topic modeling

Researches have demonstrated that many organizations leverage the analysis of user-generated opinions and content on social networks to cultivate knowledge and enhance decision-making. This study aims to analyze the financial performance of non-profit organizations by examining user-generated opinions. The dataset comprises 26,714 user tweets from 23 global Twitter accounts and ten years of financial data (2010-2020) from five selected non-profit organizations. The results of topic modeling and sentiment analysis were converted into panel data. These results were analyzed using ordinary least squares (OLS) and dynamic ordinary least squares (DOLS) methods.Findings indicate a significant relationship between user-generated content and the financial performance of non-profit organizations. While positive sentiments did not exhibit a significant impact on financial performance, negative sentiments and retweets demonstrated a negative relationship, whereas likes showed a positive correlation with financial performance. By harnessing insights from user-generated content, non-profit organizations can optimize their content strategies to improve financial outcomes. Nevertheless, it is imperative to consider the adverse effects of negative sentiments and retweets to achieve substantial improvements in financial performance.

In recent years, the rapid growth of virtual space has made people devote more of their time in virtual space, especially to social networks, which can be attributed to the remarkable features of virtual space; including increasing the speed of information exchange, easy and free access to information and variety of knowledge topics. In this regard, the opinions recorded by users in virtual networks have grown day by day and have become very important, and extracting the opinions and feelings of users' opinions for more informed decision-making is of great help to businesses, on the other hand, virtual reality technology in the past few decades It has undergone technical changes and improved immersion and the feeling of remote presence; This technology is used in various fields such as education, tourism, health, sports, entertainment, architecture and construction, etc. The increasing progress of virtual reality technology has caused many businesses to operate in this field, but due to changes Continuous market and the need for timely information, companies should use differentiation and growth strategies, in this regard, they need to ask users' opinions and in line with that, try to grow and improve their business, considering that Users' comments are textual, and reading and summarizing them is time-consuming and difficult. Based on this, the aim of the current research was to categorize comments related to virtual reality technology using machine learning methods and a dictionary-based approach. Therefore, about one million tweets in the field of virtual reality technology were collected by the web crawler, and after data preprocessing, 480,432 samples remained in the data, then Dirichlet's hidden allocation topic modeling was implemented on the data. This modeling separated different topics by examining the distribution of words in tweets; The tweets whose distribution of words were similar were placed into a topic and the number of topics with the highest coherence score was selected, the number of topics 9 had higher coherence and the data were grouped into 9 topics, so once again the Dirichlet hidden allocation modeling was set to 9. The topic was done, with this the tweets were grouped into 9 different topics. To evaluate the model, considering that we had a probability distribution, the confusion criterion was used, the value of which was -9.44, and the coherence score was used for the degree of semantic similarity between words and the distinction between subjects, and the result was 0.47. The lower the confusion criterion and the higher the coherence score, the more efficient the model is. With the help of keyword weights obtained by Dirichlet hidden allocation modeling and examining at least 5 different tweets from each topic, 9 topics related to virtual reality technology were identified: "New Technology", "Creation and Make", "Technological Business", "Education", "Virtual Games", "Progress", "Gadget", "Metaverse", and "Indiegame", the topics were analyzed with the help of several graphs. We found that the number of neutral comments on topics such as "New Technology" and "Metaverse" is more than positive and negative comments, which indicates the lack of sufficient information or the lack of use of these technologies, and it is necessary for businesses in this field, to try more in this regard, in the same way, if we observe the graph of "Virtual Games" and "Technological Business", we can see that it changes almost with the same ratio in different years, in the sense that this The two graphs are related, in fact, businesses should keep in mind that the factors affecting these two issues are the same, but users pay more attention to the issue of "Virtual Games", as a result, if the creators of "Technological Business" Focus specifically on "Virtual Games", they will grow more due to the more attention of users, also the creators of games should consider that "Virtual Games" are a topic of more attention than "Indiegame". Is. In the subjects of "Education" and "Gadget", users lost their attention to these subjects in the field of virtual reality over time, in fact they showed their attention to other subjects, so it is better for businesses that operate in this field to take measures To advertise and attract users or change their user area if there is no growth.

