|
|
- Trend Analysis with Interpretability and Cold-Start Problems for
Recommender Systems-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Matrix factorization is a common method in recommender systems. However, distinguishing between continuously and temporarily popular items is challenging because basic matrix factorization relies on the accumulated user rating records for an item. Moreover, recent trends emphasize that recommender systems should be both accurate and interpretable, necessitating clear reasons behind each recommendation. In this paper, we propose temporal positive collective matrix factorization (TPCMF), which improves the interpretability and temporality of collective matrix factorization. We make the factor matrix obtained by collective matrix factorization non-negative to increase the interpretability. In addition, we take into account the time variations of the factor matrices and make time-series predictions, which enables temporal recommendations. Moreover, in experiments using real-world datasets, we determined that factor interpretation under TPCMF provides substantial insights into the interpretability and temporality of recommendations with accuracy surpassing existing methods. In addition to these features, we also propose a preprocessing method to address the cold-start problem, which is a common issue in matrix factorization.
PubDate: 2024-08-07
- Firm Default Prediction by GNN with Gravity-Model Informed Neighbor Node
Sampling-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Firm default prediction is important in credit risk management and understanding economic trends. Both practitioners and academic researchers have long studied it. While traditional statistical methods such as discriminant analysis and logistic regression have been used recently, machine learning and deep learning methods have been widely applied. The graph neural network (GNN) is one of the latest applications of deep-learning approaches. With the use of GNNs, it is possible to reflect the non-linear relationships of features among neighboring companies around the target company, whereas ordinary machine learning and deep learning methods focus only on the features of the target company. However, when handling large-scale graphs such as inter-firm networks, it is difficult to apply vanilla GNNs naively. Although uniform neighbor node sampling is commonly used for large-scale graphs, to the best of our knowledge, no research has focused on better sampling methods for GNN applications for default prediction. From the practical viewpoint, it means which companies should be considered with priority for firm default prediction. In this study, we propose a novel gravity model-informed neighbor sampling method based on the estimated transaction volume by utilizing knowledge from econophysics. The scope of this research is to determine whether we can improve default predictions by considering neighboring companies with larger transaction amounts compared to ordinary uniform sampling. We also verified that the proposed method improves the prediction performance and stability compared to GNNs with other sampling techniques and other machine learning methods using real large-scale inter-firm network data.
PubDate: 2024-07-29
- Development of a Product Recommendation Method for the Real-Time Cart
Context in Supermarkets by Integrating Cooking Recipes and Purchase
Co-occurrences-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract The effectiveness of product recommendation systems is critical to enhancing customer experiences and boosting sales in the rapidly evolving retail domain, especially in supermarkets. Thus, in this study, we design an innovative recommendation approach for physical supermarkets, and our approach integrates insights from previous purchasing patterns with current shopping cart compositions augmented with recipe-based information. As this approach deviates from traditional strategies, which primarily rely on historical data, it dynamically addresses shoppers’ immediate preferences and recommends products that suit their intended purchases. Furthermore, we evaluate the effectiveness of this technique using data from smart shopping carts in a brick-and-mortar supermarket, revealing significant improvements in key performance indicators, such as Recall, Precision, and the F1 score, than with the existing methods. These results highlight the benefits of integrating real-time cart data with historical purchasing patterns, offering a path to more personalized and efficient recommendations in retail environments. This study illustrates the potential of such integrated approaches toward significantly improving in-store shopping experiences.
PubDate: 2024-07-27
- Investigating Factors that Influence Purchase Intentions in Live-Streaming
Contexts Through the Elaboration Likelihood Model: The Perspectives of
Para-Social Interaction and Information Quality-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract This research is grounded in the Elaboration Likelihood Model and investigates the determinants of purchase intentions within live-streaming settings. It differentiates between peripheral cues (para-social interaction and source credibility) and central cues (information quality), which are pivotal in shaping consumer attitudes and, subsequently, purchase intentions. Furthermore, this study assesses the moderating roles of product involvement and product consistency. Data were gathered through an online survey, yielding 299 valid responses. The findings indicate that para-social interaction, source credibility, and information quality significantly and positively influence consumer attitudes when neither product involvement nor consistency is accounted for. The analysis of the moderating variables presented diverse outcomes. Specifically, product involvement significantly influences the cognitive processing routes of the audience, aligning with prior research; viewers with low involvement form attitudes predominantly through para-social interaction and source credibility, whereas those with high involvement depend more on information quality. Conversely, despite its significance in prior studies, product consistency did not exhibit a substantial moderating effect in this research context. These results provide valuable insights for both academic research and practical applications in online marketing strategies.
