• Journal of the Korean Society for information Management
• /
• v.22 no.3 s.57
• /
• pp.261-287
• /
• 2005

The purpose of this paper is to explore the ways to improve the performance of SVM (Support Vector Machines) text classifier using inter-document similarities. SVMs are powerful machine learning systems, which are considered as the state-of-the-art technique for automatic document classification. In this paper text categorization via SVMs approach based on feature representation with document vectors is suggested. In this approach, document vectors instead of index terms are used as features, and vector similarities instead of term weights are used as feature values. Experiments show that SVM classifier with document vector features can improve the document classification performance. For the sake of run-time efficiency, two methods are developed: One is to select document vector features, and the other is to use category centroid vector features instead. Experiments on these two methods show that we can get improved performance with small vector feature set than the performance of conventional methods with index term features.

• Smart Media Journal
• /
• v.13 no.4
• /
• pp.9-15
• /
• 2024

Tree-based algorithms have been the dominant methods used build a prediction model for tabular data. This also includes personal credit data. However, they are limited to compatibility with categorical and numerical data only, and also do not capture information of the relationship between other features. In this work, we proposed an ensemble model using the Transformer architecture that includes text features and harness the self-attention mechanism to tackle the feature relationships limitation. We describe a text formatter module, that converts the original tabular data into sentence data that is fed into FinBERT along with other text features. Furthermore, we employed FT-Transformer that train with the original tabular data. We evaluate this multi-modal approach with two popular tree-based algorithms known as, Random Forest and Extreme Gradient Boosting, XGBoost and TabTransformer. Our proposed method shows superior Default Recall, F1 score and AUC results across two public data sets. Our results are significant for financial institutions to reduce the risk of financial loss regarding defaulters.

• Journal of Information Science Theory and Practice
• /
• v.10 no.spc
• /
• pp.86-95
• /
• 2022

Online recruitment websites discuss job demands in various fields, and job postings contain detailed job specifications. Analyzing this text can elucidate the features that determine job salaries. Text embedding models can learn the contextual information in a text, and explainable artificial intelligence frameworks can be used to examine in detail how text features contribute to the models' outputs. We collected 733,625 job postings using the WORKNET API and classified them into low, mid, and high-range salary groups. A text embedding model that predicts job salaries based on the text in job postings was trained with the collected data. Then, we applied the SHapley Additive exPlanations (SHAP) framework to the trained model and discovered the significant words that determine each salary class. Several limitations and remaining words are also discussed.

• International Journal of Computer Science & Network Security
• /
• v.22 no.1
• /
• pp.283-287
• /
• 2022

Machine Learning is the most popular method used in data science. Growth of data is not only numeric data but also text data. Most of the algorithm of supervised and unsupervised machine learning algorithms use numeric data. Now it is required to convert text data into numeric. There are many techniques for this conversion. Researcher confuses which technique is best in what situation. Here in proposed work BOW (Bag-of-Words) and TF-IDF (Term-Frequency-Inverse-Document-Frequency) has been studied based on different features to determine best method. After experimental results on text data, TF-IDF and BOW both provide better performance at range from 100 to 150 number of features.

• Journal of Information Processing Systems
• /
• v.13 no.5
• /
• pp.1397-1409
• /
• 2017

For text categorization task, distinctive text features selection is important due to feature space high dimensionality. It is important to decrease the feature space dimension to decrease processing time and increase accuracy. In the current study, for text categorization task, we introduce a novel statistical feature selection approach. This approach measures the term distribution in all collection documents, the term distribution in a certain category and the term distribution in a certain class relative to other classes. The proposed method results show its superiority over the traditional feature selection methods.

