Performance Of Text Classification Research Articles

A novel method of detecting malware on Android mobile devices with explainable artificial intelligence

The increasing prevalence of malware targeting android mobile devices has raised significant concerns regarding user privacy and security. In response, effective methods for malware classification and detection are crucial to protect users from malicious applications. This paper presents an approach that leverages deep learning techniques and explainable artificial intelligence (XAI) for android mobile malware classification and detection. Convolutional neural networks (CNNs) are deep learning model that has shown impressive performance in several application areas, including image and text classification. In the context of android mobile malware, CNNs have shown promising results in capturing intricate patterns and features inherent in malware samples. By training these models on large datasets of benign and malicious applications, accurate classification can be achieved. To enhance transparency and interpretability, XAI techniques are integrated into the classification process. These techniques provide insights into the decision-making process of the deep learning models, enabling the identification of critical features and characteristics that contribute to the classification results. This research, by combining deep learning and XAI methods, presents a fresh strategy for identifying and categorizing Android malware. This research paper will focus on a fascinating CNN-based malware categorization technique.

OPTIMIZING LONG TEXT CLASSIFICATION PERFORMANCE THROUGH KEYWORD-BASED SENTENCE SELECTION: A CASE STUDY ON ONLINE NEWS CLASSIFICATION FOR INDONESIAN GDP GROWTH-RATE DETECTION

Efficiently managing lengthy textual data, particularly in online news, is crucial for enhancing the performance of long text classification. This study delves into innovative approaches to streamline the Gross Domestic Product (GDP) computation process by harnessing modern data analytics, Natural Language Processing (NLP), and online news sources. Leveraging online news data introduces real-time information, promising to improve the accuracy and timeliness of economic indicators like GDP. However, handling the complexity of extensive textual data poses a challenge, demanding advanced NLP techniques. This research shifts from traditional word-weight-based methods to keyword-based extractive summarization techniques to address this. These tailored approaches ensure that selected sentences align precisely with specific keywords relevant to the research case, such as GDP growth rate detection. The study emphasizes the necessity of adapting summarization methods to capture information in unique research contexts effectively. According to classification results, the implementation of sentence selection successfully demonstrated improved performance in terms of classification accuracy. Specifically, there was an average accuracy increase of 0.0226 for machine learning and 0.0164 for transfer learning models. Additionally, in terms of computational efficiency, sentence selection also accelerates processing time during hyperparameter tuning and fine-tuning, as observed using the same computational resources.

Harnessing large language models over transformer models for detecting Bengali depressive social media text: A comprehensive study

In an era where the silent struggle of underdiagnosed depression pervades globally, our research delves into the crucial link between mental health and social media. This work focuses on early detection of depression, particularly in extroverted social media users, using LLMs such as GPT 3.5, GPT 4 and our proposed GPT 3.5 fine-tuned model DepGPT, as well as advanced Deep learning models(LSTM, Bi-LSTM, GRU, BiGRU) and Transformer models(BERT, BanglaBERT, SahajBERT, BanglaBERT-Base). The study categorized Reddit and X datasets into “Depressive” and “Non-Depressive” segments, translated into Bengali by native speakers with expertise in mental health, resulting in the creation of the Bengali Social Media Depressive Dataset (BSMDD). Our work provides full architecture details for each model and a methodical way to assess their performance in Bengali depressive text categorization using zero-shot and few-shot learning techniques. Our work demonstrates the superiority of SahajBERT and Bi-LSTM with FastText embeddings in their respective domains also tackles explainability issues with transformer models and emphasizes the effectiveness of LLMs, especially DepGPT (GPT 3.5 fine-tuned), demonstrating flexibility and competence in a range of learning contexts. According to the experiment results, the proposed model, DepGPT, outperformed not only Alpaca Lora 7B in zero-shot and few-shot scenarios but also every other model, achieving a near-perfect accuracy of 0.9796 and an F1-score of 0.9804, high recall, and exceptional precision. Although competitive, GPT-3.5 Turbo and Alpaca Lora 7B show relatively poorer effectiveness in zero-shot and few-shot situations. The work emphasizes the effectiveness and flexibility of LLMs in a variety of linguistic circumstances, providing insightful information about the complex field of depression detection models.

