Document Classification Research Articles

Modelling Misinformation in Swahili-English Code-switched Texts

Code-switching, which is the mixing of words or phrases from multiple, grammatically distinct languages, introduces semantic and syntactic complexities to sentences which complicate automated text classification. Despite code-switching being a common occurrence in informal text-based communication among most bilingual or multilingual users of digital spaces, its use to spread misinformation is relatively less explored. In Kenya, for instance, the use of code-switched Swahili-English is prevalent on social media. Our main objective in this paper was to systematically re- view code-switching, particularly the use of Swahili-English code-switching to spread misinformation on social media in the Kenyan context. Additionally, we aimed at pre-processing a Swahili-English code-switched dataset and developing a misinformation classification model trained on this dataset. We discuss the process we took to develop the code- switched Swahili-English misinformation classification model. The model was trained and tested using the PolitiKweli dataset which is the first Swahili-English code-switched dataset curated for misinformation classification. The dataset was collected from Twitter (now X) social media platform, focusing on text posted during the electioneering period of the 2022 general elections in Kenya. The study experimented with two types of word embeddings - GloVe and FastText. FastText uses character n-gram representations that help generate meaningful vectors for rare and unseen words in the code-switched dataset. We experimented with both the classical machine learning algorithms and deep learning algo- rithms. Bidirectional Long Short-Term Memory Networks (BiLSTM) algorithm showed the best performance with an f-score of 0.89. The model was able to classify code-switched Swahili-English political misinformation text as fake, fact or neutral. This study contributes to recent research efforts in developing language models for low-resource languages.

Tackling Real-world Complexity: Hierarchical Modeling and Dynamic Prompting for Multimodal Long Document Classification

Tackling Real-world Complexity: Hierarchical Modeling and Dynamic Prompting for Multimodal Long Document Classification

Artificial Intelligence Methods for Spanish Documents Classification

La rápida globalización y la creciente necesidad de comunicación interlingüística requieren corpus modernos y en tiempo real para ayudar a los estudiantes de idiomas. Los métodos tradicionales para crear dichos corpus, especialmente en español, son inadecuados debido a su incapacidad para procesar la gran cantidad de datos no estructurados disponibles en internet. En este estudio se exploran las metodologías de inteligencia artificial (IA) para la adquisición automática de documentos en español de la web, preprocesándolos y clasificándolos con el fin de construir un corpus vasto y flexible para el aprendizaje del español. La investigación aplica el rastreo web mediante el framework Scrapy para recopilar datos, que luego se limpian y clasifican utilizando modelos avanzados de procesamiento del lenguaje natural (PLN). En concreto, el estudio emplea el algoritmo BERT (Bidirectional Encoder Representations from Transformers) y su variante mejorada RoBERTa para lograr la clasificación de documentos. Mediante una combinación de técnicas de aumento de datos y modelos de aprendizaje profundo, el estudio logra una alta precisión en la clasificación de texto en español, lo que demuestra el potencial del uso de la IA para superar las limitaciones de los enfoques tradicionales de creación de corpus.

A Deep Learning Framework for Automated Classification and Archiving of Orthodontic Diagnostic Documents.

Background Orthodontic diagnostic workflows often rely on manual classification and archiving of large volumes of patient images, a process that is both time-consuming and prone to errors such as mislabeling and incomplete documentation. These challenges can compromise treatment accuracy and overall patient care. To address these issues, we propose an artificial intelligence (AI)-driven deep learning framework based on convolutional neural networks (CNNs) to automate the classification and archiving of orthodontic diagnostic images. Our AI-based framework enhances workflow efficiency and reduces human errors. This study is an initial step towards fully automating orthodontic diagnosis and treatment planning systems, specifically focusing on the automation of orthodontic diagnostic record classification using AI. Methods This study employed a dataset comprising 61,842 images collected from three dental clinics, distributed across 13 categories. A sequential classification approach was developed, starting with a primary model that categorized images into three main groups: extraoral, intraoral, and radiographic. Secondary models were applied within each group to perform the final classification. The proposed model, enhanced with attention modules, was trained and compared with pre-trained models such as ResNet50 (Microsoft Corporation, Redmond, Washington, United States) and InceptionV3 (Google LLC, Mountain View, California, United States). External validation was performed using 13,729 new samples to assess the artificial intelligence (AI) system's accuracy and generalizability compared to expert assessments. Results The deep learning framework achieved an accuracy of 99.24% on an external validation set, demonstrating performance almost on par with human experts. Additionally, the model demonstrated significantly faster processing times compared to manual methods. Gradient-weighted class activation mapping (Grad-CAM) visualizations confirmed that the model effectively focused on clinically relevant features during classification, further supporting its clinical applicability. Conclusion This study introduces a deep learning framework for automating the classification and archiving of orthodontic diagnostic images. The model achieved impressive accuracy and demonstrated clinically relevant feature focus through Grad-CAM visualizations. Beyond its high accuracy, the framework offers significant improvements in processing speed, making it a viable tool for real-time applications in orthodontics. This approach not only reduces the workload in healthcare settings but also lays the foundation for future automated diagnostic and treatment planning systems in digital orthodontics.