Constant changes in the market and the need for timely information force companies to use differentiation and growth strategies appropriate to the needs of customers. (Sánchez, Folgado-Fernández, & Sánchez, 2022). Companies can check and analyze their customers' opinions through microblogging sites (Facebook, Twitter, etc.) and finally improve the desired products or services (Ahmad, Aftab, Bashir, & Hameed, 2018). Today, users express their opinions and feelings and review products in online social networks. Therefore, user comments and the analysis of these comments have become a valuable resource for businesses (Kim et al., 2015; Loureiro et al., 2019).
Virtual reality and augmented reality have undergone technical developments in the past few decades and have improved immersion and the feeling of remote presence. Several examples of applications of such techniques can be found in stores, the tourism industry, hotels, restaurants, etc. (Loureiro, Guerreiro, & Ali, 2020). Due to the constant changes in the market and the need for timely information, companies should use differentiation and growth strategies, nowadays, due to the rapid evolution of the Internet, instead of collecting their opinions through time-consuming and expensive methods such as questionnaires and interviews, etc., they express in the context of social networks, which is very useful for businesses in their development, and they can measure the feelings of customers towards products and services, and understand the needs of users, and finally make appropriate and appropriate decisions in the direction of adopt growth, but in order to use the produced content correctly, text mining and sentiment analysis techniques should be used, which has not been researched in Iran so far. Analysis of users' opinions and feelings about virtual reality technology can help businesses that operate in the field of metaverse, virtual game production, virtual education, virtual tourism, etc., to make better decisions and plans.

Social media generates a large amount of real-time social signals that can provide new insights into human behavior and emotions. People around the world are constantly engaged with social media. (Al-Samarraie, Sarsam, & Alzahrani, 2023).
On the other hand, the amount of data is increasing day by day. Almost all institutions, organizations and business industries store their data electronically. A huge amount of text is circulating on the Internet in the form of digital libraries, repositories, and other textual information such as blogs, social media networks, and emails (Sagayam, Srinivasan, & Roshni, 2012).
Topic modeling is one of the most powerful techniques in text mining for data mining, discovering hidden data and finding relationships between data and textual documents (Jelodar et al., 2017).
The technological advances of the last century have confronted societies with new realities that have indisputably improved daily life, making it more convenient and interesting. In recent decades, technology using virtual reality and wearable devices have had a significant impact in the fields of education, tourism, health, sports, entertainment, architecture and construction, etc. (Kosti et al., 2023).
Virtual reality is a technology that allows a user to interact with a computer-simulated environment, whether that environment is a simulation of the real world or an imaginary one. With virtual reality, we can experience the most frightening and overwhelming situations with safe play and a learning perspective (Mandal, 2013). Most people are curious about the possibilities and future of new technologies, considering the various applications it is supposed to offer such as virtual meetings, learning environments and many others, however, there are also concerns about potential negative effects. because real world signals can be transmitted in the virtual world. In this regard, people express their feelings in different social networks (Bhattacharyya et al., 2023).

According to the main goal of the research, which is to classify comments related to virtual reality technology using machine learning methods and a dictionary-based approach, therefore, about one million tweets in the field of virtual reality technology were collected by the web crawler and After data preprocessing, 480,432 samples remained in the data, then Dirichlet hidden allocation thematic modeling was implemented on the data. By examining the distribution of words in tweets, this modeling tries to separate different topics by detecting the distribution of words; The tweets whose distribution of words are similar were put into a topic, and the number of topics with the highest score was selected, the number of topics 9 has higher coherence, and the data was grouped into 9 topics, so once again, Dirichlet hidden allocation modeling was applied 9 topics were done, whereby the tweets were grouped into 9 different topics. Considering that we have a probability distribution, the confusion criterion was used to evaluate the model. The lower the confusion criterion and the higher the coherence score, the more efficient the model is. With the help of keyword weights obtained by Dirichlet hidden allocation modeling and examining at least 5 different tweets from each topic, 9 topics related to virtual reality technology were identified: "New Technologies", "Creation and Make", "Technological Business", "Education", "Virtual Games", "Progress", "Gadget", "Metaverse" and "Indiegame" were named.

In this research, by examining topics in different years, we observed that the topic of "Progress" was the most popular topic among users from 2017 to the end of 2021, in early 2022, this topic gave way to "Metaverse", currently "Metaverse" is one of the most popular topics being discussed by users. Businesses in the field of virtual reality should strive for the attractiveness of "Metaverse" and attract users. Likewise, if we observe the "Virtual Games" and "Technological Business" graphs, we can see that they change with almost the same ratio in different years, meaning that these graphs are related to each other, in fact, business and keep in mind that the factors affecting these two issues are the same, but in the case of "Virtual Games" it has more effects, and if "Technological Businesses" specifically focus on virtual games, they will grow more due to the greater attention of users. had Similarly, "Indiegame" which have had a series of changes but in recent years have had a declining trend and then no change, now the creators of these games should check, and in general "Virtual Games" are a more interesting topic than "Indiegame". In the subjects of "Education" and "Gadget" it has been decreasing since the beginning of 2017, which shows that users lost their attention to these subjects in the field of virtual reality over time, in fact to other topics showed their attention, so it is better for businesses that are active in this field to take measures to advertise and attract users, or change their user field if there is no growth.

