This paper proposes AEDA (An Easier Data Augmentation) technique to help improve the performance on text classification tasks. AEDA includes only random insertion of punctuation marks into the original text. This is an easier technique to implement f
or data augmentation than EDA method (Wei and Zou, 2019) with which we compare our results. In addition, it keeps the order of the words while changing their positions in the sentence leading to a better generalized performance. Furthermore, the deletion operation in EDA can cause loss of information which, in turn, misleads the network, whereas AEDA preserves all the input information. Following the baseline, we perform experiments on five different datasets for text classification. We show that using the AEDA-augmented data for training, the models show superior performance compared to using the EDA-augmented data in all five datasets. The source code will be made available for further study and reproduction of the results.
This paper describes our system participated in Task 6 of SemEval-2021: the task focuses on multimodal propaganda technique classification and it aims to classify given image and text into 22 classes. In this paper, we propose to use transformer base
d architecture to fuse the clues from both image and text. We explore two branches of techniques including fine-tuning the text pretrained transformer with extended visual features, and fine-tuning the multimodal pretrained transformers. For the visual features, we have tested both grid features based on ResNet and salient region features from pretrained object detector. Among the pretrained multimodal transformers, we choose ERNIE-ViL, a two-steam cross-attended transformers pretrained on large scale image-caption aligned data. Fine-tuing ERNIE-ViL for our task produce a better performance due to general joint multimodal representation for text and image learned by ERNIE-ViL. Besides, as the distribution of the classification labels is very unbalanced, we also make a further attempt on the loss function and the experiment result shows that focal loss would perform better than cross entropy loss. Last we have won first for subtask C in the final competition.
Transformer architecture achieves great success in abundant natural language processing tasks. The over-parameterization of the Transformer model has motivated plenty of works to alleviate its overfitting for superior performances. With some explorat
ions, we find simple techniques such as dropout, can greatly boost model performance with a careful design. Therefore, in this paper, we integrate different dropout techniques into the training of Transformer models. Specifically, we propose an approach named UniDrop to unites three different dropout techniques from fine-grain to coarse-grain, i.e., feature dropout, structure dropout, and data dropout. Theoretically, we demonstrate that these three dropouts play different roles from regularization perspectives. Empirically, we conduct experiments on both neural machine translation and text classification benchmark datasets. Extensive results indicate that Transformer with UniDrop can achieve around 1.5 BLEU improvement on IWSLT14 translation tasks, and better accuracy for the classification even using strong pre-trained RoBERTa as backbone.
In this paper, we have used electrospinning technique to obtain nonwoven networks of nanofibers from polyacrylic nitrile, where we dissolved polyacryl nitrile in its dimethylformamide solvent to form the polymeric solution.
Modern and future optical networks rely on wavelength divisional multiplexing
technology, it was necessary to developed the whole network elements to keep up with the
increasing need to offer a wide band and a very short time delay and high reliabi
lity, and
replacement of electro optic equipment with optical equipment.
Optical amplifiers have taken an important part in this evolution, and the Raman amplifier
(RAMAN) had a great deal of attention, for its high gain and flattened gain.
In this paper, we have examined the effect of both fiber parameters and pump parameters
on the performance of Raman optical amplifier in terms of gain and bandwidth and the
used pump power.
This research demonstrated the effect of multi pump on this amplifier performance and its
flattening and bandwidth, and we reached a flattening gain on a wide bandwidth.
The simulation was done by using MATLAB and OPTISYSTEM program based on the
mathematical equations that describe the amplifier model for 1450nm-1650nm bandwidth.
This research shows a reference study about the technique of the nanotechnology,
kinds of nano materials, methods of its manufacturing, its characteristics, its applications,
and the role of nano in electronics and optoelectronic. As well this rese
arch explains the
possibility of manufacturing of a computer with carbon nanotubes and the technique of the
paper battery and the capacitor with high capacity.
This study gives too an overvieo of the devices and the used technique to in vestigate
the nano materials.The article shows too the behavior of the quantum mechanics with the
nano electronics. Finally it gives a general overvieo about the quantum devices and their
own simulation processes.
The present study is concerned with the creation of an organized
framework to the teaching of architectural design that addresses
the individual differences of learners, such as differences in
abilities, tendencies, experiences, speed of learning
and self
development through appointing the self-learning skills required
for architectural students, and how much these skills are
practiced, as an important aspect of the design process. This
method enhances the education and development of mental
ability which considered the most important for students to solve
design problems in compliance with the society variables.
Aim of the Study: To evaluate the changes to the radiographic
bone density of the mental bone grafts applied in terms of ring
technique until 18 months post operatively and to evaluate the
effect of the adjacency of neighboring teeth.
Current studies in most developing countries focus on the development of cooling technologies due to the unreliable electrical supply and the urgent need for reliable cooling in areas connected to or independent of the network. In this research, the
performance of a 112L DC refrigerator was evaluated, whereby a single screw AC compressor was replaced by a variable speed VSDC compressor that can be operated by a solar PV system without an inverter.
This study examines the most important component of NG-PON2
networks and the characteristics of these network through a model of a
system consisting of four pairs of wavelengths to serve 8 subscribers
with symmetrical transmission on both tow links equal to 40Gbps, The
proposed model designed using OptiSystem program shows good
results that give a clear view of the behavior of such systems.
