Improving deep neural network design with new text data representations

Prusa, Joseph D.; Khoshgoftaar, Taghi M.

doi:10.1186/s40537-017-0065-8

Cited by 57 publications

(32 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, the parameters settings used were strongly influenced by the characteristics of the data set used, such as the number of data set rows. In addition to parameters setting, accuracy was also influenced by the CNN architecture used, as well in previous studies which used a more complex CNN architecture on the twitter data set and got the better results [16], [17].…”

Section: Resultsmentioning

confidence: 86%

“…Although Parameters Setting IV and Parameters Setting III have the same parameters setting values, they have not able yet to make the proposed study (indicated by bold numbers) getting the greater accuracy (79.3%, 81.10%, 80.67%). The higher accuracies are (82.7%, 86.8%) [4], (82.3%) [17], and (88.3%) [16], indicated by numbers in italics), as shown in Table IX. It shows that although the Parameters Setting III and IV values were able to produce a better accuracy on the movie review data set, the parameters setting is not capable in producing better accuracy when applied to the Twitter data set.…”

Section: Resultsmentioning

confidence: 92%

“…This Parameters Setting IV used the same parameters setting values as Parameters Setting III since Parameters Setting III has been shown to provide better results than Parameters Setting I and II. The results of the accuracy are shown in Table VIII and Table IX. [4] 82.3% [17] 88 .3% [16] TF + w2v 81.10% 86.8% [4] 82.3% [17] 88 .3% [16] TH+ w2v 80.67% 86.8% [4] 82.3% [17] 88 .3% [16] TF: Tensorflow, TH: Theano, w2v: Pre-Training Word2vec…”

Section: Resultsmentioning

confidence: 99%

See 2 more Smart Citations

Performance Improvement Using CNN for Sentiment Analysis

Nasichuddin

Adji

Widyawan

2018

IJITEE

View full text Add to dashboard Cite

The approach using Deep Learning method provides great results in various field implementations, especially in the field of Sentiment Analysis. One of Deep Learning methods is CNN which has the ability to provide great accuracy in some previous research. However, there are some parts of the training process which can be improved to upgrade the accuracy level and the training time. In this paper, we try to improve the accuracy and processing time of sentiment analysis using CNN model. By tuning the filter size, frameworks, and pre-training, the results show that the use of smaller filter size and pre-training word2vec provide greater accuracy than some previous studies.

show abstract

Section: Resultsmentioning

confidence: 86%

Section: Resultsmentioning

confidence: 92%

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Performance Improvement Using CNN for Sentiment Analysis

Nasichuddin

Adji

Widyawan

2018

IJITEE

View full text Add to dashboard Cite

show abstract

“…Dropout randomly ignores neural connections during training and this significantly reduces overfitting, often obtaining major improvements when compared with other regularization methods. 77 The grid search ranges for the MLP hyperparmeters were set to: 78 and Mahendhiran et al 73 that used a fixed number of epochs (e.g., 100) for each experiment, this hyperparameter value was set to 100.…”

Section: Multilayer Perceptronmentioning

confidence: 99%

Social Media Cross-Source and Cross-Domain Sentiment Classification

Zola

Cortez

Ragno

et al. 2019

Int. J. Info. Tech. Dec. Mak.

View full text Add to dashboard Cite

Due to the expansion of Internet and Web 2.0 phenomenon, there is a growing interest in sentiment analysis of freely opinionated text. In this paper, we propose a novel cross-source cross-domain sentiment classification, in which cross-domain-labeled Web sources (Amazon and Tripadvisor) are used to train supervised learning models (including two deep learning algorithms) that are tested on typically nonlabeled social media reviews (Facebook and Twitter). We explored a three-step methodology, in which distinct balanced training, text preprocessing and machine learning methods were tested, using two languages: English and Italian. The best results were achieved using undersampling training and a Convolutional Neural Network. Interesting cross-source classification performances were achieved, in particular when using Amazon and Tripadvisor reviews to train a model that is tested on Facebook data for both English and Italian.

show abstract

“…They use a combination of text mining and machine-learning techniques to identify ideas hidden in large amounts of texts. Prusa and Khoshgoftaar (2017) propose a new method of creating character-level representations of text to reduce the computational costs associated with training a deep convolutional network. Further, they show that we can use the proposed embedding with padded convolutional layers to enable the use of current convolutional network architectures, while still facilitating faster training and higher performance than the previous approach for learning from character level.…”

Section: Literature Reviewmentioning

confidence: 99%

An analysis of the 2016 US presidential election using Chanakya - a knowledge discovery platform for text mining

Malhotra

2018

IJKEDM

View full text Add to dashboard Cite

In this era of information overload, discovering knowledge is a challenge. However, a new generation of text mining tools enables researchers and practitioners to analyse large volumes of data. This paper illustrates the design of knowledge discovery system -Chanakya using text mining. Chanakya works in two stages. Stage 1 uses naive Bayes classifier, a supervised machine-learning algorithm to train for classes, as we explicitly provide training data that is labelled with classes. Stage 2 uses k-means analysis, an unsupervised machine-learning algorithm to determine what categories are emerging from the mentions of each class. We use the 2016 presidential elections Twitter feeds to illustrate the use of Chanakya. Chanakya offers a commentary on the current state of the political arena after analysing the candidate tweets and how people are reacting to these tweets.

show abstract

Improving deep neural network design with new text data representations

Cited by 57 publications

References 16 publications

Performance Improvement Using CNN for Sentiment Analysis

Performance Improvement Using CNN for Sentiment Analysis

Social Media Cross-Source and Cross-Domain Sentiment Classification

An analysis of the 2016 US presidential election using Chanakya - a knowledge discovery platform for text mining

Contact Info

Product

Resources

About