Chih-yuan Li

Prof. Dr. Chih-yuan Li

Computer Science, St. Francis College, Brooklyn, Connecticut, USA
orcid
Artificial Intelligence and Machine Learning toward Human De...

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 November 2022 in Information
Verified
Journal Article
Open access
PDF file
Chih-yuan LiChih-yuan Li
Chih-yuan Li
MRMichael Renda
Michael Renda
FYFatima Yusuf
Fatima Yusuf
JGJames Geller
James Geller
SCSoon Chun
Soon Chun
Since the start of the COVID-19 pandemic, government authorities have responded by issuing new public health policies, many of which were intended to contain its spread but ended up limiting economic and social activities. The citizen responses to these policies are diverse, ranging from goodwill to fear and anger. It is challenging to determine whether or not these public health policies achieved the intended impact. This requires systematic data collection and scientific studies, which can be very time-consuming. To overcome such challenges, in this paper, we provide an alternative approach to continuously monitor and dynamically make sense of how public health policies impact citizens. Our approach is to continuously collect Twitter posts related to COVID-19 policies and to analyze the public reactions. We have developed a web-based system that collects tweets daily and generates timelines and geographical displays of citizens’ “concern levels”. Tracking the public reactions towards different policies can help government officials assess the policy impacts in a more dynamic and real-time manner. For this paper, we collected and analyzed over 16 million tweets related to ten policies over a 10-month period. We obtained several findings; for example, the “COVID-19 (General)” and ”Ventilators” policies engendered the highest concern levels, while the “Face Coverings” policy caused the lowest. Nine out of ten policies exhibited significant changes in concern levels during the observation period.

15 June 2020
Verified
Conference Paper
Open access
SCSoon Ae Chun
Soon Ae Chun
Chih-yuan LiChih-yuan Li
Chih-yuan Li
ATAmir Toliyat
Amir Toliyat
JGJames Geller
James Geller
COVID-19, the disease caused by the Corona Virus, started from Wuhan, China, in late December 2019, and quickly swept the Asian countries with confirmed cases and deaths. Within two and half months, it started spreading to European countries, and to the US, triggering the pandemic declaration by WHO. Governments around the globe have declared a public health crisis in specific regions and nationwide, with drastic measures taken to contain the spread of the disease. Citizens in this public health crisis are going through a wide range of emotions, such as disbelief, shock, concerns about health, fear about food supplies, anxiety, panic, etc., through directly and viscerally experiencing the disease spreading. We present an approach to measure and monitor citizens’ concern levels using public sentiments in Twitter data. Our approach shows temporal and geographic spread of citizens’ concerns during the COVID-19 public health crisis.

13 May 2024 in The International FLAIRS Conference Proceedings
Verified
Journal Article
Open access
PDF file
Chih-yuan LiChih-yuan Li
Chih-yuan Li
SCSoon Chun
Soon Chun
JGJames Geller
James Geller
The spread of fake news has emerged as a critical challenge in the digital era. Confusion and conflict can arise if people mistake fake news for real news. Thus, advanced detection methodologies are desired. This paper aims to identify fake news, while addressing the issue of class imbalances. We employ multi-class fake news detection, an advanced methodology beyond traditional binary classification. We highlight CNN’s better performance over the baseline BERT model in the literature, with improvements in accuracy, precision, recall, and F1-Score. We uniquely experimented with four model variants: CNN and BERT with both trainable embeddings and BERT embeddings. Our experiment demonstrates CNN's effectiveness in identifying text patterns. To address class imbalances, we experimented with three different balancing methods. Our study includes fine-tuning ChatGPT for multi-class classification. The result indicates notable limitations in ChatGPT's automated classification, which highlights the complexities of AI-based categorization. Our findings demonstrate the CNN model's efficiency and effectiveness, and show the intricacies of fake news detection. These insights confirm the need for advanced AI methodologies in combating misleading information.

24 February 2025 in Preprints.org
Verified
Preprint
Open access
PDF file
Chih-yuan LiChih-yuan Li
Chih-yuan Li
SCSoon Chun
Soon Chun
JGJames Geller
James Geller
Social media allows people to express and share a variety of users’ experiences, opinions, beliefs, interpretations, or viewpoints on a single topic. Summarizing a collection of social media textposts (microblogs) on one topic may be challenging and can result in an incoherent summary due to these multiple perspectives by different users. We introduce an approach of microblog summarization based on user perspectives, called Multiple View Summarization Framework (MVSF), a novel approach designed to efficiently generate multiple summaries from the same social media dataset depending on a chosen perspective, delivering personalized and fine-grained summaries. The MVSF leverages component-of-perspective computing that can recognize the perspectives expressed in microblogs, such as sentiments, critical views, political orientations, or unreliable opinions (fake news), etc. The perspective computing can filter social media data to summarize them according to a specific user-selected perspective. For the summarization methods, our framework implements three extractive summarization methods: Entity-based, Social Signal-based, and Triple-based. We conduct comparative evaluations of MVSF summarizations against state-of-the-art summarization models, including BertSum, SBert, T5, and Bart-Large-CNN, by using a gold standard BBC news dataset and Rouge scores. Furthermore, we utilized a dataset of 18,047 tweets about COVID-19 vaccines to demonstrate the applications of MVSF. Our contributions include the innovative approach of using user perspectives in summarization methods as a unified framework, capable of generating multiple summaries that reflect different perspectives, in contrast to prior approaches of one summary for one dataset. The practical implication of MVSF is that it offers end-users diverse perspectives from social media data. Our prototype web application is also implemented using ChatGPT to show the feasibility of our approach.

4 May 2022 in The International FLAIRS Conference Proceedings
Verified
Journal Article
Open access
PDF file
Chih-yuan LiChih-yuan Li
Chih-yuan Li
SCSoon Chun
Soon Chun
JGJames Geller
James Geller
While the world has been combating COVID, there has also been an ongoing “Infodemic,” caused by the spread of fake news about the pandemic. Due to the rapid data sharing on social media, the impact of fake news can be quite damaging. Citizens might mistake fakes news for real news. Human lives have been lost due to fake information about COVID. Our goal is to identify fake news on social media and help stem the spread by deep learning approaches. To understand the different characteristics in fake and real news, we conducted behavioral and sentiment analyses between fake and real news regarding the COVID pandemic. We then further built detection models based on feature elimination, and we identified differences of model robustness based on selected features.

8 May 2023 in The International FLAIRS Conference Proceedings
Verified
Journal Article
Open access
PDF file
Chih-yuan LiChih-yuan Li
Chih-yuan Li
SCSoon Chun
Soon Chun
JGJames Geller
James Geller
Social Media provide voluminous posts about current topics and events. When a user desires to investigate a popular topic, it is not feasible as there are many posts. Besides, posts show different biases, viewpoints, perspectives, and emotions. Thus, providing summaries of large post sets with different viewpoints is necessary. We develop a multiple view summa-rization framework to generate different view-based summar-ies of Twitter posts. Users can apply different methods to generate summaries: 1) Entity-centered, 2) Social feature-based, 3) Event-based summarization, using all triple embed-dings and 4) Sentiment-based summarization to generate summaries of positive or negative views of tweets. These summarization methods are compared with BertSum, SBert, T5, and Bart-Large-CNN with a gold standard dataset. Our results, based on Rouge scores, were better than these pub-lished extractive and abstractive summarization models.