FROM AUTHOR]

As the processing power of mobile terminals increases, wireless network applications such as voice assistants can put more context-sensitive tasks on the mobile terminals, thus reducing the wireless network bandwidth needed and the cost of data storage in the cloud. Co-reference annotation, identifying the same semantics in context, is one of the critical techniques in these tasks. However, there are some problems with the existing co-reference annotation standards. First, the annotation is incomplete. Second, the types of annotated mentions are inconsistent. Third, there are currently no metrics for the above characteristics. Analyzing the above-mentioned issues, this paper proposes a new co-reference annotation standard. The new standard can annotate more semantics and co-reference relations and only adopts two types of mentions for annotation. Meanwhile, this paper presents a performance evaluation corpus and designs three performance metrics for evaluating the new standard according to the completeness of semantic annotation, the completeness of co-reference annotation, and the consistency of mention. The experiment shows that the new standard outperforms all the baseline methods and achieves 0.95 in the completeness of semantic annotation, 0.68 in the completeness of co-reference annotation, and 0.57 in the consistency of types of mentions. [ABSTRACT FROM AUTHOR]

As social media and web-based forums have grown in popularity, the fast-spreading trend of fake news has become a major threat to the government and other agencies. With the rise of social media and internet platforms, misinformation may quickly spread over borders and language boundaries. Detecting and neutralizing fake news in several languages can help to protect the integrity of global elections, political discourse, and public opinion. The lack of a robust multilingual database for training the classification models makes detecting fake news a difficult task. This paper looks at it by describing several forms of fake news (like serious fabrications, large-scale hoaxes, stance news, deceptive news, satire news, clickbait, misinformation, rumour). This review paper includes different steps, features, tools for mitigating the scourge of information pollution, and different available datasets. This study presented a taxonomy for detecting fake news, which gives a comprehensive overview and analysis of existing DL-based algorithms focusing on diverse techniques. This paper also includes the monolingual and multilingual fake news detection models. Finally, this paper ends with the technical challenges. [ABSTRACT FROM AUTHOR]

This paper examines Russia's propaganda discourse on Twitter during the 2022 invasion of Ukraine. The study employs network analysis, natural language processing (NLP) techniques, and qualitative analysis to identify key communities and narratives associated with the prevalent and damaging narrative of "fascism/Nazism" in discussions related to the invasion. The paper implements a methodological pipeline to identify the main topics, and influential actors, as well as to examine the most impactful messages in spreading this disinformation narrative. Overall, this research contributes to the understanding of propaganda dissemination on social media platforms and provides insights into the narratives and communities involved in spreading disinformation during the invasion. [ABSTRACT FROM AUTHOR]

Emission data collection has always been a significant burden and challenge for Chinese counties to develop a CO2 emission inventory. This paper proposed a fuel consumption-based method to develop a local-specific CO2 emission rate database for Chinese counties using only open-source big data. Localized vehicle fuel consumption data is obtained through natural language processing (NLP) algorithm and large language model (LLM). The emission rates derived by our proposed method are consistent with field test results in literature. Besides, the CO2 emission estimation results using local-specific traffic activity data indicate that our method could effectively improve the accuracy of vehicle emission assessment. Compared with conventional method, the novel approach proposed in this paper can provide a pathway for convenient, universal, and cost-saving assessment for local scale CO2 emission rates. With this method, it is possible to formulate a local-specific CO2 emission database in various Chinese counties using only open-access big data. [ABSTRACT FROM AUTHOR]

Natural language inference (NLI) is a fundamental task of natural language processing (NLP). Most recent NLI research has focused on explaining the model's decisions in generating causal explanations (i.e., why did a premise/hypothesis pair as input lead to their inference relation as output?). As layer-based language models can learn language structure information, this paper conducts a sample-by-sample analysis of the linguistic feature relation between premise and hypothesis that is expected to guide NLI modeling and interpretation better. Our empirical study verifies that the linguistic feature relation of premise/hypothesis pairs can be seen in NLI inference models, which can be used to interpret inference samples. Meanwhile, experimental results show that these linguistic features relation interpretation can help the NLI model achieve comparable inference accuracy compared with state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The transformer model is a famous natural language processing model proposed by Google in 2017. Now, with the extensive development of deep learning, many natural language processing tasks can be solved by deep learning methods. After the BERT model was proposed, many pre-trained models such as the XLNet model, the RoBERTa model, and the ALBERT model were also proposed in the research community. These models perform very well in various natural language processing tasks. In this paper, we describe and compare these well-known models. In addition, we also apply several types of existing and well-known models which are the BERT model, the XLNet model, the RoBERTa model, the GPT2 model, and the ALBERT model to different existing and well-known natural language processing tasks, and analyze each model based on their performance. There are a few papers that comprehensively compare various transformer models. In our paper, we use six types of well-known tasks, such as sentiment analysis, question answering, text generation, text summarization, name entity recognition, and topic modeling tasks to compare the performance of various transformer models. In addition, using the existing models, we also propose ensemble learning models for the different natural language processing tasks. The results show that our ensemble learning models perform better than a single classifier on specific tasks. [ABSTRACT FROM AUTHOR]

This paper delves into the intricate relationship between Artificial Intelligence (AI) and the art ecosystem, emphasizing the need for a decolonizing approach in the face of AI's growing influence. It argues that the development of AI is not just a technological leap but also a significant cultural and societal moment, akin to the advent of moving images that Walter Benjamin famously analyzed. The paper examines how AI, particularly in its current oligarchical and corporate-driven form, perpetuates and magnifies the existing social inequalities, thereby necessitating a critical and radical rethinking of its role in society and the arts. At the heart of the discussion is the concept of AI as a broad term encompassing various forms of machine intelligence, from natural language processing to computer vision. The paper criticizes the dominant anthropocentric view of intelligence and creativity, proposing a more inclusive approach that considers the diverse forms of intelligence present in other species and potentially in AI itself. It underscores the role of AI in shaping the art ecosystem, not just in the creative process but also in gatekeeping and decision-making. The paper proposes a framework for decolonizing AI in the art ecosystem, focusing on four key tasks: recognizing access as a form of power, understanding and addressing biases inherent in AI, assessing the impact of AI on marginalized communities, and challenging dominant narratives and epistemologies to create space for alternative voices and perspectives. It emphasizes the need for artists and the art community to engage actively with AI, shaping its development towards more equitable and just outcomes. In conclusion, the paper calls for a radical reimagination of AI's role in society and the arts, advocating for a future where AI is not just about technological advancement but also about fostering a more inclusive, equitable, and creatively diverse world. It invites artists, thinkers, and innovators to join in this journey of reimagining and reshaping the future of AI and the art ecosystem. [ABSTRACT FROM AUTHOR]

