Fusing Logical Relationship Information of Text in Neural Network for Text Classification

With the development of computer science and information science, text classification technology has been greatly developed and its application scenarios have been widened. In traditional process of text classification, the existing method will lose much logical relationship information of text. The logical relationship information of a text refers to the relationship information among different logical parts of the text, such as title, abstract, and body. When human beings are reading, they will take title as an important part to remind the central idea of the article, abstract as a brief summary of the content of the article, and body as a detailed description of the article. In most of the text classification studies, researchers concern more about the relationship among words (word frequency, semantics, etc.) and neglect the logical relationship information of text. It will lose information about the relationship among different parts (title, body, etc.) and have an influence on the performance of text classification. Therefore, we propose a text classification algorithm—fusing the logical relationship information of text in neural network (FLRIOTINN), which complements the logical relationship information into text classification algorithms. Experiments show that the effect of FLRIOTINN is better than the conventional backpropagation neural networks which does not consider the logical relationship information of text.

Download Full-text

A deep neural network improves endoscopic detection of early gastric cancer without blind spots

Endoscopy ◽

10.1055/a-0855-3532 ◽

2019 ◽

Vol 51 (06) ◽

pp. 522-531 ◽

Cited By ~ 33

Author(s):

Lianlian Wu ◽

Wei Zhou ◽

Xinyue Wan ◽

Jun Zhang ◽

Lei Shen ◽

...

Keyword(s):

Neural Network ◽

Gastric Cancer ◽

Early Gastric Cancer ◽

Predictive Value ◽

The Third ◽

Visual Tasks ◽

Deep Convolution Neural Network ◽

Different Parts ◽

Better Than ◽

Blind Spots

Abstract Background Gastric cancer is the third most lethal malignancy worldwide. A novel deep convolution neural network (DCNN) to perform visual tasks has been recently developed. The aim of this study was to build a system using the DCNN to detect early gastric cancer (EGC) without blind spots during esophagogastroduodenoscopy (EGD). Methods 3170 gastric cancer and 5981 benign images were collected to train the DCNN to detect EGC. A total of 24549 images from different parts of stomach were collected to train the DCNN to monitor blind spots. Class activation maps were developed to automatically cover suspicious cancerous regions. A grid model for the stomach was used to indicate the existence of blind spots in unprocessed EGD videos. Results The DCNN identified EGC from non-malignancy with an accuracy of 92.5 %, a sensitivity of 94.0 %, a specificity of 91.0 %, a positive predictive value of 91.3 %, and a negative predictive value of 93.8 %, outperforming all levels of endoscopists. In the task of classifying gastric locations into 10 or 26 parts, the DCNN achieved an accuracy of 90 % or 65.9 %, on a par with the performance of experts. In real-time unprocessed EGD videos, the DCNN achieved automated performance for detecting EGC and monitoring blind spots. Conclusions We developed a system based on a DCNN to accurately detect EGC and recognize gastric locations better than endoscopists, and proactively track suspicious cancerous lesions and monitor blind spots during EGD.

Download Full-text

Sentiment Classification Using Convolutional Neural Networks

Applied Sciences ◽

10.3390/app9112347 ◽

2019 ◽

Vol 9 (11) ◽

pp. 2347 ◽

Cited By ~ 18

Author(s):

Hannah Kim ◽

Young-Seob Jeong

Keyword(s):

Neural Network ◽

Neural Networks ◽

Convolutional Neural Networks ◽

Text Classification ◽

State Of The Art ◽

Sentiment Classification ◽

Learning Models ◽

Text Data ◽

Textual Data ◽

Better Than

As the number of textual data is exponentially increasing, it becomes more important to develop models to analyze the text data automatically. The texts may contain various labels such as gender, age, country, sentiment, and so forth. Using such labels may bring benefits to some industrial fields, so many studies of text classification have appeared. Recently, the Convolutional Neural Network (CNN) has been adopted for the task of text classification and has shown quite successful results. In this paper, we propose convolutional neural networks for the task of sentiment classification. Through experiments with three well-known datasets, we show that employing consecutive convolutional layers is effective for relatively longer texts, and our networks are better than other state-of-the-art deep learning models.

Download Full-text

CrowdTC: Crowd-powered Learning for Text Classification

ACM Transactions on Knowledge Discovery from Data ◽

10.1145/3457216 ◽

2021 ◽

Vol 16 (1) ◽

pp. 1-23

Author(s):

Keyu Yang ◽

Yunjun Gao ◽

Lei Liang ◽

Song Bian ◽

Lu Chen ◽

...