Software Performance Optimization for Classification and Linking of Administrative Documents

Software Performance Optimization for Classification and Linking of Administrative Documents

Automatic TNM staging of colorectal cancer radiology reports using pre-trained language models

Background and Objective:Colorectal cancer is one of the major causes of cancer death worldwide. Essential for prognosis and treatment planning, TNM staging offers critical insights into the advancement of colorectal cancer. However, manual TNM staging from colon magnetic resonance imaging (MRI) is a laborious and error prone process. This study introduces an automated text classification system for TNM staging of colon MRI images in Spanish. Methods:A dataset of 1319 Spanish colon MRI reports was collected and manually labeled with TNM staging. In order to automate the task of TNM staging, a multimodal system was proposed. The system is based on RoBERTa language model pre-trained on a combination of biomedical and clinical Spanish language corpora and uses Natural Language Processing (NLP) techniques to extract relevant categorical and numerical features from MRI reports. Results:The performance of the system was evaluated using different metrics and the results obtained are very promising: the best performance among the proposed systems reached 0.7464, 0.8792 and 0.6776 of macro F1-score for T, N and M respectively. Conclusions:This study demonstrates the feasibility of using a language model for automatic TNM staging based on Spanish clinical reports of colorectal cancer patients. The proposed system can be a useful tool to improve the efficiency and accuracy of colorectal cancer diagnosis.

Automated classification of online reviews of otolaryngologists.

The study aimed to extract online comments of otolaryngologists in the 20 most populated cities in the United States from healthgrades.com, develop and validate a natural language processing (NLP) logistic regression algorithm for automated text classification of reviews into 10 categories, and compare 1- and 5-star reviews in directly-physician-related and non-physician-related categories. 1977 1-star and 12,682 5-star reviews were collected. The primary investigator manually categorized a training dataset of 324 1-star and 909 5-star reviews, while a validation subset of 100 5-star and 50 1-star reviews underwent dual manual categorization. Using scikit-learn, an NLP algorithm was trained and validated on the subsets, with F1 scores evaluating text classification accuracy against manual categorization. The algorithm was then applied to the entire dataset with comparison of review categorization according to 1- and 5-star reviews. F1 scores for NLP validation ranged between 0.71 and 0.97. Significant associations emerged between 1-star reviews and treatment plan, accessibility, wait time, office scheduling, billing, and facilities. 5-star reviews were associated with surgery/procedure, bedside manner, and staff/mid-levels. The study successfully validated an NLP text classification system for categorizing online physician reviews. Positive reviews were found to have an association with directly-physician related context. 1-star reviews were related to treatment plan, accessibility, wait time, office scheduling, billing, and facilities. This method of text classification effectively discerned the nuances of human-written text, providing valuable insights into online healthcare feedback that is scalable.Level of evidence: Level 3.

Remote Assistance for Bone-Fractured Patients using Deep Learning Models

ABSTRACT Remote diagnosis enables healthcare professionals to evaluate and diagnose patients from a distance using telecommunication technologies, enhancing healthcare delivery by improving accessibility, especially for those in remote or underserved areas. One of the significant sustainability challenges in remote medical diagnostics is offering timely assistance to vulnerable groups like the elderly, disabled, mentally impaired individuals, and wounded military personnel in combat zones. This becomes particularly difficult in emergencies when rapid analysis of medical records is needed, especially if the data is stored on secure blockchain networks. The proposed work addresses these challenges by deploying a comprehensive framework for large-scale analysis, utilizing both document and image classification for dual validation. It integrates advanced techniques such as Inception V3, VGG-16, VGG-19, RESNET-50, and Densenet-201 for bone fracture detection, with Inception V3 achieving the highest accuracy of 95.1%. In addition, a Document Classification Analysis (DCA) method is proposed, which automatically classifies the severity of fractures. Object detection techniques are also introduced for detecting minor fractures using region-based image segmentation, ensuring precise diagnosis even for subtle injuries. This pioneering integration of technologies provides a holistic solution for remote medical diagnostics.