Prolonged and over-excessive interaction with cyberspace poses a threat to people's health and leads to the occurrence of Cyber-Syndrome, which covers not only physiological but also psychological disorders. This paper aims to create a tree-shaped gold-standard corpus that annotates the Cyber-Syndrome, clinical manifestations, and acupoints that can alleviate their symptoms or signs, designating this corpus as CS-A. In the CS-A corpus, this paper defines six entities and relations subject to annotation. There are 448 texts to annotate in total manually. After three rounds of updating the annotation guidelines, the inter-annotator agreement (IAA) improved significantly, resulting in a higher IAA score of 86.05%. The purpose of constructing CS-A corpus is to increase the popularity of Cyber-Syndrome and draw attention to its subtle impact on people's health. Meanwhile, annotated corpus promotes the development of natural language processing technology. Some model experiments can be implemented based on this corpus, such as optimizing and improving models for discontinuous entity recognition, nested entity recognition, etc. The CS-A corpus has been uploaded to figshare. [ABSTRACT FROM AUTHOR]

Copyright of Frontiers of Information Technology & Electronic Engineering is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

People worldwide have suffered tremendously in terms of their mental health due to years of exposure to stress, anxiety, and the pressures of today's fast-paced lifestyles. The digitization of the data made possible by advances in health-care technology worldwide has allowed for a more precise mapping of the many variations of human biology than was possible before. People's methods of interaction with one another are evolving due to the rapid development of technology. Twitter, Facebook, Telegram, and Instagram have all risen to prominence as platforms where users can openly discuss their innermost thoughts, psyche, and feelings with one another. Texts are put through a psychological analysis process to pull out relevant details, characteristics, and user feedback. Psychological analysts rely on social media for the early identification of depressive symptoms and patterns of behavior. Machine learning has been recognized as a powerful method for sifting through the vast quantities of data in the health-care industry. Predicting the likelihood of mental diseases and executing likely treatment outcomes is a common application of ML techniques in mental health. This paper compiles a list of different mental health disorders along with the methods used in detecting and diagnosing mental health-related issues using online social media. [ABSTRACT FROM AUTHOR]

Introduction: The emergence of artificial intelligence (AI) chat programs has opened two distinct paths, one enhancing interaction and another potentially replacing personal understanding. Ethical and legal concerns arise due to the rapid development of these programs. This paper investigates academic discussions on AI in medicine, analyzing the context, frequency, and reasons behind these conversations. Methods: The study collected data from the Web of Science database on articles containing the keyword "ChatGPT" published from January to September 2023, resulting in 786 medically related journal articles. The inclusion criteria were peer-reviewed articles in English related to medicine. Results: The United States led in publications (38.1%), followed by India (15.5%) and China (7.0%). Keywords such as "patient" (16.7%), "research" (12%), and "performance" (10.6%) were prevalent. The Cureus Journal of Medical Science (11.8%) had the most publications, followed by the Annals of Biomedical Engineering (8.3%). August 2023 had the highest number of publications (29.3%), with significant growth between February to March and April to May. Medical General Internal (21.0%) was the most common category, followed by Surgery (15.4%) and Radiology (7.9%). Discussion: The prominence of India in ChatGPT research, despite lower research funding, indicates the platform's popularity and highlights the importance of monitoring its use for potential medical misinformation. China's interest in ChatGPT research suggests a focus on Natural Language Processing (NLP) AI applications, despite public bans on the platform. Cureus' success in publishing ChatGPT articles can be attributed to its open-access, rapid publication model. The study identifies research trends in plastic surgery, radiology, and obstetric gynecology, emphasizing the need for ethical considerations and reliability assessments in the application of ChatGPT in medical practice. Conclusion: ChatGPT's presence in medical literature is growing rapidly across various specialties, but concerns related to safety, privacy, and accuracy persist. More research is needed to assess its suitability for patient care and implications for non-medical use. Skepticism and thorough review of research are essential, as current studies may face retraction as more information emerges. [ABSTRACT FROM AUTHOR]

While the link between financial market movement and economic policy uncertainty indices is well-established in literature, uncertainty in the form of 'foreboding' emanating from catastrophic events has not been explored in literature. This paper explores "foreboding", which reflects uncertainty at its extreme, following the Covid-19 pandemic. Using Natural Language Processing on minute-by-minute news data, I construct two Foreboding Indices, representing 'foreboding' or 'fearful apprehension', for 28,622 Covid-related news for the period July 2020–August 2021. The impact of foreboding on financial market volatility is explored using a logistic regression model. Both the indices show a marked increase in June–July, 2020, in January 2021, April, 2021, and July–August, 2021 and have a positive impact on volatility for hourly S&P 500 Index. Understanding of foreboding sentiment is crucial for central banks looking to monitor financial market volatility. Appropriate signaling in accordance to sentiment can help central banks handle detrimental impacts of market volatility. Moreover, FI can be used for market practitioners to gauge the sentiment and take effective trading decisions. [ABSTRACT FROM AUTHOR]