Keyword(s):

Neural Network ◽

Neural Networks ◽

Text Classification ◽

Deep Neural Networks ◽

Semantic Information ◽

Human Beings ◽

Hybrid Neural Network ◽

Public Datasets ◽

Almost All ◽

The Cost

Text classification is a fundamental task in content analysis. Nowadays, deep learning has demonstrated promising performance in text classification compared with shallow models. However, almost all the existing models do not take advantage of the wisdom of human beings to help text classification. Human beings are more intelligent and capable than machine learning models in terms of understanding and capturing the implicit semantic information from text. In this article, we try to take guidance from human beings to classify text. We propose Crowd-powered learning for Text Classification (CrowdTC for short). We design and post the questions on a crowdsourcing platform to extract keywords in text. Sampling and clustering techniques are utilized to reduce the cost of crowdsourcing. Also, we present an attention-based neural network and a hybrid neural network to incorporate the extracted keywords as human guidance into deep neural networks. Extensive experiments on public datasets confirm that CrowdTC improves the text classification accuracy of neural networks by using the crowd-powered keyword guidance.

Download Full-text

Effectiveness of Domain-Based Lexicons vis-à-vis General Lexicon for Aspect-Level Sentiment Analysis: A Comparative Analysis

Journal of Information & Knowledge Management ◽

10.1142/s0219649219500333 ◽

2019 ◽

Vol 18 (03) ◽

pp. 1950033

Author(s):

Madan Lal Yadav ◽

Basav Roychoudhury

Keyword(s):

Machine Learning ◽

Sentiment Analysis ◽

Text Classification ◽

Machine Learning Techniques ◽

Classification Algorithms ◽

Actual Performance ◽

Learning Techniques ◽

Improved Performance ◽

Text Content ◽

Better Than

One can either use machine learning techniques or lexicons to undertake sentiment analysis. Machine learning techniques include text classification algorithms like SVM, naive Bayes, decision tree or logistic regression, whereas lexicon-based sentiment analysis uses either general or domain-based lexicons. In this paper, we investigate the effectiveness of domain lexicons vis-à-vis general lexicon, wherein we have performed aspect-level sentiment analysis on data from three different domains, viz. car, guitar and book. While it is intuitive that domain lexicons will always perform better than general lexicons, the actual performance however may depend on the richness of the concerned domain lexicon as well as the text analysed. We used the general lexicon SentiWordNet and the corresponding domain lexicons in the aforesaid domains to compare their relative performances. The results indicate that domain lexicon used along with general lexicon performs better as compared to general lexicon or domain lexicon, when used alone. They also suggest that the performance of domain lexicons depends on the text content; and also on whether the language involves technical or non-technical words in the concerned domain. This paper makes a case for development of domain lexicons across various domains for improved performance, while gathering that they might not always perform better. It further highlights that the importance of general lexicons cannot be underestimated — the best results for aspect-level sentiment analysis are obtained, as per this paper, when both the domain and general lexicons are used side by side.

Download Full-text

Plato on Pure Pleasure and the Best Life

Phronesis ◽

10.1163/15685284-12341263 ◽

2014 ◽

Vol 59 (2) ◽

pp. 113-142 ◽

Cited By ~ 2

Author(s):

Emily Fletcher

Keyword(s):

Good Life ◽

Human Life ◽

Detailed Examination ◽

The Good Life ◽

Human Beings ◽

Instrumental Value ◽

Different Types ◽

The Relationship ◽

Better Than ◽

Positive Evaluation

Abstract In the Philebus, Socrates maintains two theses about the relationship between pleasure and the good life: (1) the mixed life of pleasure and intelligence is better than the unmixed life of intelligence, and: (2) the unmixed life of intelligence is the most divine. Taken together, these two claims lead to the paradoxical conclusion that the best human life is better than the life of a god. A popular strategy for avoiding this conclusion is to distinguish human from divine goods; on such a reading, pleasure has merely instrumental value, and it benefits human beings only as a result of their imperfect nature. I argue that certain ‘pure’ pleasures are full-fledged, intrinsic goods in the Philebus, which are even worthy of the gods (thus Socrates ultimately rejects thesis 2). This positive evaluation of pure pleasure results from a detailed examination of pleasure, which reveals that different types of pleasures have fundamentally different natures.

