Social Media Word Embeddings : The Studies
Here are some excellent Social Media Word Embeddings studies that are still relevant today.
Cultural Bias and the Analysis of Language
An analysis about cultural biases in word embedding yielded interesting findings. By analyzing a large dataset of cultural texts, researchers found that there is a significant difference in the way that people treat words that are associated with different cultures. For example, Europeans tend to associates words such as taxation with the legislature while Asians tend to associate these words with their families and friends. It was also found that there exists a significant cultural bias in how people analyze language data.
An evaluation of word embeddings and their impact on political science research
A study about word embeddings has shown that they are valuable for political science research because they can help to develop more accurate models of social networks. However, it is unknown how well these techniques work when used with specific parameters in order to get the most reliable results. This study looked at three different parameters - the context window length, embedding vector dimensions, and pretrained versus locally fit variant - in order to see which affected the efficiency and accuracy of word embeddings.
Fake News on Social Media Shocks World
An article about the use of social media to share fake news has shown that the way in which news is created and published has changed. The use of social media has made it easy for people to access fake news, which has had a negative effect on the way people think about the news.
Gender Stereotypes and the Sentiment of Texts
An inquiry about word embedding revealed that there is a significant difference in the sentiment of texts between men and women. The study found that the gender stereotypes that people have are based on their role in society as women or men. People become more likely to think that someone who is a woman is a weaker asset, and more likely to think that someone who is a man is more strong.
HowVolcanoes and Climate Change are Embedded in Scientific Journals
A journal about unsupervised word embeddings across 51,000 scientific journal articles was published in May 2019. The study found that the results from these écrits can provide valuable insights about volcanoes and climate change. The research was conducted by a former material physics colleague of mine who discovered new science from older scientific publications.
swarm of bees: evidence that website content is affecting public opinion
An analysis about the swarm of bees was conducted using a word network analysis. This study found that the topic of bees Swarm was highly related to other topics on the website, such as life and what it means to be a Stockbroker.
renaissance of fentanyl: a new opioid to watch
A study about a drug that is newly detected in overdose victims seeking rescue treatment shows that the drug is fentanyl, a powerful and highly addictive opioid pain medication. In this study, fentanyl was found in 81% of overdose victims. The study also found that fentanyl is quickly becoming an emerging Drug of Abuse in the United States.
The Use of Topics in Social Media: A User-Based Topic Modeling Study
An inquiry about the usage of topics in social media using user-based Topic Modeling found that the sparsity of material leads to multiple meanings and perceptions among users. To address this issue, we employed a semantic modelling method, SM-UTM, to construct a user topic model. The yielded model was found to aggregate relevant short text into individual topics with rich semantics.
The Negative Reaction to Positive Things
An evaluation about sentiment analysis reveals that people are usually more negative towards items if the item is seen as harmful or troubling to them. This occurs even if the individual feels no personal connection to the product, service, organization, or individual.
Clustering Rumoured Standpoints using an Approximate Heaviside Method
An analysis about node embedding has been carried out in the social media networks for the purpose of classifying Rumour Standpoints. This study found that there are four classes of Rumour Standpoints, and Combinatorial Clustering wasano-Heaviside Method can be used to predict these classes. The study found that nodes in the network tend to clustered according to their reported sentiment level. Furthermore, the use of word embeddings based on Hastie and Brown basis function improved the accuracy of predicting sentiment levels for rumoured standpoints.