In this editorial, we explore the urgent challenges created by the rise of infodemics —a term used to describe the epidemic spread of fake news, misinformation, and disinformation through social networks initially associated with the COVID-19 pandemic. This issue has drawn significant attention from various academic fields, including computer science, artificial intelligence, mathematics, physics, biology, sociology, and psychology, among others. This special issue is dedicated to advancing infodemics research across various academic domains. The selected papers include relevant contributions advancing the state of the art in the area, ranging from network analysis for identifying influential nodes and communities in networks to language processing for text classification and filtering relevant messages within extensive corpora. [ABSTRACT FROM AUTHOR]

The article discusses the recent advancements in cheminformatics, particularly in the representation of chemical information for drug design. However, there is a perceived lack of transformative ideas in science compared to previous decades, with a potential explanation being the "low hanging fruit" problem where easier problems have already been solved. Another possible reason for declining research productivity is a shift towards more defensive research and regulatory requirements. Despite this, there is an increase in the number of research papers being published, leading to information overload for researchers. The Journal of Cheminformatics aims to address this by introducing a Scientific Contribution Statement, allowing authors to highlight their contributions and differentiate their work from prior research. The statement is intended to be a helpful tool for both editors and readers in assessing the value of a research paper. [Extracted from the article]

Deep learning has seen significant growth recently and is now applied to a wide range of conventional use cases, including graphs. Graph data provides relational information between elements and is a standard data format for various machine learning and deep learning tasks. Models that can learn from such inputs are essential for working with graph data effectively. This paper identifies nodes and edges within specific applications, such as text, entities, and relations, to create graph structures. Different applications may require various graph neural network (GNN) models. GNNs facilitate the exchange of information between nodes in a graph, enabling them to understand dependencies within the nodes and edges. The paper delves into specific GNN models like graph convolution networks (GCNs), GraphSAGE, and graph attention networks (GATs), which are widely used in various applications today. It also discusses the message-passing mechanism employed by GNN models and examines the strengths and limitations of these models in different domains. Furthermore, the paper explores the diverse applications of GNNs, the datasets commonly used with them, and the Python libraries that support GNN models. It offers an extensive overview of the landscape of GNN research and its practical implementations. [ABSTRACT FROM AUTHOR]

Document similarity recognition is one of the most important problems in natural language processing. This paper proposes a plagiarism comparison mechanism called JCF. Initially, the TF–IDF scheme is applied to build a bag of words as the representation of the common features of all documents. Then, the plagiarism comparison is carried out in a coarse-grained manner, which speeds up the similarity comparison. Finally, the most similar documents can then be compared in detail based on a fine-grained approach. In addition, the JCF detects plagiarism at both syntax level and semantic-like level. To prevent the distortion of similarity comparison, this paper further develops a similarity restoration approach such that the proposed JCF can obtain both advantages of quickness and accuracy. Performance studies confirm that the proposed JCF outperforms existing studies in terms of precision, recall and F1 score. [ABSTRACT FROM AUTHOR]

Differential diagnosis is a crucial aspect of medical practice, as it guides clinicians to accurate diagnoses and effective treatment plans. Traditional resources, such as medical books and services like UpToDate, are constrained by manual curation, potentially missing out on novel or less common findings. This paper introduces and analyzes two novel methods to mine etiologies from scientific literature. The first method employs a traditional Natural Language Processing (NLP) approach based on syntactic patterns. By using a novel application of human-guided pattern bootstrapping patterns are derived quickly, and symptom etiologies are extracted with significant coverage. The second method utilizes generative models, specifically GPT-4, coupled with a fact verification pipeline, marking a pioneering application of generative techniques in etiology extraction. Analyzing this second method shows that while it is highly precise, it offers lesser coverage compared to the syntactic approach. Importantly, combining both methodologies yields synergistic outcomes, enhancing the depth and reliability of etiology mining. [ABSTRACT FROM AUTHOR]

Disaster management requires detailed data from past disasters for policy planning as well as for the generation of disaster exercises and simulations. Post-analysis of disasters is often distributed in official reports, which provide detailed analysis of the events that have happened. However, some information in such reports is often only given implicitly as part of natural language and thus not accessible to classical natural language processing-based text mining. To address this problem, in this paper, we propose to consider the information extraction tasks related to post-disaster report analysis as algebra stories, that can be treated with computer algebra systems together with natural language processing. We applied our enhanced information extraction approach in preliminary experiments to the report of a bushfire in 2009 in Victoria, Australia and used four different tools for solving fire-specific algebra story problems (ASP). Our evaluation shows that these tools have difficulty handling the occurring ASPs. [ABSTRACT FROM AUTHOR]

When processing language, the brain is thought to deploy specialized computations to construct meaning from complex linguistic structures. Recently, artificial neural networks based on the Transformer architecture have revolutionized the field of natural language processing. Transformers integrate contextual information across words via structured circuit computations. Prior work has focused on the internal representations ("embeddings") generated by these circuits. In this paper, we instead analyze the circuit computations directly: we deconstruct these computations into the functionally-specialized "transformations" that integrate contextual information across words. Using functional MRI data acquired while participants listened to naturalistic stories, we first verify that the transformations account for considerable variance in brain activity across the cortical language network. We then demonstrate that the emergent computations performed by individual, functionally-specialized "attention heads" differentially predict brain activity in specific cortical regions. These heads fall along gradients corresponding to different layers and context lengths in a low-dimensional cortical space. The extent to which transformer-based language models provide a good model of human brain activity during natural language comprehension is unclear. Here, the authors show that the internal transformations performed by the network predict brain activity in cortical language network as measured by fMRI. [ABSTRACT FROM AUTHOR]