Due to the global attention to cryptocurrencies especially bitcoin, governments, banks, and other industries are expected to use cryptocurrencies in their daily transactions. Therefore, as any other financial field, there is a need to identify the challenges in this field to safe investment. On the other hand, by the expansion of social networks, unstructured data is increasing, which can be used to create added value in various areas such as sentiment analysis. Therefore, this study was conducted to investigate the impact of perceived risk by social network users on the price of Bitcoin. According to the research findings, the identified risks in the field of bitcoin include social, economic, security, technological and legal risks., The conversations on the Bitcoin Talk Site were used to extract the bitcoin risks. After collecting these conversations by the web crawler, the conversations were clustered into thematic clusters using the Latent Dirichlet Allocation algorithm, which is one of the most popular methods in Topic Modeling. were analyzed using vocabulary-based method and AFINN dictionary. NARX Neural Network was used to measure the effect of Userschr('39') Sentiment on the price of Bitcoin. The results show a correlation of 0.99 and a mean square error of 0.001, which means that there is a correlation between the actual price and the predicted price of Bitcoin. The findings of this study can attract the attention of financial actors and businessmen in the field of bitcoin to plan a safe investment and reduce risk.

With the development of mobile communications and the advent of the Fifth Generation, increasing Internet speed and capabilities as enablers of the Internet of Things is creating changes in people's daily lives and work. On the other hand, there are environmental issues, the possibility of pathogenicity, and doubts about the realization of the imagined features, which are much debated today. Most researches has paid attention to technical and developmental dimensions of it, while from the social science perspective, the acceptance of new technology and from the marketing perspective, consumer satisfaction can be important for service providers. Accordingly, the purpose of this study is to investigate the perception of Twitter social network users as a microblogging platform about the fifth generation of mobile communication using machine learning methods and Sentiment Analysis .Therefore, a collection of more than 40,000 tweets on this topic was compiled through the Twitter user interface and the Recurrent Neural Network model was created with79% accuracy. Finally, the topic modeling was done by LDA method for further deepening. The results shows that although there is dissatisfaction with the quality provided, the cost and coverage of the fifth generation of mobile communications, health concerns, rumors that Covid19 is related to the fifth generation, but still more people are optimistic about the future of this technology in various fields such as the Internet of Things and artificial intelligence..

Today, social media has become a place for users to comment on various topics, including consumer products, and these comments have become a valuable resource for sentiments analyzing and extracting customer expectations of products. This subject provides companies with a good opportunity to redesign their products based on user feedback. In this study, to provide recommendations to the design units of detergent products, we used sentiment analysis of customers and consumers of this products on social media. More than 4200 tweets were extracted from Twitter in 2019 based on the research topic and refined and tagged during the pre-processing of the texts. Afterwards, we analyzed the emotions using fuzzy logic and topic modeling. We have used topic modeling to find the features mentioned in the comments for a better approach in the design units, and fuzzy logic to obtain the degree of polarity of ideas into 5 categories: very positive, positive, neutral, negative and very negative. We used confusion matrix for evaluating research model and an accuracy of 86.15 % has been recorded.In this research Python libraries are used for data gathering, cleansing and analysis.

The dynamic capabilities perspective seeks to provide an appropriate explanation for how successful firms are coping with rapidly changing business environments and creating competitive advantages to respond to consistent environmental changes. Although the dynamic capabilities perspective is one of the most prominent theories in the management discipline and has affected the different subfields of this discipline, remarkable concerns have arisen about the intellectual structure of this perspective’s literature and its field of research. Dispersion and clutter of theoretical foundations, confusing plurality and diversity of conceptualizations, lack of a holistic framework of existing views, disconnected understandings and findings, and lack of coherence in the body of knowledge are instances of these concerns. Accordingly, the current study aims to develop a holistic insight into the topical structure and trend of the field and contribute to addressing the concerns challenging the identity of the literature as a whole unit by comprehensively reviewing the scholarly documents on dynamic capabilities.  

The present study has used the text mining approach and topic modeling method. This approach and method is a powerful tool that can be used to analyze a huge amount of documents in a short time and with the required efficiency. For this purpose, first, articles in English-language journals related to the subject of dynamic capabilities and indexed in one of the Web of Science or Scopus databases until the beginning of 2019 were collected as the target collection of documents. Then, considering that the abstract of a research is the extract of its main findings and reflects the highlights of the study, the abstracts of the collected documents were considered as the set of texts required for textual analysis. After storing the bibliographic information of the retrieved documents and deleting duplicate documents or items without abstract, the number of abstracts reached 4206. In the next step, in order to pre-process and clean the collected text, measures such as unifying the uppercase and lowercase letters of the words, removing numbers and punctuation marks, deleting stop words, and merging the words with the same stem were performed. In addition, words with insufficient value were identified and eliminated using one of the common valuation criteria. Finally, using the latent Dirichlet allocation algorithm, topic modeling of the collected abstracts was performed, as a result of which the topical structure of the corpus and the temporal trend of each topic were obtained.

