Sharif Digital Repository

Large data sources labeled by experts at cost are essential for deep learning success in various domains. But, when labeling is expensive and labeled data is scarce, deep learning generally does not perform well. The goal of semi-supervised learning is to leverage abundant unlabeled data that one can easily collect. New semi-supervised algorithms based on data augmentation techniques have reached new advances in this field. In this work, by studying different textual augmentation techniques, a new approach is proposed that can obtain effective information signals from unlabeled data. The method encourages the model to generate the same representation vectors for different augmented versions... 

One way for businesses to grow and compete, in any age (especially the digital age), is to create a Brand Relevance through creating or finding, and then owning new categories or subcategories. In this way, instead of beating competitors; they become irrelevant by enticing customers to buy a new category or subcategory for which other alternative brands are not considered relevant. Firms traditionally rely on interviews and focus groups to identify these subcategories and customer needs. Nowadays, with the growth of social media, user-generated content (UGC) is also a good alternative source. However, Due to the large size of UGC and the non-informative or repetitive data it contains,... 

Text classification accompanied by high precision has become a challenging issue in computational linguistics and natural language processing science. Proper data set accessibility, utilizing the best method and prominent linguistics features has been always regarded as the basic concern of this process. The following study relying on Bijan Khan Corpus is tried to represent keywords vectors of different documents using tf_idf. These vectors are regarded as an input for latent topic models algorithms including probabilistic latent semantic analysis. The output of this algorithm will be the documents feature vectors which will be later used in order to train different classifiers like K... 

In recent years, neural networks have been widely used in most machine learning domains. However, it has been shown that these networks are vulnerable to adversarial examples. adversarial examples are small and imperceptible perturbations applied to the input which lead to producing wrong output and thus, fooling the network. This will become an important issue in security related applications of deep neural networks, such as self-driving cars and medical diagnostics. Since, in the wort-case scenario, even human lives could be threatened. Although, many works have focused on crafting adversarial examples for image data, only a few studies have been done on textual data due to the existing... 

With the widespread usage of the internet among all layers of societies and the fast growth of social networks’ impact, researchers found a new source to study people’s habits, interests and culture. In order to combine these two important aspects of social networks, we used data from social networks to perform a cross-cultural study. The proposed method includes steps of gathering data from twitter, automatic classification of tweets into news categories and calculating cultural distance and cultural similarities from the overall distribution of tweets among the selected classes. By applying the proposed method on a sample of tweets in 2016, we examined the overall tendencies of users of...