SPContrastNet: A Self-Paced Contrastive Learning Model for Few-Shot Text Classification

Meta-learning has recently promoted few-shot text classification, which identifies target classes based on information transferred from source classes through a series of small tasks or episodes. Existing works constructing their meta-learner on Prototypical Networks need improvement in learning discriminative text representations between similar classes that may lead to conflicts in label prediction. The overfitting problems caused by a few training instances need to be adequately addressed. In addition, efficient episode sampling procedures that could enhance few-shot training should be utilized. To address the problems mentioned above, we first present a contrastive learning framework that simultaneously learns discriminative text representations via supervised contrastive learning while mitigating the overfitting problem via unsupervised contrastive regularization, and then we build an efficient self-paced episode sampling approach on top of it to include more difficult episodes as training progresses. Empirical results on eight few-shot text classification datasets show that our model outperforms the current state-of-the-art models. The extensive experimental analysis demonstrates that our supervised contrastive representation learning and unsupervised contrastive regularization techniques improve the performance of few-shot text classification. The episode-sampling analysis reveals that our self-paced sampling strategy improves training efficiency.

A Heterogeneous Directed Graph Attention Network for inductive text classification using multilevel semantic embeddings

In the current study, a novel network model is proposed for text classification based on Graph Attention Networks (GATs) and sentence-transformer embeddings. Most existing methods with a pretraining model as an input layer still treat words as the minimum processing unit. However, word embedding is not an efficient and appropriate solution when dealing with long texts containing many professional words. This study aims to design a model capable of handling text classification tasks at multilevel semantic segmentation. The main contribution of this study is that a novel GAT variant is designed using global nodes and Squeeze-and-Excitation Networks (SENet) to capture semantic information. Moreover, a novel unidirectional attention mechanism is introduced for our model to avoid the message passing of irrelevant noisy information within global nodes. The numerical results show that according to the characteristics of datasets, specific semantic information combinations can effectively improve the accuracy and performance of text classification. Without fine-tuning of the pretrained encoder, the new state-of-the-art performance is achieved on three benchmark datasets. In addition, a comprehensive analysis of the graph attention mechanism in the model for specific cases suggests that the unidirectional attention mechanism and the use of global nodes are key contributing factors to multilevel semantic fusion.

Comparative Analysis of XGBoost Performance for Text Classification with CPU Parallel and Non-Parallel Processing

This paper shows the findings of a study that looks at how CPU parallel processing changes the way Extreme Gradient Boosting (XGBoost) classifies text. XGBoost models can sort news stories into set groups faster and more accurately, with or without CPU parallelism. This is the main goal of the study. The Keras dataset is used to prepare the text so that the TF-IDF (Term Frequency-Inverse Document Frequency) features can be found. These features will then be used to train the XGBoost model. This is used to check out two different kinds of the XGBoost classifier. There is parallelism between one of them and not it in the other. How well the model works can be observed by how accurate it is. This includes both how long it takes to learn and estimate and how well predictions work. The models take very different amounts of time to compute, but they are all pretty close in terms of how accurate they are. Parallel processing on the CPU has made tasks proceed more rapidly, and XGBoost is now better at making the most of that speed to do its task. The purpose of the study is to show that parallel processing can speed up XGBoost models without affecting their accuracy. This is helpful for putting text into categories.

Comparing Hierarchical Approaches to Enhance Supervised Emotive Text Classification

The performance of emotive text classification using affective hierarchical schemes (e.g., WordNet-Affect) is often evaluated using the same traditional measures used to evaluate the performance of when a finite set of isolated classes are used. However, applying such measures means the full characteristics and structure of the emotive hierarchical scheme are not considered. Thus, the overall performance of emotive text classification using emotion hierarchical schemes is often inaccurately reported and may lead to ineffective information retrieval and decision making. This paper provides a comparative investigation into how methods used in hierarchical classification problems in other domains, which extend traditional evaluation metrics to consider the characteristics of the hierarchical classification scheme, can be applied and subsequently improve the classification of emotive texts. This study investigates the classification performance of three widely used classifiers, Naive Bayes, J48 Decision Tree, and SVM, following the application of the aforementioned methods. The results demonstrated that all the methods improved the emotion classification. However, the most notable improvement was recorded when a depth-based method was applied to both the testing and validation data, where the precision, recall, and F1-score were significantly improved by around 70 percentage points for each classifier.

