Visual sentiment analysis has received increasing attention in recent years. However, the quality of the dataset is a concern because the sentiment labels are crowd-sourcing, subjective, and prone to mistakes. This poses a severe threat to the data-d
riven models including the deep neural networks which would generalize poorly on the testing cases if they are trained to over-fit the samples with noisy sentiment labels. Inspired by the recent progress on learning with noisy labels, we propose a robust learning method to perform robust visual sentiment analysis. Our method relies on an external memory to aggregate and filter noisy labels during training and thus can prevent the model from overfitting the noisy cases. The memory is composed of the prototypes with corresponding labels, both of which can be updated online. We establish a benchmark for visual sentiment analysis with label noise using publicly available datasets. The experiment results of the proposed benchmark settings comprehensively show the effectiveness of our method.
There is evidence of misinformation in the online discourses and discussions about the COVID-19 vaccines. Using a sample of 1.6 million geotagged English tweets and the data from the CDC COVID Data Tracker, we conduct a quantitative study to understa
nd the influence of both misinformation and fact-based news on Twitter on the COVID-19 vaccine uptake in the U.S. from April 19 when U.S. adults were vaccine eligible to May 7, 2021, after controlling state-level factors such as demographics, education, and the pandemic severity. We identify the tweets related to either misinformation or fact-based news by analyzing the URLs. By analyzing the content of the most frequent tweets of these two groups, we find that their structures are similar, making it difficult for Twitter users to distinguish one from another by reading the text alone. The users who spread both fake news and fact-based news tend to show a negative attitude towards the vaccines. We further conduct the Fama-MacBeth regression with the Newey-West adjustment to examine the effect of fake-news-related and fact-related tweets on the vaccination rate, and find marginally negative correlations.
Computer science has grown rapidly since its inception in the 1950s and the pioneers in the field are celebrated annually by the A.M. Turing Award. In this paper, we attempt to shed light on the path to influential computer scientists by examining th
e characteristics of the 72 Turing Award laureates. To achieve this goal, we build a comprehensive dataset of the Turing Award laureates and analyze their characteristics, including their personal information, family background, academic background, and industry experience. The FP-Growth algorithm is used for frequent feature mining. Logistic regression plot, pie chart, word cloud and map are generated accordingly for each of the interesting features to uncover insights regarding personal factors that drive influential work in the field of computer science. In particular, we show that the Turing Award laureates are most commonly white, male, married, United States citizen, and received a PhD degree. Our results also show that the age at which the laureate won the award increases over the years; most of the Turing Award laureates did not major in computer science; birth order is strongly related to the winners success; and the number of citations is not as important as one would expect.
