Fine-Grained Sentiment Analysis of Arabic COVID-19 Tweets Using BERT-Based Transformers and Dynamically Weighted Loss Function

The outbreak of coronavirus disease (COVID-19) has affected almost all of the countries of the world, and has had significant social and psychological effects on the population. Nowadays, social media platforms are being used for emotional self-expression towards current events, including the COVID-19 pandemic. The study of people’s emotions in social media is vital to understand the effect of this pandemic on mental health, in order to protect societies. This work aims to investigate to what extent deep learning models can assist in understanding society’s attitude in social media toward COVID-19 pandemic. We employ two transformer-based models for fine-grained sentiment detection of Arabic tweets, considering that more than one emotion can co-exist in the same tweet. We also show how the textual representation of emojis can boost the performance of sentiment analysis. In addition, we propose a dynamically weighted loss function (DWLF) to handle the issue of imbalanced datasets. The proposed approach has been evaluated on two datasets and the attained results demonstrate that the proposed BERT-based models with emojis replacement and DWLF technique can improve the sentiment detection of multi-dialect Arabic tweets with an F1-Micro score of 0.72.

Improving Accuracy using The ASERLU layer in CNN-BiLSTM Architecture on Sentiment Analysis

There have been 350,000 tweets generated by the interaction of social networks with different cultures and educational backgrounds in the last ten years. Various sentiments are expressed in the user comments, from support to hatred. The sentiments regarded the United States General Election in 2020. This dataset has 3,000 data gotten from previous research. We augment it becomes 15,000 data to facilitate training and increase the required data. Sentiment detection is carried out using the CNN-BiLSTM architecture. It is chosen because CNN can filter essential words, and BiLSTM can remember memory in two directions. By utilizing both, the training process becomes maximum. However, this method has disadvantages in the activation. The drawback of the existing activation method, i.e., "Zero-hard Rectifier" and "ReLU Dropout" problem to become the cause of training stopped in the ReLU activation, and the exponential function cannot be set become the activation function still rigid towards output value in the SERLU activation. To overcome this problem, we propose a novel activation method to repair activation in CNN-BiLSTM architecture. It is namely the ASERLU activation function. It can adjust positive value output, negative value output, and exponential value by the setter variables. So, it adapts more conveniently to the output value and becomes a flexible activation function because it can be increased and decreased as needed. It is the first research applied in architecture. Compared with ReLU and SERLU, our proposed method gives higher accuracy based on the experiment results.

Automatic Sentiment Detection in Naturalistic Audio

This paper proposed another Audio notion investigation utilizing programmed discourse acknowledgment is an arising research territory where assessment or opinion showed by a speaker is identified from regular sound. It is moderately under-investigated when contrasted with text-based notion identification. Separating speaker estimation from common sound sources is a difficult issue. Nonexclusive techniques for feeling extraction by and large use records from a discourse acknowledgment framework, and interaction the record utilizing text-based estimation classifiers. In this examination, we show that this standard framework is imperfect for sound assessment extraction. Then again, new engineering utilizing watchword spotting (UWS) is proposed for assumption discovery. In the new engineering, a book-based assessment classifier is used to naturally decide the most helpful and discriminative feeling bearing watchword terms, which are then utilized as a term list for UWS. To get a minimal yet discriminative assumption term list, iterative element enhancement for most maximum entropy estimation model is proposed to diminish model intricacy while keeping up powerful grouping precision. The proposed arrangement is assessed on sound acquired from recordings in youtube.com and UT-Opinion corpus. Our exploratory outcomes show that the proposed UWS based framework fundamentally outflanks the conventional engineering in distinguishing assumption for testing reasonable undertakings.

Sentiment Analysis Technologies in AliMe — An Intelligent Assistant for E-commerce

In this paper, we introduce a sentiment analysis framework and its corresponding key techniques used in AliMe, an artificial intelligent (AI) assistant for e-commerce customer service, whose fundamental ability of sentiment analysis provides support for five upper-layer application modules: user sentiment detection, user sentiment comfort, sentimental generative chatting, user service quality control and user satisfaction prediction. Detailed implementation of each module is demonstrated and experiments show our framework not only performs well on each single task but also manifests its competitive business value as a whole.