Cigarette tasting Chinese text classification for low-resource scenarios

At present, the Chinese text field is facing challenges from low resource issues such as data scarcity and annotation difficulties. Moreover, in the domain of cigarette tasting, cigarette tasting texts tend to be colloquial, making it difficult to obtain valuable and high-quality tasting texts. Therefore, in this paper, we construct a cigarette tasting dataset (CT2023) and propose a novel Chinese text classification method based on ERNIE and Comparative Learning for Low-Resource scenarios (ECLLR). Firstly, to address the issues of limited vocabulary diversity and sparse features in cigarette tasting text, we utilize Term Frequency-Inverse Document Frequency (TF-IDF) to extract key terms, supplementing the discriminative features of the original text. Secondly, ERNIE is employed to obtain sentence-level vector embedding of the text. Finally, contrastive learning model is used to further refine the text after fusing the keyword features, thereby enhancing the performance of the proposed text classification model. Experiments on the CT2023 dataset demonstrate an accuracy rate of 96.33% for the proposed method, surpassing the baseline model by at least 11 percentage points, and showing good text classification performance. It is thus clear that the proposed approach can effectively provide recommendations and decision support for cigarette production processes in tobacco companies.

Machine learning for text document classification-efficient classification approach

<p>Numerous alternative methods for text classification have been created because of the increase in the amount of online text information available. The cosine similarity classifier is the most extensively utilized simple and efficient approach. It improves text classification performance. It is combined with estimated values provided by conventional classifiers such as Multinomial Naive Bayesian (MNB). Consequently, combining the similarity between a test document and a category with the estimated value for the category enhances the performance of the classifier. This approach provides a text document categorization method that is both efficient and effective. In addition, methods for determining the proper relationship between a set of words in a document and its document categorization is also obtained.</p>

A Class Specific Feature Selection Method for Improving the Performance of Text Classification

Recently, a significant amount of research work has been carried out in the field of feature selection. Although these methods help to increase the accuracy of the machine learning classification, the selected subset of features considers all the classes and may not select recommendable features for a particular class. The main goal of our paper is to propose a new class-specific feature selection algorithm that is capable of selecting an appropriate subset of features for each class. In this regard, we first perform class binarization and then select the best features for each class. During the feature selection process, we deal with class imbalance problems and redundancy elimination. The Weighted Average Voting Ensemble method is used for the final classification. Finally, we carry out experiments to compare our proposed feature selection approach with the existing popular feature selection methods. The results prove that our feature selection method outperforms the existing methods with an accuracy of more than 37%.

Enhanced text classification through an improved discrete laying chicken algorithm

AbstractThe exponential growth of digital text documents presents a significant challenge for text classification algorithms, as the vast number of words in each document can hinder their efficiency. Feature selection (FS) is a crucial technique that aims to eliminate irrelevant features and enhance classification accuracy. In this study, we propose an improved version of the discrete laying chicken algorithm (IDLCA) that utilizes noun‐based filtering to reduce the number of features and improve text classification performance. Although LCA is a newly proposed algorithm, it has not been systematically applied to discrete problems before. Our enhanced version of LCA employs different operators to improve both exploration and exploitation of this algorithm to find better solutions in discrete mode. To evaluate the effectiveness of the proposed method, we compared it with some conventional nature‐inspired feature selection methods using various learning models such as decision trees (DT), K‐nearest neighbor (KNN), Naive Bayes (NB), and support vector machine (SVM) on five benchmark datasets with three different evaluation metrics. The experimental results demonstrate the effectiveness of the proposed algorithm in comparison to the existing one. The code is available at https://github.com/m0javad/Improved-Discrete-Laying-Chicken-Algorithm.

MCICT: Graph convolutional network-based end-to-end model for multi-label classification of imbalanced clinical text

The rapid growth of clinical text data requires accurate and powerful automated classification methods to support medical decision making and personalized healthcare. The multi-label classification task for clinical texts is designed to assign the most relevant set of labels to each clinical text. However, this task presents two significant challenges: (1) how to accurately extract fine-grained semantic features from complex clinical texts, and (2) how to effectively mitigate the issue of label class imbalance. To overcome these problems, we innovatively propose a novel Multi-label Classification of Imbalanced Clinical Text (MCICIT) model. In order to obtain fine-grained semantic features from clinical texts, we utilize the specialized pre-trained language model BioBERT, tailored for biomedical texts. To tackle the challenge of label class imbalance, we present a Co-occurrence Based and Embeddings with Additional Information Enhanced Graph Convolutional Network (CoEAI-GCN) module. On one hand, we enrich the label content by incorporating additional information to acquire more accurate word embeddings as the feature matrix. On the other hand, we combine the co-occurrence relationship of labels to construct a correlation matrix. Ultimately, label representations are learned through a graph convolutional network. By conducting multi-label classification experiments on two clinical text datasets extracted from real medical systems, our model achieves a 3.2% and 0.5% improvement in F1 scores, respectively, compared to state-of-the-art deep learning models. Additionally, we conduct ablation studies to explore the behaviors of the proposed model. These results together demonstrate that our proposed MCICT effectively enhances the classification performance of imbalanced clinical texts.