In the globalization trend, China's cultural heritage is in danger of gradually disappearing. The protection and inheritance of these precious cultural resources has become a critical task. This paper focuses on the Miao batik culture in Guizhou Province, China, and explores the application of knowledge graphs, natural language processing, and deep learning techniques in the promotion and protection of batik culture. We propose a dual-channel mechanism that integrates semantic and visual information, aiming to connect batik pattern features with cultural connotations. First, we use natural language processing techniques to automatically extract batik-related entities and relationships from the literature, and construct and visualize a structured batik pattern knowledge graph. Based on this knowledge graph, users can textually search and understand the images, meanings, taboos, and other cultural information of specific patterns. Second, for the batik pattern classification, we propose an improved ResNet34 model. By embedding average pooling and convolutional operations into the residual blocks and introducing long-range residual connections, the classification performance is enhanced. By inputting pattern images into this model, their categories can be accurately identified, and then the underlying cultural connotations can be understood. Experimental results show that our model outperforms other mainstream models in evaluation metrics such as accuracy, precision, recall, and F1-score, achieving 94.46%, 94.47%, 93.62%, and 93.8%, respectively. This research provides new ideas for the digital protection of batik culture and demonstrates the great potential of artificial intelligence technology in cultural heritage protection. [ABSTRACT FROM AUTHOR]

Ransomware-as-a-service (RaaS) is increasing the scale and complexity of ransomware attacks. Understanding the internal operations behind RaaS has been a challenge due to the illegality of such activities. The recent chat leak of the Conti RaaS operator, one of the most infamous ransomware operators on the international scene, offers a key opportunity to better understand the inner workings of such organizations. This paper analyzes the main discussion topics in the Conti chat leak using machine learning techniques such as Natural Language Processing (NLP) and Latent Dirichlet Allocation (LDA), as well as visualization strategies. Five discussion topics are found: (1) Business, (2) Technical, (3) Internal tasking/Management, (4) Malware, and (5) Customer Service/Problem Solving. Moreover, the distribution of topics among Conti members shows that only 4% of individuals have specialized discussions while almost all individuals (96%) are all-rounders, meaning that their discussions revolve around the five topics. The results also indicate that a significant proportion of Conti discussions are non-tech related. This study thus highlights that running such large RaaS operations requires a workforce skilled beyond technical abilities, with individuals involved in various tasks, from management to customer service or problem solving. The discussion topics also show that the organization behind the Conti RaaS operator shares similarities with a large firm. We conclude that, although RaaS represents an example of specialization in the cybercrime industry, only a few members are specialized in one topic, while the rest runs and coordinates the RaaS operation. [ABSTRACT FROM AUTHOR]

This special issue features the selected works of authors who have presented papers at the 2022 iteration of the Joint Conference on Digital Libraries (JCDL) in Cologne, Germany. The motto of the conference was "Bridging Worlds" and was run as a fully hybrid event. Ten papers covering all aspects of Digital Libraries, namely Natural Language Processing, Information Retrieval, User Behavior, Scholarly Communication, Classification, Information Extraction are included in this issue. [ABSTRACT FROM AUTHOR]

Requirements Engineering (RE) has undergone several transitions over the years, from traditional methods to agile approaches emphasising increased automation. In many software development projects, requirements are expressed in natural language and embedded within large volumes of text documents. At the same time, RE activities aim to define software systems' functionalities and constraints. However, manually executing these tasks is time-consuming and prone to errors. Numerous research efforts have proposed tools and technologies for automating RE activities to address this challenge, which are documented in published works. This review aims to examine empirical evidence on automated RE and analyse its impact on the RE sub-domain and software development. To achieve our goal, we conducted a Systematic Literature Review (SLR) following established guidelines for conducting SLRs. We aimed to identify, aggregate, and analyse papers on automated RE published between 1996 and 2022. We outlined the output of the support tool, the RE phase covered, levels of automation, development approach, and evaluation approaches. We identified 85 papers that discussed automated RE from various perspectives and methodologies. The results of this review demonstrate the significance of automated RE for the software development community, which has the potential to shorten development cycles and reduce associated costs. The support tools primarily assist in generating UML models (44.7%) and other activities such as omission of steps, consistency checking, and requirement validation. The analysis phase of RE is the most widely automated phase, with 49.53% of automated tools developed for this purpose. Natural language processing technologies, particularly POS tagging and Parser, are widely employed in developing these support tools. Controlled experimental methods are the most frequently used (48.2%) for evaluating automated RE tools, while user studies are the least employed evaluation method (8.2%). This paper contributes to the existing body of knowledge by providing an updated overview of the research literature, enabling a better understanding of trends and state-of-the-art practices in automated RE for researchers and practitioners. It also paves the way for future research directions in automated requirements engineering. [ABSTRACT FROM AUTHOR]

Named entity recognition (NER) aims to identify the required entities and their types from unstructured text, which can be utilized for the construction of knowledge graphs. Traditional methods heavily rely on manual feature engineering and face challenges in adapting to large datasets within complex linguistic contexts. In recent years, with the development of deep learning, a plethora of NER methods based on deep learning have emerged. This paper begins by providing a succinct introduction to the definition of the problem and the limitations of traditional methods. It enumerates commonly used NER datasets suitable for deep learning methods and categorizes them into three classes based on the complexity of named entities. Then, some typical deep learning-based NER methods are summarized in detail according to the development history of deep learning models. Subsequently, an in-depth analysis and comparison of methods achieving outstanding performance on representative and widely used datasets is conducted. Furthermore, the paper reproduces and analyzes the recognition results of some typical models on three different types of typical datasets. Finally, the paper concludes by offering insights into the future trends of NER development. [ABSTRACT FROM AUTHOR]

Gynecological health remains a critical aspect of women's overall well-being, with profound implications for maternal and reproductive outcomes. This comprehensive review synthesizes the current state of knowledge on four pivotal aspects of gynecological health: preterm birth, breast cancer and cervical cancer and infertility treatment. Machine learning (ML) has emerged as a transformative technology with the potential to revolutionize gynecology and women's healthcare. The subsets of AI, namely, machine learning (ML) and deep learning (DL) methods, have aided in detecting complex patterns from huge datasets and using such patterns in making predictions. This paper investigates how machine learning (ML) algorithms are employed in the field of gynecology to tackle crucial issues pertaining to women's health. This paper also investigates the integration of ultrasound technology with artificial intelligence (AI) during the initial, intermediate, and final stages of pregnancy. Additionally, it delves into the diverse applications of AI throughout each trimester. This review paper provides an overview of machine learning (ML) models, introduces natural language processing (NLP) concepts, including ChatGPT, and discusses the clinical applications of artificial intelligence (AI) in gynecology. Additionally, the paper outlines the challenges in utilizing machine learning within the field of gynecology. [ABSTRACT FROM AUTHOR]

