Inevitable-Metaverse: A Novel Twitter Dataset for Public Sentiments on Metaverse

Kadhim Hayawi,Sakib Shahriar,Mohamed Adel Serhani,Eiman Alothali

2024-03-02

Abstract:Metaverse has emerged as a novel technology with the objective to merge the physical world into the virtual world. This technology has seen a lot of interest and investment in recent times from prominent organizations including Facebook which has changed its company name to Meta with the goal of being the leader in developing this technology. Although people in general are excited about the prospects of metaverse due to potential use cases such as virtual meetings and virtual learning environments, there are also concerns due to potential negative consequences. For instance, people are concerned about their data privacy as well as spending a lot of their time on the metaverse leading to negative impacts in real life. Therefore, this research aims to further investigate the public sentiments regarding metaverse on social media. A total of 86565 metaverse-related tweets were used to perform lexicon-based sentiment analysis. Furthermore, various machine and deep learning models with various text features were utilized to predict the sentiment class. The BERT transformer model was demonstrated to be the best at predicting the sentiment categories with 92.6% accuracy and 0.91 F-measure on the test dataset. Finally, the implications and future research directions were also discussed.

Computers and Society

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to understand the public's views and emotional tendencies towards the emerging technology of Metaverse. Specifically, researchers hope to evaluate people's attitudes towards the Metaverse, including positive, negative or neutral emotions, by analyzing public discussions on social media, especially relevant posts on Twitter.

### Main problems
1. **Public acceptance of the Metaverse**: With the development of Metaverse technology, especially after Facebook changed its name to Meta, what do the public think of it? Is there widespread acceptance or concern?
2. **Technical implementation of sentiment analysis**: How to accurately analyze and classify sentiment data on social media through natural language processing (NLP) and machine - learning techniques?

### Research objectives
- **Explore public sentiment**: Understand the public's emotional tendencies towards the Metaverse by analyzing Metaverse - related posts on Twitter.
- **Construct a data set**: Create a new Metaverse - related data set containing sentiment scores and sentiment categories, and publish it publicly for other researchers to use.
- **Model comparison**: Use multiple machine - learning and deep - learning models for sentiment classification, and evaluate the impact of different features and models on prediction performance.

### Method overview
1. **Data collection**: From January 1, 2019 to April 9, 2022, 749,842 English - language tweets related to the Metaverse were collected through the Twitter API.
2. **Data pre - processing**: Sampling, cleaning, removing stop words and other pre - processing steps were performed on the data, and finally 86,565 tweets were obtained for analysis.
3. **Sentiment analysis**: Use the TextBlob library for dictionary - based sentiment analysis, calculate the polarity score of each tweet, and classify it as positive, negative or neutral sentiment accordingly.
4. **Model training and evaluation**: Use multiple machine - learning (such as KNN, Naïve Bayes, Random Forest, SVM) and deep - learning (such as LSTM, BERT) models for sentiment classification, and evaluate model performance through indicators such as accuracy, precision, recall and F - measure.

### Main contributions
- Proposed a novel Metaverse - related data set containing sentiment scores and categories.
- Compared the performance of multiple machine - learning and deep - learning models in the sentiment classification task, and found that the BERT model performs excellently in predicting sentiment categories, with an accuracy rate of 92.6% and an F - measure of 0.91.

Through these methods, researchers can not only gain in - depth understanding of the public's views on the Metaverse, but also provide valuable data resources and technical references for future research.