۱.
With the fast growth of social media, they have become the most important platform for posting multimodal content generated by users. Much of the data on social networks such as Instagram and Telegram is multimodal data. With the aim of analyzing such multimodal data in social networks, multimodal sentiment analysis has become one of the most significant subjects for researchers in the field of emotion recognition and data mining. Although multimodal sentiment analysis of social media data for English language has been addressed in several researches recently, few studies addressed the problem for the Persian language which is the official language of more than 120 million of people around the word. In this study, a multimodal deep learning model is proposed to address this problem. The proposed method utilizes a bi-directional long short-term memory (bi-LSTM) for processing text posts and a VGG16 convolutional network for analyzing images. A new dataset of Instagram and Telegram posts, MPerSocial, containing 1000 pairs of images and Persian comments is introduced in the current study and used for evaluating the proposed method. The results of experiments show that using the fusion of textual and image modalities improves sentiment polarity detection accuracy by 20% and 8% compared with the scenario in which image and text modalities in isolation. Also, the performance of the proposed model is better than three similar deep and four traditional machine learning models. All codes and dataset used in the current study are publicly available at GitHub.
۲.
The human chest contains vital organs such as the heart, lungs, and other organs. Chest radiology is one of the best and least costly methods to diagnose chest diseases. In this study, proposed a new method to diagnose 14 main diseases of the chest such as (cardiomegaly, emphysema, effusion, hernia, nodule, pneumothorax, atelectasis, pleural - thickening, mass, edema, integration, penetration, fibrosis, pneumonia) using the neural network and deep learning to increase accuracy, sensitivity, and specificity. The proposed method is implemented in the form of a web application and is available as a decision-making system for physicians to diagnose chest diseases.The results of the simulation on the sample dataset showed that the diagnosis of chest diseases was 98.93%, indicating the high efficiency of the new method. Finally, the proposed method was compared with other deep learning architectures such as densenet121, vgg16, exception architecture on the same dataset, which showed a 5% higher accuracy than them.
۳.
Q&A forums are designed to help users in finding useful information and accessing high-quality content posted by other users in text forums. Automatically identifying high-quality replies posted in response to the initial posts not only provides users with appropriate content, but also saves their time. Existing methods for classifying user replies based on their quality, try to extract quality features from both the textual content and metadata of the replies. This feature engineering step is a time and labor-intensive task. The current study addresses this problem by proposing new model based on deep learning for detecting quality user replies using only raw textual content. Specifically, we propose a long short-term memory (LSTM) model that exploits the embeddings from language models (ELMo) for representing words as contextual numerical vectors. We compared the effectiveness of the proposed model with four traditional machine learning models on the TripAdvisor for New York City (NYC) and the Ubuntu Linux distribution online forums datasets. Experimental results indicated that the proposed model significantly outperformed the four traditional algorithms on both datasets. Moreover, the proposed model achieved about 16% higher accuracy compared to that obtained by the traditional algorithms trained on both textual and quality dimension features.
۴.
Iran is facing low levels of all three types of children's nutrition like nutrient and micronutrients deficiency and overeating. The most common nutritional problems and child deaths are vitamin deficiencies and food quality. The purpose of this research is to plan food recommended system to control malnutrition in children 6 to 12 years old using hybrid machine learning algorithms. The results of this research are applicable in terms of target research. In terms of the implementation method, it is a descriptive survey and the process of gathering information is quantitative data. The dataset used includes 1001 data points collected from the health centers of Mianeh city located in East Azerbaijan in Iran from the integrated apple web system. In this research, the Python programming language has been used to analyze the child nutrition dataset, and AdaBoost and Decision Tree hybrid algorithms have been utilized for the child nutrients recommender system. We concluded that the number of meal features using the Decision Tree algorithm with 98.5% accuracy was more important than other nutritional features of children in recognizing malnutrition in them. From a review of 1001 data into the child nutrition dataset, 807 children are underweight and malnourished, 170 children are normal weight, 20 children are obese and four children are overweight. Therefore, the high exactness of hybrid algorithms in these studies has been able to have a high alignment with the opinion of nutritionists from 2019 to 2020.
۵.
Recent developments in Question Answering (QA) have improved state-of-the-art results, and various datasets have been released for this task. Since substantial English training datasets are available for this task, the majority of works published are for English Question Answering. However, due to the lack of Persian datasets, less research has been done on the latter language, making comparisons difficult. This paper introduces the Persian Question Answering Dataset (ParSQuAD) based on the machine translation of the SQuAD 2.0 dataset. Many errors have been discovered within the process of translating the dataset; therefore, two versions of ParSQuAD have been generated depending on whether these errors have been corrected manually or automatically. As a result, the first large-scale QA training resource for Persian has been generated. In addition, we trained three baseline models, i.e., BERT, ALBERT, and Multilingual-BERT (mBERT), on both versions of ParSQuAD. mBERT achieves scores of 56.66% and 52.86% for F1 score and exact match ratio respectively on the test set with the first version and scores of 70.84% and 67.73% respectively with the second version. This model obtained the best results out of the three on each version of ParSQuAD.
