International Journal of Web Research
Volume:1 Issue: 2, Autumn-Winter 2018

Twitter has provided a convenient platform to express feelings and opinions in different areas. Opinion mining in Twitter can be considered as studying the overall sentiment of a tweet. There are two general categories of sentiment analysis methods in the Persian language, linked-base methods and, content-based methods. In this study, we implement a new link-based method for improving opinion classification in the Persian language. To compare with the content-based method, we implement a content-based method using Naïve Bayes Method with two different weighting Methods TF/IDF and Chi-Square. The TF/IDF method has good results in previous Persian language studies. The Chi-Square method has not been used in the Persian language researches, but the accuracy is fairly good in English. The results show that the improvement in the language-independent methods is remarkable and is in accordance with this research, the precision of the proposed algorithm for positive and negative comments was 98.87% and 97.87%, and the recall value for positive and negative comments was 99.24% and 96.84% respectively. The results also show that because of complexities in Persian syntax and lack of proper natural language processing tools in Persian, content-based algorithms operate poorly compared to English.

In recent years, discriminative learning methods have widely been used in various areas of Natural Language Processing (NLP). These methods achieve the best performance, when the set of training and testing samples have the same distribution. However, in many applications of NLP, the lack of labeled datasets for some domains is a serious challenge. In such conditions, we need to develop a model based on domains with rich labeled instances and apply it to the domain with no labeled instances. In this research, a method for sentiment classification of opinions into positive and negative groups, which represent the users' feelings, is offered based on multi-source transfer learning. The proposed method here employs Spectral Feature Alignment algorithm to adapt different domains. Furthermore, according to the Majority Voting, accuracy is assigned to classifications trained on different domains based on the Majority Voting Error. Ultimately, decisions are made for each classification based on the calculated error. The Amazon datasets for four different categories, each of which contains 1000 positive and 1000 negative samples, are exploited to train the proposed model. Meanwhile, each category includes unlabeled samples that are used to select pivot features. The accuracy values of 85.5%, 86.4%, 83.5% and 90.1% obtained for Electronics, DVD, Books and Kitchen domains respectively, show the effectiveness of the proposed method compared with similar methods.

In this paper, we focus on improving the performance of recommender systems. To do this, we propose a new algorithm named PBloofI which is a kind of hierarchical bloom filter. Actually, the Bloom filter is an array-based technique for showing the items’ features. Since the feature vectors of items are sparse, the Bloom filter reduces the space usage by using the hashing technique. And also, to reduce the time complexity we used the hierarchical version of bloom filter which is based on B+ tree of order d. Since Bloom filters can make a tradeoff between space and time, proposing a new hierarchical Bloom filter causes a remarkable reduction in space and time complexity of recommender systems. To increase the accuracy of the recommender systems we use Probabilistic version of hierarchical Bloom filter. By measuring the accuracy of the algorithm we show that the proposed algorithms not only decrease the time complexity but also have no significant effect on accuracy.

The convergence of the stock market and the Internet has created new marketing channels in the e-commerce and stock market brokerage industry. Given the fact that the brokerage industry is customer-oriented, the principal challenge for firms active in this industry is attracting and maintaining customers and traders. This study aimed to assess the impact of the quality of financial service provider websites on customer responses among brokerages. The conceptual framework was developed Based on the S-O-R paradigm so that, corporate reputation and perceived value used as the mediator factors. This research adopted a questionnaire survey. All factors were measured with multiple items. The conceptual model was tested using PLS-SEM and Smart PLS software. Findings show that website quality is a determinant factor in predicting customer responses. The impact of website quality directly on customer perceived value and corporate reputation was significant. Also, corporate reputation and perceived value have a positive and significant impact on purchase intention and word-of-mouth. The Sobel test results, at 95% confidence level, accepted the indirect impact of website quality on word-of-mouth and the purchase intention, taking into account the mediating role of corporate reputation and perceived value. In general, this research developed a proposed framework on the role of website quality on customer and trader behaviour in the brokerage industry

Opinion formation is a collective behavior, describing the dynamics of people’s opinions due to their interactions. Nowadays, social media are broadly used and cause a lot of interactions among users who mainly know each other merely as a username, but significantly influence each others’ opinions and emotions. Both emotions and opinions spread across users in social media via their exchanged posts. Furthermore, based on psychology research, emotion affects people’s opinion. In this research, we implemented two binary classifiers to predict the users’ next opinions considering previous posts sent in online community: an original classifier, a classifier based on the social impact model of opinion formation; and an emotion-integrated classifier, a classifier based on the social impact model of opinion formation integrated with an emotion model to achieve an improved model. To evaluate the improved classifier, we used a dataset containing some debates from the CreateDebate.com website and compared the performance of the original classifier with the performance of the emotion integrated classifier. The experiment results show that considering emotions improves the accuracy and precision of the social impact model of opinion formation in social media.

Today's online communities, as a multifaceted platform, have many applications in e-commerce, marketing and e-learning. Online blogging services are one of the most popular environment for user interactions. Users share their ideas, opinions, and information in this environment. The spread of information between users plays an essential role in the success of such online communities. However, these communities face challenges in post management and information spread. Modeling the life cycle of a post provides an opportunity to examine how information is disseminated among users. In these communities, each post after creation is reposted and transmitted by users. Depending on their content and online community structure, posts are spread in different ways in the network. Some posts are rapidly becoming epidemic and some are not welcomed by users. In this article, we are looking for a method that estimates the probability of an epidemic of a post. For this purpose, a learning method based on learning automata has been used. The evaluations show that this method is efficient in three evaluation datasets. Furthermore, we will introduce self-organized posts that facilitate the management of posts in online communities.

In this paper, we describe our proposed methodology for constructing an ontology of natural language processing (NLP). We use a semi-automatic method; a combination of rule-based and machine learning techniques; to construct and populate an ontology with bilingual (English-Persian) concept labels (lexicon) and evaluate it manually. This methodology results in a complete ontology in the natural language processing domain with 1333 classes (containing concepts, tools, applications, etc.), 88 object properties, and 2437 annotation assertions for different classes. The built ontology is populated with about 428K NLP related papers and 38K authors, and also about 5M "is Related to" relations between papers and ontology classes and 1M "is Author of" relations between papers and authors. The evaluation results show that the ontology achieved a good result. The instantiation is done to enable applications find experts, publications and institutions (such as universities or research laboratories) related to various topics in NLP field.