PubDate: 2024-07-20
- Event Identification for Supply Chain Risk Management Through News
Analysis by Using Large Language Models-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Event identification is important in many areas of the business world. In the supply chain risk management domain, the timely identification of risk events is vital to ensure the success of supply chain operations. One of the important sources of real-time information from across the world is news sources. However, the analysis of large amounts of daily news cannot be done manually by humans. On the other hand, extracting related news depends on the query or the keyword used in the search engine and the news content. Recent advancements in artificial intelligence have opened up opportunities to leverage intelligent techniques to automate this analysis. This paper introduces the LUEI framework, a lightweight framework that, with only the event’s name as input, can autonomously learn all the related phrases associated with that event. It then employs these phrases to search for relevant news and presents the search engine results with a label indicating their relevance. Hence, by conducting this analysis, the LUEI framework is able to identify the occurrence of the event in the real world. The framework’s novel contribution lies in its ability to identify those events (termed as the Contributing Events (CEs)) that contribute to the occurrence of a risk event, offering a proactive approach to risk management in supply chains. Pinpointing CEs from vast news data gives supply chain managers actionable insights to mitigate risks before they escalate.
PubDate: 2024-07-15
- Application of Deep Learning in the Classification of Maritime Safety
Information-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Maritime safety information (MSI) refers to urgent information concerning navigation warnings, weather warnings, weather forecasts, and other safety-related information for navigation. MSI is primarily disseminated through two systems: the international NAVTEX (Navigational Telex) system and the Enhanced Group Calling (EGC) system. NAVTEX receivers on ships can automatically receive MSI, enhancing safety of life and property at sea and reducing the workload for communication personnel. Due to the narrowband direct printing telegraphy (NBDP) technology used in the broadcast of MSI, the information must be concise, it makes the information difficult for navigators to interpret. To address this challenge, various machine learning solutions have been proposed. Among these, deep neural networks have shown superior performance in MSI classification. This study provides a more detailed analysis of deep neural networks applied to MSI classification. It utilizes collected MSI datasets, including thousands of navigation warnings, and compares the performances of different models based on accuracy, precision, recall, and F1-score.
PubDate: 2024-07-04
- The Integration of Artificial Intelligence and Video Production Skills in
Workplace Development: A Study from the Perspective of Vocational Training
-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract This paper explores the potential impact of artificial intelligence (A.I.) and video production skills on workplace development, focusing on the perspective of vocational training. With the rapid growth of A.I. technology, it is essential to investigate how learners acquire video production skills under the guidance of A.I. and imagine the possibilities for their future career development. This study examines the process of learning video production skills in the A.I.-dominated era and explores learners’ perceptions and expectations of their career development trajectory. This study adopts a qualitative approach, primarily through focus group interviews and thematic analysis. The findings of this study indicate that video-making skills are crucial in online marketing. While artificial intelligence enhances efficiency, human creativity remains indispensable. Integrating A.I. into training can improve learning outcomes. Recommendations include emphasizing hands-on experience, personalized learning paths, and collaborative learning.
PubDate: 2024-06-25
- Evaluating the Network Performance of the Ensembled-Based Veracity
Architecture for Fake News Detection in Infrastructureless Social Networks
-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Infrastructureless social networks (ISNs) are created by the interconnection of spectral-constrained mobile devices. One such type of ISN are the mobile ad hoc networks (MANETs). One of the issues that consumers of content on ISNs face is the inability to detect fake news in messages sent through the network. To address the fake news detection issue in ISNs, the ensemble-based veracity architecture, an ensemble-based computational social system for fake news detection in infrastructureless social networks, has been proposed. Ensemble-based Veracity detects fake news using both the publisher’s credibility and the content of the news. To understand the effect that ensemble-based Veracity has on network performance, this work investigates the network performance of the ensemble-based Veracity architecture. Ensemble-based Veracity is fully evaluated using a MANET-based experimental design and simulation environment. The network performance results of the experiments on ensemble-based Veracity are thoroughly analysed, and all the observations are noted. According to the experimental results, the throughputs were 2,445,528 bps, 2,391,905 bps and 2,236,778 bps for 20, 50 and 100 nodes, respectively. The experimental results show that ensemble-based Veracity negligibly affects the throughput, queuing time, queue length and number of packets passed to the upper layers of the network and the network performance.