This paper analyses Twitter data to detect the political lean of a profile by extracting and classifying sentiments expressed through tweets. The work utilizes natural language processing, augmented with sentiment analysis algorithms and machine learning techniques, to classify specific keywords. The proposed methodology initially performs data pre-processing, followed by multi-aspect sentiment analysis for computing the sentiment score of the extracted keywords, for precisely classifying users into various clusters based on similarity score with respect to a sample user in each cluster. The proposed technique also predicts the sentiment of a profile towards unknown keywords and gauges the bias of an unidentified user towards political events or social issues. The proposed technique was tested on Twitter dataset with 1.72 million tweets taken from over 10,000 profiles and was able to successfully identify the political leniency of the user profiles with 99% confidence level, and also on a synthetic dataset with 2500 tweets, where the predicted accuracy and F1 score were 0.99 and 0.985 respectively, and 0.97 and 0.975 when neutral users were also considered for classification. The paper could also identify the impact of political decisions on various clusters, by analyzing the shift in the number of users belonging to the different clusters. [ABSTRACT FROM AUTHOR]

The objective of advanced topic modeling is not only to explore latent topical structures, but also to estimate relationships between the discovered topics and theoretically relevant metadata. Methods used to estimate such relationships must take into account that the topical structure is not directly observed, but instead being estimated itself in an unsupervised fashion, usually by common topic models. A frequently used procedure to achieve this is the method of composition, a Monte Carlo sampling technique performing multiple repeated linear regressions of sampled topic proportions on metadata covariates. In this paper, we propose two modifications of this approach: First, we substantially refine the existing implementation of the method of composition from the R package stm by replacing linear regression with the more appropriate Beta regression. Second, we provide a fundamental enhancement of the entire estimation framework by substituting the current blending of frequentist and Bayesian methods with a fully Bayesian approach. This allows for a more appropriate quantification of uncertainty. We illustrate our improved methodology by investigating relationships between Twitter posts by German parliamentarians and different metadata covariates related to their electoral districts, using the structural topic model to estimate topic proportions. [ABSTRACT FROM AUTHOR]

Aspect-base sentiment analysis (ABSA) refers to the analysis of sentiments in users' notes from the perspective of different aspects, which has attracted much attention in the last decade. With the development of social networks, user-generated content on various Internet platforms is growing rapidly. This content contains valuable information, especially in e-commerce, which helps people and consumers make better decisions. Therefore, improving the performance of ABSA methods has recently attracted the attention of many researchers. In this paper, the basic concepts in sentiment analysis, applications, challenges, evaluation metrics, common benchmark datasets, required pre-processing and textual representation methods have been defined. An overview of all available methods for ABSA is also provided, and then a detailed review of deep learning methods in ABSA with a focus on graph neural networks has been done. Finally, this paper reviews the latest studies that have used deep learning and graph neural networks to solve sentiment analysis problems. We hope that our summaries in this paper can provide necessary guidelines for beginners and new researchers. [ABSTRACT FROM AUTHOR]

The restricted access to data in healthcare facilities due to patient privacy and confidentiality policies has led to the application of general natural language processing (NLP) techniques advancing relatively slowly in the health domain. Additionally, because clinical data is unique to various institutions and laboratories, there are not enough standards and conventions for data annotation. In places without robust death registration systems, the cause of death (COD) is determined through a verbal autopsy (VA) report. A non-clinician field agent completes a VA report using a set of standardized questions as guide to identify the symptoms of a COD. The narrative text of the VA report is used as a case study to examine the difficulties of applying NLP techniques to the healthcare domain. This paper presents a framework that leverages knowledge across multiple domains via two domain adaptation techniques: feature extraction and fine-tuning. These techniques aim to improve VA text representations for COD classification tasks in the health domain. The framework is motivated by multi-step learning, where a final learning task is realized via a sequence of intermediate learning tasks. The framework builds upon the strengths of the Bidirectional Encoder Representations from Transformers (BERT) and Embeddings from Language Models (ELMo) models pretrained on the general English and biomedical domains. These models are employed to extract features from the VA narratives. Our results demonstrate improved performance when initializing the learning of BERT embeddings with ELMo embeddings. The benefit of incorporating character-level information for learning word embeddings in the English domain, coupled with word-level information for learning word embeddings in the biomedical domain, is also evident. [ABSTRACT FROM AUTHOR]

Automatic grading models are valued for the time and effort saved during the instruction of large student bodies. Especially with the increasing digitization of education and interest in large-scale standardized testing, the popularity of automatic grading has risen to the point where commercial solutions are widely available and used. However, for short answer formats, automatic grading is challenging due to natural language ambiguity and versatility. While automatic short answer grading models are beginning to compare to human performance on some datasets, their robustness, especially to adversarially manipulated data, is questionable. Exploitable vulnerabilities in grading models can have far-reaching consequences ranging from cheating students receiving undeserved credit to undermining automatic grading altogether—even when most predictions are valid. In this paper, we devise a black-box adversarial attack tailored to the educational short answer grading scenario to investigate the grading models' robustness. In our attack, we insert adjectives and adverbs into natural places of incorrect student answers, fooling the model into predicting them as correct. We observed a loss of prediction accuracy between 10 and 22 percentage points using the state-of-the-art models BERT and T5. While our attack made answers appear less natural to humans in our experiments, it did not significantly increase the graders' suspicions of cheating. Based on our experiments, we provide recommendations for utilizing automatic grading systems more safely in practice. [ABSTRACT FROM AUTHOR]