EXPLOITING OPTIMAL SELF-ATTENTION DEEP LEARNING-BASED RECOGNITION OF TEXTUAL EMOTIONS FOR DISABLED PERSONS

A disability is a significant issue that has posed and continues to pose a challenge. Disability is a basis of frustration because it can be observed as a mental, constraint, cognitive, and physical handicap that inhibits the individual’s growth and involvement. Consequently, significant effort has been put into removing these kinds of restrictions. These initiatives address the trouble that disabled people encounter. People with disabilities often need to rely on others to meet their requirements. Machine learning (ML) is excelling in producing smart cities and offering a secure environment for disabled individuals. Emotional detection is an important research domain that can expose many appreciated inputs. Emotion is expressed differently through speech and facial expressions, gestures, and written text. Emotion detection in a text document is fundamentally a content-based classification task, utilizing models from deep learning (DL), complex systems and natural language processing (NLP). This paper presents an Optimal Self-Attention DL-based Recognition of Textual Emotions (OSADL-RTE) technique for Disabled Persons. The presented OSADL-RTE technique focuses on identifying distinct types of emotions in the textual data. As a primary preprocessing step, the OSADL-RTE technique comprises different phases to transform the input in a useful way. For word embedding, the bag of words (BoWs) approach is exploited. The OSADL-RTE technique derives self-attention long short-term memory (SA-LSTM) approach to identify emotions. Lastly, the arithmetic fractals optimization algorithm (AOA) approach correctly tunes the hyperparameter selection of the SA-LSTM approach. The experimental study of the OSADL-RTE approach occurs on the emotion database. The investigational outcome of the OSADL-RTE approach portrayed a superior accuracy outcome of 99.59% over existing methods.

Effective Document Classification using Novel Enhanced Long Short Term Memory-based Namib Beetle Optimization Algorithm

Effective Document Classification using Novel Enhanced Long Short Term Memory-based Namib Beetle Optimization Algorithm

Automated text classification of opinion vs. news French press articles. A comparison of transformer and feature-based approaches

This study explores Natural Language Processing (NLP) methods for distinguishing between press articles belonging to the journalistic genres of ‘objective’ news and ‘subjective’ opinion. Two classification models are compared: CamemBERT, a French transformer model fine-tuned for the task, and a machine learning model using 32 linguistic features. Trained on 8000 Belgian French articles, both models are evaluated on 1000 Canadian French articles. Results show CamemBERT’s superiority but highlight potential for hybrid approaches and emphasizes the need for robust and transparent methods in NLP. The research contributes to understanding NLP’s role in journalism by addressing challenges of point of view detection in press discourse.

Perioperative Hypotension in Chronic Kidney Disease Patients with Dialysis Undergoing Noncardiac Surgery: A Retrospective Cohort Study.

Patients with chronic kidney disease who require maintenance dialysis suffer excess morbidity and mortality for reasons that are not clearly elucidated. There are few targets for intervention to improve their outcomes. The authors hypothesized that perioperative hypotension is more common in patients receiving dialysis. A retrospective cohort study was conducted of adult patients who had inpatient noncardiac surgery lasting greater than 2 h with general anesthesia between January 2012 and December 2021 at the Cleveland Clinic (Cleveland, Ohio). Using International Classification of Diseases codes and dialysis documentation in the electronic medical record, was derived age, American Society of Anesthesiologists (Schaumburg, Illinois) Physical Status, surgery type, and propensity score (predicting dialysis dependence) matched groups of patients (1) with dialysis-dependent chronic kidney disease, (2) with nondialysis chronic kidney disease, and (3) without kidney disease. The primary outcomes were total area under the curve of mean arterial pressure (MAP) less than 65 mmHg during the surgery and postoperative MAP less than 70 mmHg during 48 h after surgery. Three 1:1:1 matched groups of 1,886 patients (total, 5,658 patients) from an overall cohort of 123,761 were derived. Cases with dialysis-dependent kidney disease had a greater intraoperative area under the curve of MAP less than 65 mmHg (difference in medians, 18.4 mmHg-min 98.75% CI, 11.3 to 25.6; P < 0.001) when compared to patients who had nondialysis kidney disease and when compared to patients without kidney disease (difference in medians, 15.5 mmHg-min; 98.75% CI, 6.6 to 24.4; P < 0.001). Patients receiving preoperative dialysis were also more likely to have postoperative MAP less than 70 mmHg compared to patients with nondialysis kidney disease (rate ratio, 1.52; 98.75% CI, 1.48 to 1.57; P < 0.0001) and compared to patients without kidney disease (rate ratio, 1.43; 98.75% CI, 1.38 to 1.47; P < 0.0001). Perioperative hypotension is more severe and common among patients who undergo chronic maintenance dialysis compared to similar patients without dialysis dependence. The management of hemodynamics in this population comes with unique considerations that warrant further systematic evaluation.

