“When you enjoy something, it doesn’t feel like work,” said Rice University alumna Karin Verspoor (BA ’93). She double majored in Computer Science and Cognitive Science and added courses in French and Japanese, taking 18-19 credits every semester.
“There was a moment when I was simultaneously taking an artificial intelligence class in Computer Science (CS) and a CogSci class on the psychology of language, taught by Livia Polanyi. My CS professor started talking about expert systems and grammars and how we could try to use computers to process grammar.
“Livia was talking about how the brain processes syntax and syntax structure, and I thought, ‘Wait a minute. They are talking about the same thing but from different perspectives.’ Psychology was looking at how humans use language, and CS was exploring how to codify it in rules a computer could understand.”
She credits Polanyi with approving a plan that eventually led her to graduate school and careers spanning research, industry, and academia. Polanyi’s students were required to set up a research experiment and submit a paper on their findings at the end of the term.
Verspoor said, “I convinced Livia to let me write a program for my research rather than conduct a psychology experiment. My program would resolve part-of-speech ambiguities, such as to determine whether ‘orange’ is being used as an adjective or noun by looking at contextual constraints. It turned out that Livia was already looking at formal models of language in her own research and said, ‘Yes, go for it.’ She was very supportive and that experience was fascinating to me on a deep level.
“Computers treat words as binary strings, so I had to consider how a computer would interpret cognitive complexities that humans deal with easily and naturally. I wanted to keep researching natural language processing so I looked into grad school.”
The University of Edinburgh was one of the few graduate programs offering a combination of both Computer Science and Cognitive Science with a focus on natural language. Verspoor earned her Ph.D. in four years and headed to Macquarie University in Australia for a year as a postdoc.
She said, “I loved working in research, but felt drawn back to the States. I was primarily applying for jobs in research labs until a conversation with a startup founder convinced me to dive into the dot com bubble.
“Ben Goertzel was launching Intelligenesis Corp, a company developing artificial intelligence software. He was one of the most intelligent and inspiring people I’d ever met, and language processing was going to be a core part of his new ‘thinking machine’ so I was hooked. Looking back, it was a 25-year research project masquerading as a startup.”
When the company imploded during the dot com bust, Verspoor moved to another startup where she helped develop a state-of-the-art natural language processor that could create an abstract of a long document. She said their summarization technique was quite good and she wanted to enter it in an evaluation conference organized by the National Institute for Standards and Technology (NIST). Verspoor felt their algorithm could be proven publicly by participating, and that their accomplishments would benefit from rigorous scientific examination. Her superiors disagreed, fearing negative commercial repercussions if they did not win first place.
Verspoor’s frustration made her realize how much she valued being able to share research findings in a scientific environment. She accepted a job at the Los Alamos National Laboratory where she would be valued for both her research orientation and her industry experience, and allowed to publish her research.
“At Los Alamos, I would be the only computational linguist and could hire my own team. Except my team money was reallocated before I arrived. When I showed up, they told me I still had a job but needed to find something else to do since the program I’d been hired to create was gone.”
One of her new colleagues invited Verspoor to collaborate on bioinformatics and protein biology research. He began teaching Verspoor about molecular biology and she started text mining research papers, to understand the connection between proteins as described in the language of scientific publications. Although she was working in computational biology, she employed the same principles she had learned in her first research project at Rice.
“I was back in my happy place. Learning about proteomics and genomics just added another dimension. For five years, it was rewarding work, but researching computational biology and computational linguistics were simply not top priorities at a nuclear weapons facility.
“I had a choice: stay and work on whatever projects were thrown at me, or look around. I was motivated by the biomedical domain and the idea that as a computer scientist I had something to contribute to medicine.
“So I turned down industry offers and took a soft money academic position at the University of Colorado School of Medicine. My husband and I had three children and I was the primary breadwinner when I walked away from a stable government job for an insecure position, but I had to do it.”
Verspoor’s first NIH grant proposal, focusing on text mining, was funded. She’d successfully transitioned back into an academic environment, attracted good students, hired a postdoc, and increased her publications. But working on soft money meant she had to continually seek grant funding.
After three years in Denver, she started thinking about stability. The hard money associated with researcher positions at the National Information and Communication Technology Australia (NICTA) felt like a safe bet. She and her family changed continents and she continued working in biomedical text mining, leading a group of five research staff members.
“We worked out of a location on the University of Melbourne campus for two years. Then the hard money disappeared and 65 people lost their jobs, including me and my entire team. I was invited to apply to the University of Melbourne where I am now a professor in the School of Computing and Information Systems.”
Despite her career success in research, startups, and academia, Verspoor understands the challenges women face in a male-dominated discipline like computer science. She never had a female CS lecturer while she was an undergraduate.
“I started asking questions about the gender imbalance as an undergraduate, and it’s still an issue today. Something needs to change,” she said.
“In our very first semester programming class at the University of Melbourne, the young men appear to have an advantage because many have dabbled in computers in high school while many of the young women have not. There should not be a gap in their performance, because we don’t assume any coding knowledge. But there is still a perceived difference in experience. That’s one of the reasons I created the Sex-differentiated Student Experiences in Tertiary Computer Science (SSET-CS) program.”
SSET-CS aims to create a friendly setting where female students feel more secure asking questions and expressing ideas. Weekly lectures may be attended by as many as 1000 students in the introductory course but department-run tutorials that augment the lectures are restricted to about 30 students.
Verspoor said, “We’ve turned one of the weekly tutorials into a women-only option for female students with a female tutor and a female demonstrator (assistant tutor). The initial feedback on the program was quite positive.
“SSET-CS has run the female-only tutorial option for two semesters. After the first semester, 50% of the women in the female tutorial continued into the next course. Out of the women in the general population, only 30% continued the next semester. We can’t say right now what is driving the difference, but we suspect there is some relationship to the supportiveness of that early learning environment.”
Verspoor wants to see the number of women in computer science go up; a more diverse discipline is a stronger and more innovative discipline.
“We won’t change the industry until we change the students who will make up that industry,” she said.