Code assistance refers to the utilization of various tools, techniques, and models to help developers in the process of software development. As coding tasks become increasingly complex, code assistant plays a pivotal role in enhancing developer productivity, reducing errors, and facilitating a more efficient coding workflow. This assistance can manifest in various forms, including code autocompletion, error detection and correction, code generation, documentation support, and context-aware suggestions. Language models have emerged as integral components of code assistance, offering developers the capability to receive intelligent suggestions, generate code snippets, and enhance overall coding proficiency. In this paper, we propose new hybrid models for code generation by leveraging pre-trained language models BERT, RoBERTa, ELECTRA, and LUKE with the Marian Causal Language Model. Selecting these models based on their strong performance in various natural language processing tasks. We evaluate the performance of these models on two datasets CoNaLa and DJANGO and compare them to existing state-of-the-art models. We aim to investigate the potential of pre-trained transformer language models to revolutionize code generation, offering improved precision and efficiency in navigating complex coding scenarios. Additionally, conducting error analysis and refining the generated code. Our results show that these models, when combined with the Marian Decoder, significantly improve code generation accuracy and efficiency. Notably, the RoBERTaMarian model achieved a maximum BLEU score of 35.74 and an exact match accuracy of 13.8% on CoNaLa, while LUKE-Marian attained a BLEU score of 89.34 and an exact match accuracy of 78.50% on DJANGO. Implementation of this work is available at https://github.com/AhmedSSoliman/Leveraging-Pretrained-Language-Models-for-Code-Generation. [ABSTRACT FROM AUTHOR]

Recently, visual question answering (VQA) has gained considerable interest within the computer vision and natural language processing (NLP) research areas. The VQA task involves answering a question about an image, which requires both language and vision understanding. Effectively extracting visual representations from images, textual embedding from questions, and bridging the semantic disparity between image and question representations pose fundamental challenges in VQA. Lately, an increasing number of studies are focusing on utilizing graph neural networks (GNNs) to enhance the performance of VQA tasks. The ability to handle graph-structured data is a major advantage of GNNs for VQA tasks, which allows better representation of relationships between objects and regions in an image. These relationships include both spatial and semantic relationships. This paper systematically reviews various graph neural networks based studies for image-based VQA. Fifty-four related publications written between 2018—Jan. 2023 were carefully synthesized for this review. The review is structured into three perspectives: the various graph neural network techniques and models that have been applied for VQA, a comparison of the model's performance and existing challenges. After analyzing these papers, 45 different models were identified, grouped into four different GNN techniques. These are Graph Convolution Network (GCN), Graph Attention Network (GAT), Graph Isomorphism Network (GIN) and Graph Neural Network (GNN). Also, the performance of these models is compared based on accuracy, datasets, subtasks, feature representation and fusion techniques. Lastly, the study provided some possible suggestions to mitigate still existing challenges for future research in visual question answering. [ABSTRACT FROM AUTHOR]

Sentiment analysis involves extracting sentiments from various forms of text, including customer reviews, tweets, blogs, and news clips expressing opinions on diverse subjects, even populist events. The advent of tools supporting regional languages has resulted in a substantial surge of regional language texts. As Hindi ranks fourth in terms of native speakers, the development of sentiment analysis mechanisms for Hindi text becomes crucial. This paper provides a comprehensive review of specific approaches used in Hindi sentiment analysis, encompassing negation handling and the evolution of SentiWordNet for the Hindi Language. Moreover, it offers an overview of available Hindi lexicons and insights into diverse stemmers and morphological analyzers designed for the language. Additionally, the paper conducts an in-depth literature review of various sentiment analysis tasks carried out in Hindi, thereby opening avenues for future research in sentiment analysis and opinion mining in the Hindi language. [ABSTRACT FROM AUTHOR]

A long-term objective of artificial intelligence is to design an abstractive text summarization (ATS) system that can produce condensed, adequate, and realistic summaries for the source documents. Deep learning approaches have contributed significantly to recent advancements in ATS, taking the state-of-the-art to new heights. Despite the considerable success of earlier approaches, producing high-quality and more human-like abstractive summaries continues to be a challenging task in reality. In this paper, we present an adversarial framework for abstractive text summarization, in which a generative model G and a proposed deep discriminative model D are trained in an adversarial manner. The goal of the generator is to generate summaries that are hard to differentiate from real summaries, whereas the discriminator's role is to estimate the probability that a summary came from the training data rather than the generator to direct the generative model's training. Experimental results on a benchmark CNN/Daily Mail dataset demonstrate that the proposed model achieves Rouge-1 and Rouge-L scores of 41.58 and 38.96 respectively which are better than the ones reported by various other methods (e.g. the base paper [1] achieved Rouge-1 and Rouge-L score of 39.92 and 36.71, while one of the recent works ACGT [2] achieved Rouge-1 score of 40.49 and Rouge-L score of 37.41). The manuscript's unique characteristic is qualitative evaluation, which, along with quantitative evaluation, shows that the proposed model is superior. [ABSTRACT FROM AUTHOR]

The technology of the Internet of Things (IoT) represents a cornerstone of the fourth industrial revolution. We adopt a machine learning approach to examine the effect of IoT technology on shipping business operations. Text mining and the probabilistic latent Dirichlet allocation are applied for an unsupervised topic modelling analysis of two hundred and twenty-eight academic papers. Our findings reveal the potential of IoT to provide more efficient approaches to business operations and improve the quality of services, highlighting the value of instant and secure information flow among all parties involved. Problematic areas of the new technology are also identified, in reference to issues of standardization and interoperability. Relatively few studies have used machine learning techniques to elicit insights into the holistic effect of emerging IoT technology in the shipping industry. The research findings highlight the potential of IoT technology to transform shipping operations, offering useful and practical implications to academics and professionals. [ABSTRACT FROM AUTHOR]