Download Full-text

Off-line Chinese Signature Verification Using Convolutional Neural Network

Recent Patents on Engineering ◽

10.2174/1872212112666180925152231 ◽

2019 ◽

Vol 13 (4) ◽

pp. 348-355

Author(s):

Shao Hong ◽

Xu Rongze ◽

Guo Xiaopeng ◽

Cui Wencheng

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Information Science ◽

Signature Verification ◽

Network Architectures ◽

Verification Method ◽

Signature Recognition ◽

Access Control System ◽

Unique Individual ◽

Better Than

Background: The unique individual biological characteristics is used for identification in biometrics, which is safe and difficult to forge. Therefore, it can help to enhance the safety of access control system. Since the developing of modern information science and technology, computerbased signature verification system enables signature verification more efficiently and automatically in comparison with traditional human identification method. Methods: In order to improve the accuracy of Chinese signature verification, an off-line Chinese signature verification method based on deep convolutional neural network is proposed. First, the machine learning library Tensorflow is build, and the volunteers are invited to establish the offline Chinese signature dataset. Second, the dataset is pre-processed, including denoising, binarization and size normalization. Finally, three different CNN architectures (AlexNet, GoogleNet, VGGNet) are adopted to implement the signature verification. Results: Experimental results show that the performance of AlexNet is better than that of the other two convolutional neural network architectures, the accuracy of classification has been up to 99.77%, and verification rate is 87.5%. Conclusion: Compared with the traditional offline Chinese signature recognition method, the method based on the convolutional neural network Alex Net-f is better than other methods to some extent, and avoids the complicated feature engineering.

Download Full-text

Multi-Resolution WNN Fault Diagnosis Model Based on Unscented Kalman Filtering for Rotating Machinery

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.687-691.1030 ◽

2014 ◽

Vol 687-691 ◽

pp. 1030-1033

Author(s):

Yan Ming Wei ◽

Xu Sheng Gan ◽

Jie Yang

Keyword(s):

Neural Network ◽

Fault Diagnosis ◽

Bp Neural Network ◽

Rotating Machinery ◽

Wavelet Neural Network ◽

Local Minima ◽

Learning Rules ◽

Fast Training ◽

The Relationship ◽

Better Than

BP neural network has a good nonlinear mapping ability, and can describe the relationship between frequency characteristics and fault. However, the multi-resolution wavelet neural network has the simple learning rules, fast training speed with the avoidance of local minima. So a multi-resolution wavelet neural network based on UKF is proposed to solve the problem of fault diagnosis for rotating machinery. The simulation result shows that the proposed multi-resolution wavelet neural network based on UKF value has a good diagnosis capability, and is better than that of traditional BP neural network and wavelet neural network.

Download Full-text

An investigation on the neighborhood of Iran’s villages and cities (case study: Kondlous Chalous Village)

Innovaciencia Facultad de Ciencias Exactas Físicas y Naturales ◽

10.15649/2346075x.768 ◽

2019 ◽

Vol 7 (2) ◽

Author(s):

Somayeh Noruzi

Keyword(s):

Direct Observation ◽

Social Life ◽

Research Method ◽

Human Beings ◽

Sense Of Satisfaction ◽

The Relationship ◽

Different Parts

The design of residential complexes, taking into account neighborhoods in different parts of the building, can improve thequality of life of the users, human beings are created socially and physically and mentally requires communication and interaction oftheir own kind. In fact, communication is one of the most important and tangible elements of social life. In many residential complexes,due to the lack of desirable neighborhoods, the relationship between residents does not occur. In this case, many residents donot find this sense of satisfaction and belonging to the environment in which they live. In this research, the research method used isdescriptive-analytic. Using library documents, surveying methods and direct observation of space, as well as field observationsand questions from space users, to collect and analyze the data needed for research. Is. The results of the analysis of texture andneighborhoods of the Kandlos village as a case study show that the traditional neighborhood and Iranian neighborhoods are moreflexible, more diverse, more complete and comprehensive, and using the principles of location Traditional Iranian can reproducethe concept of interaction and neighborhoods and used in the design of contemporary Iranian residential complexes.

Download Full-text

Analisis Sentimen pada Data Saran Mahasiswa Terhadap Kinerja Departemen di Perguruan Tinggi Menggunakan Convolutional Neural Network

Jurnal Teknologi Informasi dan Ilmu Komputer ◽

10.25126/jtiik.2021854842 ◽

2021 ◽

Vol 8 (5) ◽

pp. 1067

Author(s):

Yuliska Yuliska ◽

Dini Hidayatul Qudsi ◽

Juanda Hakim Lubis ◽

Khairul Umum Syaliman ◽

Nina Fadilah Najwa

Keyword(s):