PubDate: 2024-06-21
DOI: 10.1007/s12626-024-00164-4
- Developing a Feature Set from Scene and Texture Features for Detecting
Neural Texture Videos Using Boosted Decision Trees-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract The prevalence of manipulated videos presents a significant challenge in today's era dominated by social media. Various types of fake videos, including notable examples such as Neural Textures, exist. Identifying such deceptive videos is a complex task. This research aims to understand the unique characteristics associated with Neural Texture videos. In the pursuit of comprehending these videos, the study explores the distinguishing traits that define them. The research employs techniques for scene and texture detection to formulate a distinct set of nineteen data features. This feature set is crafted to determine whether a video exhibits Neural Texture characteristics. To validate this set, a standard dataset of video attributes is utilized. These attributes undergo analysis using a machine learning classification model. The results of these experiments are assessed through four distinct methodologies. The evaluation reveals favourable performance outcomes when the machine learning approach and the proposed feature set are used. Based on these findings, it can be concluded that employing the suggested feature set enables the prediction of whether a video displays Neural Texture characteristics. This confirms the hypothesis that a correlation exists between a video's attributes and its authenticity, specifically in determining whether the video qualifies as a Neural Texture.
PubDate: 2024-06-19
DOI: 10.1007/s12626-024-00165-3
- Impact of Computer Usage on Organizational Memory and Learning from
Failure: A Case Study of a Japanese Company-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Few organizations are good at learning from failure, and failures are sometimes repeated. One reason could be the inadequate functioning of organizational memory. According to previous research, IT is often assumed to complement organizational memory. However, studies have also reported that information technology weakens memory; hence, no definite conclusion has been drawn. This study aimed to identify the impact of information technology on organizational memory in the context of organizational learning from failure. We focus on a Japanese company that has successfully implemented the use of information technology for organizational learning from failure. A quantitative analysis was conducted using a questionnaire survey, which revealed the following: (1) if the matter is of high interest in the organization, it is recorded, and knowledge storage and sharing media that employ information technology are used; (2) recording activity is strengthened using knowledge storage and sharing media that employ information technology; (3) even if the matter is of high interest in the organization, it is not always possible to strengthen individual memory, but it is possible to enhance individual memory with knowledge storage/sharing media that utilize information technology; (4) knowledge storage/sharing media that utilize information technology do not directly strengthen organizational memory; and (5) knowledge storage and sharing media using information technology strengthen records and personal memory, which ultimately strengthens organizational memory.
PubDate: 2024-06-14
DOI: 10.1007/s12626-024-00163-5
- All-to-All Data Exchange Method for MANET in Congestion Environment
-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Mobile Ad-hoc Network (MANET) are expected to be a means of communication for disaster victim support systems and information distribution systems at event sites where it is challenging to use existing infrastructure owing to their feature of being able to build a network immediately among nodes that are present on the spot. An operating scenario of such a post-disaster application requires an all-to-all exchange of information that quickly and reliably exchanges a lot of information among a large number of nodes present at the scene to collect more decision-making materials. All-to-all communication places a lot of load on all nodes in the network as they exchange information about all nodes with each other. The MANET routing protocol is also heavily loaded due to route discovery by flooding, which, together with the all-to-all information exchange, is a double challenge to MANET. In this article, we propose an information exchange method that is robust to high-density nodes and high-traffic environments, assuming a situation where communication becomes difficult as a result of requests to send and receive large amounts of information between overcrowded nodes in crowded environments, such as evacuation centers, during a disaster and event sites. The proposed method overcomes the problems of conventional methods and achieves smooth information exchange by applying a delay tolerant network(DTN), a clustering method, and a gossip protocol method. Simulation experiments have shown that the proposed method is highly tolerant of network congestion and increase in the number of nodes due to all-to-all information exchange.
