The Computing Research Association (CRA) awarded Serina Chang as one of four awardees of the 2019 Outstanding Undergraduate Researcher Award. Three other computer science students were also nominated, including Ruiqi Zhong, Justin Whitehouse, Hamed Nilforoshan, all of whom received honorable mentions. All four undergraduates were recognized for their work in an area of computing research.
“I am always looking for innovative ways to combine computer science and social science,” said Serina Chang, a senior studying computer science and sociology. “Not only looking at research methods but to have a social impact.”
Chang’s work on detecting emotion in social media, specifically tweets of Chicago gang members, led to a paper in EMNLP ’18. Under the guidance of Professor Kathy Mckeown she also published a paper on automatic text summarization in EACL ’17.
The paper on gang-involved youth in Chicago is one example of how computer science and sociology can intersect. The research continued the work of Desmond Patton, a professor at the School of Social Work and a Data Science Institute member, who found expression of loss over the death of a friend often led to tweets of aggression and real-world violence.
One of their key insights include that domain-specificity and context are extremely important. Interpreting the tweets was also context-critical, as it often depends on the emotional state of the user or what they had been tweeting prior to the current one, and may also refer to context such as the names of rival gangs or deceased friends.
The researchers hypothesized that domain-specificity and context would be important, but were surprised by the degree to which it affected convolutional neural network performance. For example, one set of experiments compared the word embeddings trained on domain-specific unlabeled corpus against word embeddings that trained on other corpora, including standard GloVE and Google News as well as an African-American vernacular English-specific corpus and a location-specific one, which they drew from the same area of Southside Chicago as their users.
The results found that their word embeddings outperformed the others by nearly 4 F1 points, which demonstrates the importance of constructing a large domain-specific corpus ourselves instead of resorting to “similar” corpora based on location or demographics. Said Chang, “Our hope is that we can eventually use our system to save critical time, scale their reach, and intervene before more young lives are lost.”
Chang’s thoughtfulness extends into other parts of her life. In her sophomore year she became a teaching assistant so she could work with students in Professor Paul Blaer’s data structures in JAVA course. She worked one-on-one with students to help them with assignments, with the additional goal of building their confidence when it comes to their abilities.
Mentoring others also came hand-in-hand with her involvement in Lean In at Columbia and Womxn in Computer Science (WiCS). In both organizations Chang held a leadership position and made it a point to foster relationships and empower women through various programs that celebrated diversity and feminism.
“I do still have time to study,” laughed Chang when asked about time management. This semester she is focused on working on her thesis. Grad school is also on the horizon and she will make her final decision in April. Part of the CRA award is financial assistance to attend a conference of their choice, she is still unsure of where she wants to go.
Looking back at all she has accomplished while at Columbia makes her feel grateful for all the opportunities she had to work with colleagues and professors, the relationships that have grown with friends and family, and the research that she hopes to continue.
“I am incredibly honored by this award and excited to be recognized for research that I am so passionate about,” said Chang.