• Journal of information and communication convergence engineering
• /
• v.15 no.1
• /
• pp.53-61
• /
• 2017

The preferred feature selection methods for text classification are filter-based. In a common filter-based feature selection scheme, unique scores are assigned to features; then, these features are sorted according to their scores. The last step is to add the top-N features to the feature set. In this paper, we propose an improved global feature selection scheme wherein its last step is modified to obtain a more representative feature set. The proposed method aims to improve the classification performance of global feature selection methods by creating a feature set representing all classes almost equally. For this purpose, a local feature selection method is used in the proposed method to label features according to their discriminative power on classes; these labels are used while producing the feature sets. Experimental results obtained using the well-known 20 Newsgroups and Reuters-21578 datasets with the k-nearest neighbor algorithm and a support vector machine indicate that the proposed method improves the classification performance in terms of a widely known metric ($F_1$).

• Journal of the Korean Institute of Intelligent Systems
• /
• v.25 no.4
• /
• pp.386-391
• /
• 2015

This paper proposed the optimal attributes for text classification based on Korean and Chinese linguistic features. The experiments committed to discover which is the best feature among n-grams which is known as language independent, morphemes that have language dependency and some other feature sets consisted with n-grams and morphemes showed best results. This paper used SVM classifier and Internet news for text classification. As a result, bi-gram was the best feature in Korean text categorization with the highest F1-Measure of 87.07%, and for Chinese document classification, 'uni-gram+noun+verb+adjective+idiom', which is the combined feature set, showed the best performance with the highest F1-Measure of 82.79%.

• ETRI Journal
• /
• v.32 no.5
• /
• pp.729-739
• /
• 2010

Due to the rapid development of mobile devices equipped with cameras, instant translation of any text seen in any context is possible. Mobile devices can serve as a translation tool by recognizing the texts presented in the captured scenes. Images captured by cameras will embed more external or unwanted effects which need not to be considered in traditional optical character recognition (OCR). In this paper, we segment a text image captured by mobile devices into individual single characters to facilitate OCR kernel processing. Before proceeding with character segmentation, text detection and text line construction need to be performed in advance. A novel character segmentation method which integrates touched character filters is employed on text images captured by cameras. In addition, periphery features are extracted from the segmented images of touched characters and fed as inputs to support vector machines to calculate the confident values. In our experiment, the accuracy rate of the proposed character segmentation system is 94.90%, which demonstrates the effectiveness of the proposed method.

• Journal of KIISE:Software and Applications
• /
• v.35 no.10
• /
• pp.608-615
• /
• 2008

In Knowledge Q&A services where information is created by unspecified users, document quality is an important factor of user satisfaction with search results. Previous work on quality prediction of Knowledge Q&A documents evaluate the quality of documents by using non-textual information, such as click counts and recommendation counts, and focus on enhancing retrieval performance by incorporating the quality measure into retrieval model. Although the non-textual information used in previous work was proven to be useful by experiments, data sparseness problem may occur when predicting the quality of newly created documents with such information. To solve data sparseness problem of non-textual features, this paper proposes new features for document quality prediction, namely text-confidence features, which indicate how trustworthy the content of a document is. The proposed features, extracted directly from the document content, are stable against data sparseness problem, compared to non-textual features that indirectly require participation of service users in order to be collected. Experiments conducted on real world Knowledge Q&A documents suggests that text-confidence features show performance comparable to the non-textual features. We believe the proposed features can be utilized as effective features for document quality prediction and improve the performance of Knowledge Q&A services in the future.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.5
• /
• pp.279-284
• /
• 2017

Currenty, in the era of big data, text mining and opinion mining have been used in many domains, and one of their most important research issues is to extract significant information from social media. Thus in this paper, we propose a logistic regression ensemble method of finding the main body text from blog HTML. First, we extract structural features and text features from blog HTML tags. Then we construct a classification model with logistic regression and ensemble that can decide whether any given tags involve main body text or not. One of our important findings is that the main body text can be found through 'depth' features extracted from HTML tags. In our experiment using diverse topics of blog data collected from the web, our tag classification model achieved 99% in terms of accuracy, and it recalled 80.5% of documents that have tags involving the main body text.