PubDate: 2024-05-20
DOI: 10.1007/s12626-024-00161-7
- Hybrid Load Balancing Technique for Cloud Environment Using Swarm
Optimization-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract One of the most challenging aspects of cloud computing is task scheduling. User needs are changing rapidly in a dynamic environment, and the resources can fluctuate depending on demand because they are virtual. This study presents a hybrid task scheduling model that combines Particle Swarm Optimization and Whale Optimization techniques to address the challenges of task scheduling and achieve the best performance. The method is analyzed and scored based on its “makespan,” “resource utilization,” and “convergence.” Test results indicate that the proposed method reduces the makespan in all cases. Additionally, it increases resource utilization compared to the existing state-of-the-art methods. Furthermore, resource utilization increases across the board with the number of tasks performed.
PubDate: 2024-04-02
DOI: 10.1007/s12626-024-00160-8
- Data Augmentation and Large Language Model for Legal Case Retrieval and
Entailment-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract The Competition on Legal Information Extraction and Entailment (COLIEE) is a well-known international competition organized each year with the goal of applying machine learning algorithms and techniques in the analysis and understanding of legal documents. Two main applications of using machine learning in this domain are entailment and information retrieval. In the realm of legal text analysis, the scarcity of annotated data poses a significant challenge for training robust models. To address this limitation, we employ data augmentation methods to artificially expand the training dataset, enhancing the model’s ability to generalize across diverse legal contexts. Additionally, our approach harnesses the power of a state-of-the-art language model, enabling the extraction of nuanced legal information and improving entailment predictions. We evaluate the performance of our methodology on datasets from the competition, showcasing its effectiveness in achieving competitive results.
PubDate: 2024-03-26
DOI: 10.1007/s12626-024-00158-2
- Preface of Special Issue on 10th Competition on Legal Information of
Extraction and Entailment (COLIEE 2023)-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
PubDate: 2024-03-18
DOI: 10.1007/s12626-024-00159-1
- Contribution Analysis of Large Language Models and Data Augmentations for
Person Names in Solving Legal Bar Examination at COLIEE 2023-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract This paper describes our system for COLIEE 2023 Task 4, which automatically answers Japanese legal bar exam problems. We propose an extension to our previous system in COLIEE 2022, which achieved the highest accuracy among all submissions using data augmentation. We focus on problems that include mentions of person names. In this paper, we present two main contributions. First, we incorporate LUKE as our deep learning component, which is a named entity recognition model trained on RoBERTa. Second, we fine-tune the pretrained LUKE model in multiple ways, comparing fine-tuning on training datasets that include alphabetical person names and ensembling different fine-tuning models. We confirmed that LUKE and its fine-tuned model on person type problems improve their accuracies. Our formal run results show that LUKE and our fine-tuning approach using alphabetical person names were effective, achieving an accuracy of 0.69 in the COLIEE 2023 Task 4 formal run.
PubDate: 2024-03-08
DOI: 10.1007/s12626-024-00155-5
- Does Travel Spread Infection'—Effects of Social Stirring
Simulated on SEIRS Circuit Grid-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract Previous models of the spread of viral infection could not explain the potential risk of non-infectious travelers and exceptional events, such as the reduction in infected cases with an increase in travelers. In this study, we provide an explanation for improving the model by considering two factors. First, we consider the travel of susceptible (S), exposed (E), and recovered (R) individuals who may become infected and infect others in the destination region in the near future, as well as infectious (I). Second, people living in a region and those moving from other regions are treated as separate but interacting groups to consider the potential influence of movement before infection. We show the results of the simulation of infection spread in a country where individuals travel across regions and the government chooses regions to vaccinate with priority. As a result, vaccinating people in regions with larger populations better suppresses the spread of infection, which turns out to be a part of a general law that the same quantity of vaccines can work efficiently by maximizing the conditional entropy Hc of the distribution of vaccines to regions. This strategy outperformed vaccination in regions with a larger effective regeneration number. These results, understandable through the new concept of social stirring, correspond to the fact that travel activities across regional borders may even suppress the spread of vaccination if processed at a sufficiently high pace. This effect can be further reinforced if vaccines are equally distributed to local regions.