As a result, the topical structure and thematic trends in the field of dynamic capabilities were obtained. The topical structure refers to twelve prominent topics in the field of dynamic capabilities that revealed as a result of topic modeling, and the thematic trends refer to the publication trend of scientific documents related to each of these topics. Each topic is a group of words that contains a probabilistic distribution of all the words in the corpus. In this group, the words that have the greatest proportions represent the dominant theme. The identified topics are: 1. Firm performance (focusing on the relationship between dynamic capabilities and firm performance), 2. Organizational implications (by the operational implications of dynamic capabilities for organizations), 3. Organizational change (by the transformational role of dynamic capabilities in organizations), 4. Theoretical framework (focusing on the theoretical concept, principles, assumptions, background, and development of dynamic capabilities), 5. Competitive advantage (based on the role of dynamic capabilities in creating firms’ competitive advantage), 6. Technological alliance (focusing on technological alliance capability as a dynamic capability), 7. Product/service innovation (focusing on innovation capability as a dynamic capability), 8. Innovation performance (focusing on the relationship between innovation as a dynamic capability and organizational performance), 9. Knowledge network (by  networking and knowledge management capabilities as dynamic capabilities), 10. International market (focusing on the role of dynamic capabilities in responding to the need of competitive advantage creation in the face of ever-changing international environments), 11. Information technology (focusing on the information technology and information systems-enabled dynamic capabilities), and 12. Supply chain (focusing on the dynamic capabilities of supply chain and operation). The revealed thematic trends stated that among these topics, topic 8 (i.e., innovation performance) is an emerging topic in the field of dynamic capabilities, and also, other topics have had a growing trend in recent years.

As a complementary study, the results of the topic modeling can be presented through some questionnaires or interviews to experts in the field of dynamic capabilities to benefit from their specialized feedback to achieve a more accurate interpretation and conceptualization. In addition, it is worth noting that the use of text mining approach and topic modeling method in comparison with the use of conventional review approaches such as meta-synthesis and systematic literature review methods, provides only an overview that some researchers consider a limitation of this type of analysis. Therefore, researchers can develop some particular methods of textual analysis to overcome this limitation to some extent, or use these methods in cases where the benefits of this type of analysis (such as rapid analysis of large volumes of texts, avoidance of mental judgments, and development of an overview perspective on the knowledge structure and research trend of a field of study) can overcome the limitation. Researchers interested in studying dynamic capabilities are also suggested to carefully examine the scope of each of the topics identified in the present study to deepen the structure presented in this study and to enumerate its underlying layers. In the meantime, according to the results, addressing performance and innovation-related issues has a higher priority in the dynamic capabilities literature. In addition, it is worth noting that due to the limited number of documents related to each of these topics, researchers can use qualitative methods along with text analysis and scienceometrics techniques to achieve this end.

v Recommender systems utilization has proven sales enhancement in most e-commerce platforms. This system objected to provide more options, comfort and flexibility to user which could make him interested, as well as providing better chance for companies to increase sells in their products and services. Flourishing popularity of web site has originated intrigue for recommendation systems. By penetrating in infinite fields, recommendation systems give deceptive suggestion on services compatible with user precedence. Integrating recommender systems by data management techniques to can targeted such issues and quality of suggestions will be improved considerably. Recent research reveals an idea of utilizing social network data to refine weakness points of traditional recommender system and improve prediction accuracy and efficiency. In this paper we represent views of recommender systems based on Twitter social network data by usage of variety interfaces, content analysis Methods, computational linguistics techniques and MALLET topic modeling algorithm. By deep exploration of objects, methodologies and available data sources, this paper will helps interested people to develop travel recommendation systems and facilitates future research by achieved direction.

Recommender systems utilization has proven sales enhancement in most e-commerce platforms. This system objected to provide more options, comfort and flexibility to user which could make him interested, as well as providing better chance for companies to increase sells in their products and services. Flourishing popularity of web site has originated intrigue for recommendation systems. By penetrating in infinite fields, recommendation systems give deceptive suggestion on services compatible with user precedence. Integrating recommender systems by data management techniques to can targeted such issues and quality of suggestions will be improved considerably. Recent research reveals an idea of utilizing social network data to refine weakness points of traditional recommender system and improve prediction accuracy and efficiency. In this paper we represent views of recommender systems based on Twitter social network data by usage of variety interfaces, content analysis Methods, computational linguistics techniques and MALLET topic modeling algorithm. By deep exploration of objects, methodologies and available data sources, this paper will helps interested people to develop travel recommendation systems and facilitates future research by achieved direction.