Neural Network ◽

Convolutional Neural Network ◽

Sentiment Analysis ◽

Text Classification ◽

Word Embedding ◽

Great Performance ◽

Word Representation ◽

The Relationship

Review atau saran dari customer dapat menjadi sangat penting bagi penyedia layanan, begitu pula saran dari mahasiswa mengenai layanan sebuah unit kerja di perguruan tinggi. Review menjadi penting karena dapat menjadi indikator kinerja penyedia layanan. Pengolahan review juga sangat penting karena dapat menjadi referensi untuk pengambilan keputusan dan peningkatan layanan yang lebih baik ke depannya. Penelitian ini menerapkan analisis sentimen pada data saran atau review mahasiswa terhadap kinerja unit kerja atau departemen di perguruan tinggi, yaitu Politeknik Caltex Riau. Analisis sentimen dilakukan dengan menggunakan Convolutional Neural Network (CNN) dan word embedding Word2vec sebagai representasi kata. CNN merupakan metode yang memiliki performa yang baik dalam mengklasifikasi teks, yaitu dengan teknik convolutional yang menggabungkan beberapa window kata pada kalimat dan mengambil window yang paling representative. Word2Vec digunakan sebagai representasi data saran dan inputan awal pada CNN, dimana Word2Vec merupakan dense vectors yang dapat merepresentasikan hubungan antar kata pada data saran dengan baik. Saran mahasiswa dapat mengandung kalimat yang sangat panjang, karena itu perpaduan Word2Vec sebagai representasi kata dan CNN dengan teknik convolutional, dapat menghasilkan representasi yang representative dari kalimat panjang tersebut. Penelitian ini menggunakan dua arsitektur CNN, yaitu Simple CNN dan DoubleMax CNN untuk mengidentifikasi pengaruh kompleksitas arsitektur terhadap hasil klasifikasi sentimen. Berdasarkan hasil pengujian, DoubleMax CNN dapat mengklasifikasi sentimen pada saran mahasiswa dengan sangat baik, yaitu mencapai Akurasi tertinggi sebesar 98%, Recall 97%, Precision 98% dan F1-Score 98%. AbstractStudent’s reviews about department performance can be essential for a college for it can be used to evaluate the department performance and to take an immediate action to improve its performance. This research applies sentiment analysis in the student’s reviews of college department in Politeknik Caltex Riau. Convolutional Neural Network and Word2Vec are employed to analyze the sentiment. CNN is known for its good performance in text classification by applying a convolutional technique to the input sentences. Word2Vec is used as word representation and as an input to the CNN. Word2Vec are dense vectors which can represent the relationship between words excellently. Student’s reviews can be a long sentence; hence the combination of Word2Vec as word representation and CNN with convolutional technique can produce a representative fiture from that long sentence. This research utilizes two CNN architectures, which are Simple CNN dan DoubleMax CNN to identify the effect of the complexity of CNN architecture to final result. Our experiments show that DoubleMax CNN has a great performance in classifying sentiment in the student’s reviews with the best Accuracy value of 98%, Recall 97%, Precision 98% and F1-Score value of 98%.

Download Full-text

Albanian Text Classification: Bag of Words Model and Word Analogies

Business Systems Research Journal ◽

10.2478/bsrj-2019-0006 ◽

2019 ◽

Vol 10 (1) ◽

pp. 74-87

Author(s):

Arbana Kadriu ◽

Lejla Abazi ◽

Hyrije Abazi

Keyword(s):

Text Classification ◽

Word Order ◽

Classification Algorithms ◽

Testing Time ◽

Bag Of Words ◽

Large Dataset ◽

Text Documents ◽

Independent Components ◽

Hierarchical Classifier ◽

Better Than

Abstract Background: Text classification is a very important task in information retrieval. Its objective is to classify new text documents in a set of predefined classes, using different supervised algorithms. Objectives: We focus on the text classification for Albanian news articles using two approaches. Methods/Approach: In the first approach, the words in a collection are considered as independent components, allocating to each of them a conforming vector in the vector’s space. Here we utilized nine classifiers from the scikit-learn package, training the classifiers with part of news articles (80%) and testing the accuracy with the remaining part of these articles. In the second approach, the text classification treats words based on their semantic and syntactic word similarities, supposing a word is formed by n-grams of characters. In this case, we have used the fastText, a hierarchical classifier, that considers local word order, as well as sub-word information. We have measured the accuracy for each classifier separately. We have also analyzed the training and testing time. Results: Our results show that the bag of words model does better than fastText when testing the classification process for not a large dataset of text. FastText shows better performance when classifying multi-label text. Conclusions: News articles can serve to create a benchmark for testing classification algorithms of Albanian texts. The best results are achieved with a bag of words model, with an accuracy of 94%.

Download Full-text