PubDate: 2024-03-01
DOI: 10.1007/s12626-024-00156-4
- NOWJ at COLIEE 2023: Multi-task and Ensemble Approaches in Legal
Information Processing-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract This paper presents the NOWJ team’s approach to the COLIEE 2023 Competition, which focuses on advancing legal information processing techniques and applying them to real-world legal scenarios. Our team tackled the four tasks in the competition, which involved legal case retrieval, legal case entailment, statute law retrieval, and legal textual entailment. We employ state-of-the-art machine learning models and innovative approaches, such as BERT, Longformer, BM25-ranking algorithm, and multi-task learning models. Our participation in the COLIEE 2023 has provided useful insights including the importance of the pre-processing and feature engineering, effectiveness of the multi-task models in combining different legal tasks to improve model’s performance. Although our team did not achieve state-of-the-art results, our findings identify areas for further research and improvements in legal information processing.
PubDate: 2024-02-22
DOI: 10.1007/s12626-024-00157-3
- Exploring Prompting Approaches in Legal Textual Entailment
-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract We report explorations into prompt engineering with large pre-trained language models that were not fine-tuned to solve the legal entailment task (Task 4) of the 2023 COLIEE competition. Our most successful strategy used simple text similarity measures to retrieve articles and queries from the training set. We report on our efforts to optimize performance with both OpenAI’s GPT-4 and FLaN-T5. We also used an ensemble approach to find the best combination of models and prompts. Finally, we analyze our results and suggest ideas for future improvements.
PubDate: 2024-01-23
DOI: 10.1007/s12626-023-00154-y
- Overview and Discussion of the Competition on Legal Information,
Extraction/Entailment (COLIEE) 2023-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract We summarize the 10th Competition on Legal Information Extraction and Entailment. In this tenth edition, the competition included four tasks on case law and statute law. The case law component includes an information retrieval task (Task 1), and the confirmation of an entailment relation between an existing case and a selected unseen case (Task 2). The statute law component includes an information retrieval task (Task 3), and an entailment/question-answering task based on retrieved civil code statutes (Task 4). Participation was open to any group based on any approach. Ten different teams participated in the case law competition tasks, most of them in more than one task. We received results from 8 teams for Task 1 (22 runs) and seven teams for Task 2 (18 runs). On the statute law task, there were 9 different teams participating, most in more than one task. 6 teams submitted a total of 16 runs for Task 3, and 9 teams submitted a total of 26 runs for Task 4. We describe the variety of approaches, our official evaluation, and analysis of our data and submission results.
PubDate: 2024-01-12
DOI: 10.1007/s12626-023-00152-0
- Legal Information Retrieval and Entailment Using Transformer-based
Approaches-
Free pre-print version: Loading...
Rate this result:
What is this?
Please help us test our new pre-print finding feature by giving the pre-print link a rating.
A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Abstract The challenge of information overload in the legal domain increases every day. The COLIEE competition has created four challenge tasks that are intended to encourage the development of systems and methods to alleviate some of that pressure: a case law retrieval (Task 1) and entailment (Task 2), and a statute law retrieval (Task 3) and entailment (Task 4). Here we describe our methods for Task 1 and Task 4. In Task 1, we used a sentence-transformer model to create a numeric representation for each case paragraph. We then created a histogram of the similarities between a query case and a candidate case. The histogram is used to build a binary classifier that decides whether a candidate case should be noticed or not. In Task 4, our approach relies on fine-tuning a pre-trained DeBERTa large language model (LLM) trained on SNLI and MultiNLI datasets. Our method for Task 4 was ranked third among eight participating teams in the COLIEE 2023 competition. For Task 4, We also compared the performance of the DeBERTa model with those of a knowledge distillation model and ensemble methods including Random Forest and Voting.
PubDate: 2024-01-11
DOI: 10.1007/s12626-023-00153-z
|
Hybrid journal (It can contain Open Access articles)
