I began working with Megumi in 1989 when she was at CSLI. She had excellent ideas about the difficulties in machine translation and how to approach them; I was interested in how inference and world knowledge could help in translation; the collaboration was a very productive one, and very exciting for me. This led eventually to her being hired at SRI.

She became a principal part of our text understanding research group. Her work was mainly in the areas of reference resolution and in Japanese language processing. In the latter area, she necessarily worked alone, since she was the only person we had who knew Japanese, and she had to build up the infrastructure for dealing with the Japanese language, from scratch, largely on her own. She had to do both the technical development and the marketing with Japanese clients, both of which she did supremely well. In the early 1990s, we were in a funding trough with the U.S. government, and it was primarily Japanese corporate clients that sustained us for several years. So Megumi's contribution was essential and substantial almost from the moment she came on board.

To give you an idea of how outstanding Megumi was technically, let me tell you about the message understanding system evaluation we went through in 1993. This was sponsored by DARPA, and a number of research sites from around the country and the world participated in it. The idea is that the organizers give each site a specification of the events of interest, and each text understanding system has to be developed to recognize that class of events automatically in newspaper articles. That year, they had two tracks, one for English and one for Japanese. About half a dozen of us worked on the English system, including Megumi working on the reference resolution component, and we came in in the top three out of about 15 systems -- a real success. But meanwhile Megumi also was working on the Japanese version of the system, essentially by herself, with a little bit of infrastructure help from Mabry Tyson. Her Japanese system did just as well as the English system half a dozen of us had developed.

In 1995, the evaluators introduced an evaluation of reference resolution components in isolation. This was for the English language only. Megumi led and did the bulk of the work for SRI's entry, and out of the seven sites entering, she placed first.

In parallel with all this system building, Megumi carried on an active program of theoretical work. Compliments are easy to give, and it's easy to cite people's papers. But when people give you travel money to come to a conference in Europe or Japan, that's real. Then you know your work is good. Megumi was always going off to workshops and conferences in Holland and Germany and Japan. Her research enjoyed a worldwide reputation.

When I sent out a message with the sad news about Megumi's death to a number of people who had known her and worked with her. I got a reply from a linguist at the University of Pennsylvania saying they had planned to read and discuss a paper of Megumi's in a seminar the very next day. She said she now believed that would be a fitting memorial to Megumi. I think that would have pleased Megumi. It would have pleased her to know that her work continues to be alive and influential, even though she has left us.

I want to switch now from her technical accomplishments to her artistry. The episode I want to relate is something I was reminded of when I saw recently what a terrific artist her daughter Nina is. For the workshop after the 1993 evaluation, we had to put together a poster presentation on our text understanding system. At the time we had a graphical interface that we used to specify our finite-state grammars. We thought we should include one or two of these finite-state diagrams in our poster presentation, to give people an idea of how our system worked. Then we looked at the diagrams Megumi had been producing, and we were blown away. Never mind that she had written good grammars; her diagrams were positively works of art. In the poster session, we devoted a complete panel to her entire grammar.

She lived her life with this kind of gentle artistry, using it to make the lives of those around her a little bit easier and a little bit brighter. I remember one incident that was completely insignificant in itself, but to me represented the kind of person Megumi was. I was sitting next to her at a talk at CSLI. There were people seated all along a table. The speaker gave the first one a stack of handouts and they were passed down the line, each person taking a handout off the top and passing the stack along. When it got to Megumi, she did something slightly different. She took the top one off for herself, took the next one off for me, and then passed the stack along. No fanfare. It was just something she casually did. With this almost microscopic deed, she was trying to make the lives of the people around her a little bit easier and a little bit more pleasant. I could easily have missed it, and it made me realize that there were probably hundreds of similar things she had done that I had missed, but that in aggregate made the world she inhabited a better place.

In October 1997, just three months before she learned she had cancer, we were at a workshop in San Diego. Megumi and I decided to play hookey one afternoon and go sea kayaking. We rented the kayaks, fought our way through the breakers, and had a pleasant two hours paddling out among the seagulls and pelicans. As we paddled out, we stayed pretty much together. But Megumi was doing aerobics several times a week at that time, and she was in terrific condition. On the way back, try as I might, I fell farther and farther behind.

And that's the way I would like to remember Megumi: In the peak of her form. Leaving me in her wake. Heading for the shore.

At Megumi's memorial service, her brother recalled two incidents from Megumi's childhood that really captured the kind of person she was. The first was her first haiku, when she was three. It had snowed the night before, covering the rocks in their rock garden. About midmorning Megumi looked out the window and saw the snow beginning to melt, allowing the rocks to show through, and she said

The rocks are waking up
Under their blankets of snow.
I am so happy!

The next incident happened in grade school. Her class had a race and she was coming in second to last. But just before the finish line she stopped and waited for the girl who was last, and joined hands with her as they crossed the finish line together. With this brilliant gesture, she turned an athletic defeat into a moral victory.