Chuck Wooters Returns to ICSI - Again

Wednesday, September 5, 2012

Last month, ICSI welcomed back one of its first employees: Chuck Wooters. This is Chuck’s third time at ICSI.

Chuck grew up in the San Gabriel Mountains in southern California and entered UC Berkeley as an undergraduate. He planned to get his bachelor’s degree in philosophy and then go to law school, but after taking a course in philosophical linguistics, he decided to switch majors to linguistics. His plans changed again when he became interested in computer science. He decided he would get his law degree, work as an attorney for a while, and pursue computer science later. But a linguistics professor, William Wang, pointed out that the work load of being an attorney would prevent him from developing his interest in computer science. Wang said, “You’ll never do it if you get a law degree.”

ICSI's Chuck WootersChuck decided to sign up for the GREs, which were only a couple weeks away, and later entered the graduate program in linguistics at UC Berkeley. He received his master’s degree in 1988 but decided didn’t want a PhD in linguistics because he wasn’t sure he’d be able to find a job working with computers. But he couldn’t switch to computer science since he hadn’t taken the math courses required by the department. In the end, he decided to design his own PhD program.

“Really what I was interested in was speech recognition,” he said.

So he created a program and pulled together a committee of professors from the Computer Science, Linguistics, and Psychology Departments. He became the first – and possibly the only – student to receive a doctorate in “speech recognition.”

He was also the first graduate student supervised by Nelson Morgan, the leader of the Realization Group (which would later become the Speech Group). While working on his master’s during a semester abroad in Taiwan, Chuck had built a neural network classifier to distinguish B, D, and G sounds. This impressed Morgan, who used neural networks extensively in his speech recognition research.

Chuck joined ICSI in 1989 and worked here through the granting of his doctorate in 1993. “I didn’t need the support of a department because I was here at ICSI,” he said. “I felt like this was my department.”

Morgan said, “Chuck’s skills spanned a range from computer programming to linguistics, and he picked up all the necessary speech engineering tools he needed over the next few years.”

Among his skills, Morgan said, is his ability to come up with bad puns. “This was part of his charm,” Morgan said. “He was really quite wonderful to have around.”

While at ICSI, he worked on the Berkeley Restaurant Project (BeRP), a spoken dialogue system that made restaurant recommendations based on criteria that users spoke into it. Chuck worked with Dan Jurafsky, at the time an ICSI postdoc who is now a professor at Stanford, to integrate a natural language understanding parser into the speech recognition system.

“I didn’t need the support of a department because I was here at ICSI.  I felt like this was my department.”Chuck went to the U.S. Department of Defense, where he worked on applied speech recognition research. He then worked at a start-up that was building a voice-activated robot arm used in medical surgeries. It’s the only speech recognition system approved by the Food and Drug Administration.

He also worked at the Department of Defense a second time and then BBN Technologies before returning to ICSI, where, among other things, he worked on the initial software for the FrameNet database.

Then, in mid-2007, he accepted a position at Next IT Corp, which builds and maintains virtual assistants on corporate Web sites. Clients include major airlines and the U.S. Army. To use the assistants, users type questions into a dialog box, and a natural language understanding (NLU) system analyzes the questions and provides a response. His first job at Next IT was to build a bridge between the NLU system and a speech recognition system so that users could speak their questions. But the company soon realized it could never compete with commercial speech recognition powerhouses like Google, AT&T, and Nuance, so it decided to focus exclusively on NLU. Among other things, Chuck worked on an algorithm to do parallel clustering of large quantities of text.

In the summer of 2009, he was given leave to attend a ten-week workshop at the Human Language Technology Center of Excellence at Johns Hopkins. It was the first iteration of the Summer Camp for Advanced Language Exploration, which has been held every year since.

"I now have a perspective that includes for-profit, government, and academic research." he said. While he has enjoyed his time in applied research, "I love the creative atmosphere and freedom provided by academic research."

In July 2012, he returned again to ICSI.  “It’s like coming home,” he said. “I grew up at ICSI.” Morgan says Chuck was his "baby," and in his office, Morgan still displays a pair of tennis shoes that he had bronzed when Chuck graduated.

Right now, Chuck is relearning the tools used for speech recognition at ICSI. He’s also building a virtual machine that provides access through one simple interface to all the tools commonly used by ICSI speech researchers, including general tools, like the Hidden Markov Model Toolkit, and those developed at ICSI, like Relative Spectral Processing. He will be working remotely from Spokane, Washington.

“I’m really happy to be back,” he said.