HaaS in Environmental Computing: Hadoop-as-a-Service for Big Data Mining in Environmental Computing Applications

This article addresses the importance of HaaS (Hadoop-as-a-Service) in cloud technologies, with specific reference to its usefulness in big data mining for environmental computing applications. The term environmental computing refers to computational analysis within environmental science and management, encompassing a myriad of techniques, especially in data mining and machine learning. As is well-known, the classical MapReduce has been adapted within many applications for big data storage and information retrieval. Hadoop based tools such as Hive and Mahout are broadly accessible over the cloud and can be helpful in data warehousing and data mining over big data in various domains. In this article, we explore HaaS technologies, mainly based on Apache's Hive and Mahout for applications in environmental computing, considering publicly available data on the Web. We dwell upon interesting applications such as automated text classification for energy management, recommender systems for ecofriendly products, and decision support in urban planning. We briefly explain the classical paradigms of MapReduce, Hadoop and Hive, further delve into data mining and machine learning over the MapReduce framework, and explore techniques such as Naïve Bayes and Random Forests using Apache Mahout with respect to the targeted applications. Hence, the paradigm of Hadoop-as-a-Service, popularly referred to as HaaS, is emphasized here as per its benefits in a domain-specific context. The studies in environmental computing, as presented in this article, can be useful in other domains as well, considering similar applications. This article can thus be interesting to professionals in web technologies, cloud computing, environmental management, as well as AI and data science in general.

Semantic Term Weighting Representation for Kannada Document Classification

Semantic Term Weighting Representation for Kannada Document Classification

Relationship of the modified Glasgow Prognostic Score with peripheral artery disease severity and procedure success in patients who had undergone endovascular treatment.

The modified Glasgow Prognostic Score (mGPS) is one type of inflammation-based index; it includes data on elevated C-reactive protein and reduced albumin content. The predictive value of mGPS for outcomes is investigated in various diseases such as cancer, heart failure, myocardial infarction, acute pulmonary embolism, and inflammatory bowel diseases. This study aimed to evaluate the link between mGPS and the severity and complexity of peripheral arterial disease (PAD) as determined by the Transatlantic Intercommunal Consensus Document (TASC-II) classification and the prediction value of mGPS for procedural success in patients undergoing endovascular treatment (EVT). Our study included 203 consecutive patients receiving EVT for atherosclerotic obstruction of aortoiliac, femoro-popliteal, and below-knee arteries between January 2019 and February 2020. The lesion characteristics were determined according to categories in the TASC-II. Operational failure is the inability to position the guidewire through the occluded lesion following percutaneous intervention or achieve distal perfusion following EVT. In our study, we observed 136 patients (%6) with TASC A-B lesions and 67 patients (%33) with TASC C-D lesions. EVT was performed on the femoro-popliteal artery in 59.4% of the patients, on the aortoiliac artery in 30.7%, and on the below-the-knee artery in 9.9%. mGPS was an independent predictor of severe PAD (OR: 17.943, 95% CI: 5.120-62.882; p < .001) and procedural success (odds ratio: 0.004; 95% CI: 0.001-0.099; p < .001). Additionally, we identified age and the presence of a TASC D lesion as independent predictors of interventional success (OR: 0.938, 95% CI: 0.819-0.979; p: .034; OR: 0.104, 95% CI: 0.107-0.643; p: .015, respectively). We determined that mGPS independently predicts PAD complexity and severity based on TASC-II classification; the EVT success rate is lower in patients with high mGPS.

Improving Probabilistic Models In Text Classification Via Active Learning

Social scientists often classify text documents to use the resulting labels as an outcome or a predictor in empirical research. Automated text classification has become a standard tool since it requires less human coding. However, scholars still need many human-labeled documents for training. To reduce labeling costs, we propose a new algorithm for text classification that combines a probabilistic model with active learning. The probabilistic model uses both labeled and unlabeled data, and active learning concentrates labeling efforts on difficult documents to classify. Our validation study shows that with few labeled data, the classification performance of our algorithm is comparable to state-of-the-art methods at a fraction of the computational cost. We replicate the results of two published articles with only a small fraction of the original labeled data used in those studies and provide open-source software to implement our method.

The incremental design of a machine learning framework for medical records processing.