Dual-Mode Method for Generating Adversarial Examples to Attack Deep Neural Networks

Deep neural networks yield desirable performance in text, image, and speech classification. However, these networks are vulnerable to adversarial examples. An adversarial example is a sample generated by inserting a small amount of noise into an original sample (with minimal distortion) such that it is recognized incorrectly by the targeted model. A typical method of attack using adversarial examples must satisfy two conditions: the distortion of the original sample must be kept to a minimum and misrecognition must be induced in the targeted deep neural network. Therefore, considerable time and numerous iterations are required to generate an adversarial example because both the conditions must be satisfied during the generation process. However, there are cases in which it may be desirable to generate an adversarial example that acts quickly to induce misrecognition in the deep neural network without considering the amount of distortion applied to the original sample. In this paper, we propose a dual-mode method for creating adversarial examples that allows the user to prioritize the malfunctioning of deep neural networks according to the situation. The proposed method generates an adversarial example using one of two modes: mode 1, which takes the level of distortion into account, and mode 0, which does not consider distortion and can generate examples rapidly. To evaluate the method experimentally, MNIST and CIFAR10 were used as the datasets. The results show that the proposed method can generate a targeted or untargeted adversarial example for MNIST with 50% fewer iterations using mode 0 than using mode 1. For CIFAR10, the reduction in the number of iterations that can be achieved using mode 0 is 80% and 88% for targeted and untargeted adversarial examples, respectively, and the attack success rate is 100%.

Enhancing the Performance of Multi-Category Text Classification via Label Relation Mining

Enhancing the Performance of Multi-Category Text Classification via Label Relation Mining

NEURAL NETWORKS OPTIMIZATION: METHODS AND THEIR COMPARISON BASED OFF TEXT INTELLECTUAL ANALYSIS

The research resulted in the development of software that implements various algorithms of neural networks optimization, which allowed to carry out their comparative analysis in terms of optimization quality. The article takes a detailed look at artificial neural networks and methods of their optimization: quantization, overcutting, distillation, Tucker’s dissolution. Algorithms and optimization tools of neural networks were explained, as well as comparative analysis of different methods was conducted with their advantages and disadvantages listed. Calculation values were given as well as recommendations on how to execute each method. Optimization is studied by text classification performance: peculiarities were removed, models were chosen and taught, parameters were adjusted. The set task was completed with the use of the following technologies: Python programming language, Pytorch framework and Jupyter Notebook developing environment. The results that were acquired can be used to reduce the demand on computing power while preserving the same level of detection and classification abilities.

Feature Extraction for Improvement Text Classification of Spam YouTube Video Comment using Deep Learning

The proposed algorithms are Bidirectional Long Short Term Memory (BiLSTM) and Conditional Random Fields (CRF) with Data Augmentation Technique (DAT). DAT integrates spam YouTube video comments into the traditional TF-IDF algorithm and generates a weighted word vector. The weighted word vector is fed into BiLSTM CRF to capture context information effectively. The result of this study is a new classification model to spam YouTube comment videos and increase the computational value of its performance. This research conducted two experiments: the first using BiLSTM CRF without DAT and the second using BiLSTM CRF with DAT. The experimental results state that the evaluation score using BiLSTM CRF with DAT shows outstanding performance in text classification, especially in spam YouTube video comment texts, with accuracy = 83.3%, precision = 83.6%, recall = 83.3%, and F-measure = 83.3%. So the combination of the BiLSTM-CRF method and the Data Augmentation Technique is very precise, so it can be used to increase the accuracy of classification texts for spam YouTube video comments

The application of printmaking art elements in visual communication design

Abstract Visual communication design and printmaking art are two independent art disciplines, but they have many similarities in graphic design, color matching and other aspects. This paper demonstrates the extraction of printmaking features using the improved LBP algorithm. The method of color histogram and color space is used to complete the color extraction of the image, and the texture features of the image are described by combining signal processing and statistical analysis. The LBP algorithm is used to locate the center pixel of the image to form a binary bit string, and the pattern value is obtained by multiplying each binary bit by the corresponding weight and summing up. The efficiency of image texture acquisition is enhanced by adding neighborhood gradient multiscale to the LBP algorithm. To obtain printmaking art elements and propose a direction for integration with visual communication design, image feature extraction is applied to printmaking art element extraction. The performance of texture classification for images has been significantly improved by the improved LBP algorithm, as shown in the results. On the Outextc10 texture dataset, DLBP5 achieves the highest classification accuracy of 92.19% by extracting colors, textures and shapes in printmaking and applying the features with elements of printmaking to visual communication design. Under the direction of integrating printmaking art style elements with visual communication design, users selected {unique, packaging design}, {vivid, branding design}, and {sharp, advertisement design} with higher satisfaction gradients, with ratings of 4.62, 4.17, and 4.36, respectively, which are all greater than 4.