Image Captioning is the process of generating textual descriptions of given images. It encompasses two major fields of deep learning, computer vision, and natural language processing. This paper presents an Image Captioning model which uses the Convolution Neural Network (CNN) model for feature extraction and a transformer architecture for the generation of sequences from these feature vectors. For feature extraction, this paper uses different CNN architectures like Xception, InceptionV3, ResNet50V2, VGG19, DenseNet201, ResNet152V2, EfficientNetV2B3, EfficientNetV2B0. The proposed method takes advantage of the transformer model for faster processing, and Beam search is implemented to get the top N most probable sequences for each image. The architecture is trained on Flickr8k dataset and the model outperforms the existing methods. The proposed model achieves a BLEU_4 score of 0.2184 on the Flickr8k dataset. [ABSTRACT FROM AUTHOR]

The utilization of word embeddings—powerful models computed through Neural Network architectures that encode words as vectors—has witnessed rapid growth across various Natural Language Processing applications, encompassing semantic analysis, information retrieval, dependency parsing, question answering, and machine translation. The efficacy of these tasks is strictly linked to the quality of the embeddings, underscoring the critical importance of evaluating and selecting optimal embedding models. While established procedures and benchmarks exist for intrinsic evaluation, the authors note a conspicuous absence of comprehensive evaluations of intrinsic embedding quality across multiple tasks. This paper introduces vec2best, a unified tool encompassing state-of-the-art intrinsic evaluation tasks across diverse benchmarks. vec2best furnishes the user with an extensive evaluation of word embedding models. It represents a framework for evaluating word embeddings trained using various methods and hyper-parameters on a range of tasks from the literature. The tool yields a holistic evaluation metric for each model called the PCE (Principal Component Evaluation). We conducted evaluations on 135 word embedding models, trained using GloVe, fastText, and word2vec, across four tasks integrated into vec2best (similarity, analogy, categorization, and outlier detection), along with their respective benchmarks. Additionally, we leveraged vec2best to optimize embedding hyper-parameter configurations in a real-world scenario. vec2best is conveniently accessible as a pip-installable Python package. [ABSTRACT FROM AUTHOR]

In today's fast-paced world, recommendation systems have become indispensable tools, aiding users in making personalized decisions amidst an overwhelming array of choices. These systems leverage user data and preferences to generate tailor-made recommendations based on individual tastes and behaviors. This research paper introduces the development and implementation of Pranik Movies, an ultimate recommendation system for personalized movie suggestions. The system incorporates collaborative and content-based filtering techniques, utilizing machine learning algorithms to analyze user behaviors, ratings, and viewing histories. A comprehensive overview of the research framework is provided, encompassing system architecture, data pre-processing, feature engineering techniques, and model selection and design. Text processing methods such as stemming, bag-of-words (BoW), and TF-IDF (Term Frequency-Inverse Document Frequency) are employed for processing and analyzing textual movie data. The accuracy of recommendations is enhanced through the assessment of film similarities, utilizing algorithms like cosine similarity and Euclidean distance. The paper concludes by outlining future directions for advanced machine learning techniques, social media integration, expanded content support, and the refinement of the evaluation framework. Pranik Movies signifies a significant advancement in recommendation systems, enabling personalized and precise movie recommendations within a vast and diverse cinematic landscape. [ABSTRACT FROM AUTHOR]

In the United States, recent changes to the National Statistical System have amplified the geographic-demographic resolution trade-off. That is, when working with demographic and economic data from the American Community Survey, as one zooms in geographically one loses resolution demographically due to very large margins of error. In this paper, we present a solution to this problem in the form of an AI based open and reproducible geodemographic classification system for the United States using small area estimates from the American Community Survey (ACS). We employ a partitioning clustering algorithm to a range of socio-economic, demographic, and built environment variables. Our approach utilizes an open source software pipeline that ensures adaptability to future data updates. A key innovation is the integration of GPT4, a state-of-the-art large language model, to generate intuitive cluster descriptions and names. This represents a novel application of natural language processing in geodemographic research and showcases the potential for human-AI collaboration within the geospatial domain. [ABSTRACT FROM AUTHOR]

Named entity recognition (NER) is a fundamental task in natural language processing that involves identifying and classifying entities in sentences into pre-defined types. It plays a crucial role in various research fields, including entity linking, question answering, and online product recommendation. Recent studies have shown that incorporating multilingual and multimodal datasets can enhance the effectiveness of NER. This is due to language transfer learning and the presence of shared implicit features across different modalities. However, the lack of a dataset that combines multilingualism and multimodality has hindered research exploring the combination of these two aspects, as multimodality can help NER in multiple languages simultaneously. In this paper, we aim to address a more challenging task: multilingual and multimodal named entity recognition (MMNER), considering its potential value and influence. Specifically, we construct a large-scale MMNER dataset with four languages (English, French, German and Spanish) and two modalities (text and image). To tackle this challenging MMNER task on the dataset, we introduce a new model called 2M-NER, which aligns the text and image representations using contrastive learning and integrates a multimodal collaboration module to effectively depict the interactions between the two modalities. Extensive experimental results demonstrate that our model achieves the highest F1 score in multilingual and multimodal NER tasks compared to some comparative and representative baselines. Additionally, in a challenging analysis, we discovered that sentence-level alignment interferes a lot with NER models, indicating the higher level of difficulty in our dataset. [ABSTRACT FROM AUTHOR]

This in-depth study journey explores the context of natural language processing and text analysis in dementia detection, revealing their importance in a variety of fields. Beginning with an examination of the widespread and influence of text data. The dataset utilised in this study is from TalkBank's DementiaBank, which is basically a vast database of multimedia interactions built with the goal of examining communication patterns in the context of dementia. The various communication styles dementia patients exhibit when communicating with others are seen from a unique perspective by this specific dataset. Thorough data preprocessing procedures, including cleansing, tokenization, and structuring, are undertaken, with a focus on improving prediction capabilities through the combination of textual and non-textual information in the field of feature engineering. In the subsequent phase, the precision, recall, and F1-score metrics of Support Vector Machines (SVM), K-Nearest Neighbours (KNN), Random Forest, and Artificial Neural Networks (ANN) are assessed. Empirical facts are synthesized using text analysis methods and models to formulate a coherent conclusion. The significance of text data analysis, the revolutionary potential of natural language processing, and the direction for future research are highlighted in this synthesis. Throughout this paper, readers are encouraged to leverage text data to embark on their own adventures in the evolving, data-centric world of dementia detection. [ABSTRACT FROM AUTHOR]