This work presents the development and evaluation of coordn8, a web-based application that streamlines fax processing in outpatient clinics using a "human-in-the-loop" machine learning framework. We demonstrate the effectiveness of the platform at reducing fax processing time and producing accurate machine learning inferences across the tasks of patient identification, document classification, spam classification, and duplicate document detection. We deployed coordn8 in 11 outpatient clinics and conducted a time savings analysis by observing users and measuring fax processing event logs. We used statistical methods to evaluate the machine learning components across different datasets to show generalizability. We conducted a time series analysis to show variations in model performance as new clinics were onboarded and to demonstrate our approach to mitigating model drift. Our observation analysis showed a mean reduction in individual fax processing time by 147.5 s, while our event log analysis of over 7000 faxes reinforced this finding. Document classification produced an accuracy of 81.6%, patient identification produced an accuracy of 83.7%, spam classification produced an accuracy of 98.4%, and duplicate document detection produced a precision of 81.0%. Retraining document classification increased accuracy by 10.2%. coordn8 significantly decreased fax-processing time and produced accurate machine learning inferences. Our human-in-the-loop framework facilitated the collection of high-quality data necessary for model training. Expanding to new clinics correlated with performance decline, which was mitigated through model retraining. Our framework for automating clinical tasks with machine learning offers a template for health systems looking to implement similar technologies.

Generative artificial intelligence in topic-sentiment classification for Arabic text: a comparative study with possible future directions

Social media platforms have become essential for disseminating news and expressing individual sentiments on various life topics. Arabic, widely used in the Middle East, presents unique challenges for sentiment analysis due to its complexity and multiple dialects. Motivated by the need to address these challenges, this article develops methods to overcome the lack of topic-based labeling techniques, compares different approaches for preparing extensive, annotated datasets, and analyzes the efficacy of machine learning (ML), deep learning (DL), and large language models (LLMs) in classifying Arabic textual data. Our research utilizes the topic-modeling technique to build a topic-based sentiment dataset of Arabic texts aimed at enhancing our understanding and processing capabilities. We present a comprehensive evaluation of dataset balancing techniques, including under-sampling, over-sampling, and using imbalanced datasets, providing insights into how these approaches impact classification outcomes. Additionally, we explore the influence of dataset sizes on the performance of various ML models, highlighting the importance of dataset scale in developing effective Arabic NLP applications. A further focus of our study is the comparative analysis of generative artificial intelligence (AI) models, including the emerging ChatGPT LLM, assessing their effectiveness in managing the complexities of Arabic language classification tasks. Our results show that support vector machines (SVM) achieved the highest performance, with F1-scores of 0.97 and 0.96 in classifying sentiment and topic, respectively, in Arabic tweets. This research not only benchmarks existing methodologies but also paves the way for more nuanced and robust models in the future, enhancing the application of generative AI in Arabic topic-based sentiment analysis.

Lymphedema Rates Following Axillary Lymph Node Dissection With and Without Immediate Lymphatic Reconstruction: A Prospective Trial.

Immediate lymphatic reconstruction (ILR) has been proposed to decrease lymphedema rates. The primary aim of our study was to determine whether ILR decreased the incidence of lymphedema in patients undergoing axillary lymph node dissection (ALND). We conducted a two-site pragmatic study of ALND with or without ILR, employing surgeon-level cohort assignment, based on breast surgeons' preferred standard practice. Lymphedema was assessed by limb volume measurements, patient self-reporting, provider documentation, and International Classification of Diseases, Tenth Revision (ICD-10) codes. Overall, 230 patients with breast cancer were enrolled; on an intention-to-treat basis, 99 underwent ALND and 131 underwent ALND with ILR. Of the 131 patients preoperatively planned for ILR, 115 (87.8%) underwent ILR; 72 (62.6%) were performed by one breast surgical oncologist and 43 (37.4%) by fellowship-trained microvascular plastic surgeons. ILR was associated with an increased risk of lymphedema when defined as ≥10% limb volume change on univariable analysis, but not on multivariable analysis, after propensity score adjustment. We did not find a statistically significant difference in limb volume measurements between the two cohorts when including subclinical lymphedema (≥5% inter-limb volume change), nor did we see a difference in grade between the two cohorts on an intent-to-treat or treatment received basis. For all patients, considering ascertainment strategies of patient self-reporting, provider documentation, and ICD-10 codes, as a single binary outcome measure, there was no significant difference in lymphedema rates between those undergoing ILR or not. We found no significant difference in lymphedema rates between patients undergoing ALND with or without ILR.

Feature Extraction of Nearest Neighbors based on Large-scale Language Model for Document Classification

Feature Extraction of Nearest Neighbors based on Large-scale Language Model for Document Classification