Is text preprocessing still worth the time? A comparative survey on the influence of popular preprocessing methods on Transformers and traditional classifiers

With the advent of the modern pre-trained Transformers, the text preprocessing has started to be neglected and not specifically addressed in recent NLP literature. However, both from a linguistic and from a computer science point of view, we believe that even when using modern Transformers, text preprocessing can significantly impact on the performance of a classification model. We want to investigate and compare, through this study, how preprocessing impacts on the Text Classification (TC) performance of modern and traditional classification models. We report and discuss the preprocessing techniques found in the literature and their most recent variants or applications to address TC tasks in different domains. In order to assess how much the preprocessing affects classification performance, we apply the three top referenced preprocessing techniques (alone or in combination) to four publicly available datasets from different domains. Then, nine machine learning models – including modern Transformers – get the preprocessed text as input. The results presented show that an educated choice on the text preprocessing strategy to employ should be based on the task as well as on the model considered. Outcomes in this survey show that choosing the best preprocessing technique – in place of the worst – can significantly improve accuracy on the classification (up to 25%, as in the case of an XLNet on the IMDB dataset). In some cases, by means of a suitable preprocessing strategy, even a simple Naïve Bayes classifier proved to outperform (i.e., by 2% in accuracy) the best performing Transformer. We found that Transformers and traditional models exhibit a higher impact of the preprocessing on the TC performance. Our main findings are: (1) also on modern pre-trained language models, preprocessing can affect performance, depending on the datasets and on the preprocessing technique or combination of techniques used, (2) in some cases, using a proper preprocessing strategy, simple models can outperform Transformers on TC tasks, (3) similar classes of models exhibit similar level of sensitivity to text preprocessing.

Text Classification Based on the Heterogeneous Graph Considering the Relationships between Documents

Text classification is the task of estimating the genre of a document based on information such as word co-occurrence and frequency of occurrence. Text classification has been studied by various approaches. In this study, we focused on text classification using graph structure data. Conventional graph-based methods express relationships between words and relationships between words and documents as weights between nodes. Then, a graph neural network is used for learning. However, there is a problem that conventional methods are not able to represent the relationship between documents on the graph. In this paper, we propose a graph structure that considers the relationships between documents. In the proposed method, the cosine similarity of document vectors is set as weights between document nodes. This completes a graph that considers the relationship between documents. The graph is then input into a graph convolutional neural network for training. Therefore, the aim of this study is to improve the text classification performance of conventional methods by using this graph that considers the relationships between document nodes. In this study, we conducted evaluation experiments using five different corpora of English documents. The results showed that the proposed method outperformed the performance of the conventional method by up to 1.19%, indicating that the use of relationships between documents is effective. In addition, the proposed method was shown to be particularly effective in classifying long documents.

Additive Feature Attribution Explainable Methods to Craft Adversarial Attacks for Text Classification and Text Regression

Deep learning (DL) models have significantly improved the performance of text classification and text regression tasks. However, DL models are often strikingly vulnerable to adversarial attacks. Many researchers have aimed to develop adversarial attacks against DL models in realistic black-box settings (i.e., assuming no model knowledge is accessible to attackers). These attacks typically operate with a two-phase framework: (1) sensitivity estimation through gradient-based or deletion-based methods to evaluate the sensitivity of each token to the prediction of the target model, and (2) perturbation execution to craft adversarial examples based on the estimated token sensitivity. However, gradient-based and deletion-based methods used to estimate sensitivity often face issues of capturing token directionality and overlapping token sensitivities, respectively. In this study, we propose a novel eXplanation-based method for Adversarial Text Attacks (XATA) that leverages additive feature attribution explainable methods, namely LIME or SHAP, to measure the sensitivity of input tokens when crafting black-box adversarial attacks on DL models performing text classification or text regression. We evaluated XATA's attack performance on DL models executing text classification on the IMDB Movie Review, Yelp Reviews-Polarity, and Amazon Reviews-Polarity datasets and DL models conducting text regression on the My Personality, Drug Review, and CommonLit Readability datasets. The proposed XATA outperformed the existing gradient-based and deletion-based adversarial attack baselines in both tasks. These findings indicate that the ever-growing research focused on improving the explainability of DL models with additive feature attribution explainable methods can provide attackers with weapons to launch targeted adversarial attacks.