Deep learning models produce impressive results in any natural language processing applications when given a better learning strategy and trained with large labeled datasets. However, the annotation of massive training data is far too expensive, especially in the legal domain, due to the need for trained legal professionals. Data augmentation solves the problem of learning without labeled big data. In this paper, we employ pre-trained language models and prompt engineering to generate large-scale pseudo-labeled data for the legal overruling task using 100 data samples. We train small recurrent and convolutional deep-learning models using this data and fine-tune a few other transformer models. We then evaluate the effectiveness of the models, both with and without data augmentation, using the benchmark dataset and analyze the results. We also test the performance of these models with the state-of-the-art GPT-3 model under few-shot setting. Our experimental findings demonstrate that data augmentation results in better model performance in the legal overruling task than models trained without augmentation. Furthermore, our best-performing deep learning model trained on augmented data outperforms the few-shot GPT-3 by 18% in the F1-score. Additionally, our results highlight that the small neural networks trained with augmented data achieve outcomes comparable to those of other large language models. [ABSTRACT FROM AUTHOR]

An emerging direction of quantum computing is to establish meaningful quantum applications in various fields of artificial intelligence, including natural language processing (NLP). Although some efforts based on syntactic analysis have opened the door to research in quantum NLP (QNLP), limitations such as heavy syntactic preprocessing and syntax-dependent network architecture make them impracticable on larger and real-world data sets. In this paper, we propose a new simple network architecture, called the quantum self-attention neural network (QSANN), which can compensate for these limitations. Specifically, we introduce the self-attention mechanism into quantum neural networks and then utilize a Gaussian projected quantum self-attention serving as a sensible quantum version of self-attention. As a result, QSANN is effective and scalable on larger data sets and has the desirable property of being implementable on near-term quantum devices. In particular, our QSANN outperforms the best existing QNLP model based on syntactic analysis as well as a simple classical self-attention neural network in numerical experiments of text classification tasks on public data sets. We further show that our method exhibits robustness to low-level quantum noises and showcases resilience to quantum neural network architectures. [ABSTRACT FROM AUTHOR]

Natural Language Processing is crucial because it clears up linguistic ambiguity and gives the data valuable quantitative structure for numerous downstream applications, including text analytics or speech recognition. In Natural Language Processing (NLP), knowledge summarization is the act of condensing information from lengthy texts for easier reading. In the present study, data were collected from the abstracts of the papers based on Friction Stir Welding Magnesium Alloys. These collected data were subjected to five Natural Language Processing based algorithms i.e., Text Rank, Lex Rank, Latent Semantic Analysis (LSA), Luhn (modulus 10), and KL-SUM for summarization purposes. The performance of these algorithms was evaluated by the ROUGE algorithm. The results demonstrated that the Luhn algorithm performs knowledge summarization with the greatest accuracy and F1 Score. By calculating the harmonic mean of a classifier's precision and recall, the F1-score integrates both into a single metric. It is mainly used to evaluate the efficiency of two classifiers. [ABSTRACT FROM AUTHOR]

Image - Caption Generator is a popular Artificial Intelligence research tool that works with image comprehension and language definition. Creating well-structured sentences requires a thorough understanding of language in a systematic and semantic way. Being able to describe the substance of an image using well-structured phrases is a difficult undertaking, but it can have a significant impact in terms of assisting visually impaired people in better understanding the images' content. Image captions has gained a lot of attention as a study subject for various computer vision and natural language processing (NLP) applications. The goal of image captions is to create logical and accurate natural language phrases that describes an image. It relies on the caption model to see items and appropriately characterise their relationships. Intuitively, it is also difficult for a machine to see a typical image in the same way that humans do. It does, however, provide the foundation for intelligent exploration in deep learning. In this review paper, we will focus on the latest in-depth advanced captions techniques for image captioning. This paper highlights related methodologies and focuses on aspects that are crucial in computer recognition, as well as on the numerous strategies and procedures being developed for the development of image captions. It was also observed that Recurrent neural networks (RNNs) are used in the bulk of research works (45%), followed by attention-based models (30%), transformer-based models (15%) and other methods (10%). An overview of the approaches utilised in image captioning research is discussed in this paper. Furthermore, the benefits and drawbacks of these methodologies are explored, as well as the most regularly used data sets and evaluation processes in this sector are being studied. [ABSTRACT FROM AUTHOR]

Fanfictions are a popular literature genre in which writers reuse a universe, for example to transform heteronormative relationships with queer characters or to bring romance into shows focused on horror and adventure. Fanfictions have been the subject of numerous studies in text mining and network analysis, which used Natural Language Processing (NLP) techniques to compare fanfictions with the original scripts or to make various predictions. In this paper, we use NLP to predict the popularity of a story and examine which features contribute to popularity. This endeavor is important given the rising use of AI assistants and the ongoing interest in generating text with desirable characteristics. We used the main two websites to collect fan stories (Fanfiction.net and Archives Of Our Own) on Supernatural, which has been the subject of numerous scholarly works. We extracted high-level features such as the main character and sentiments from 79,288 of these stories and used the features in a binary classification supported by tree-based methods, ensemble methods (random forest), neural networks, and Support Vector Machines. Our optimized classifiers correctly identified popular stories in four out of five cases. By relating features to classification outcomes using SHAP values, we found that fans prefer longer stories with a wider vocabulary, which can inform the prompts of AI chatbots to continue generating such successful stories. However, we also observed that fans wanted stories unlike the original material (e.g., favoring romance and disliking when characters are hurt), hence AI-powered stories may be less popular if they strictly follow the original material of a show. [ABSTRACT FROM AUTHOR]