Public Q&A transcript

October 9, 2026 dissertation defense

← Back to defense materials

Timestamps refer to the original defense recording.

The unfinished fragment after the keyboard-layout question is omitted at the presenter’s request.

Download this transcript (.txt)

[43:25] Yu Liang

Okay. Now is the question to the, I think. Audience. Yeah. From the audience.

[43:36] Yu Liang

Any questions.

[43:39] Dan Mailman

Any questions. Everybody's asleep.

[43:42] Yu Liang

No, no, no. It's exciting. Interesting. I learned a lot. Okay.

[43:47] Audience member (Zoom label: Schwab iPad)

I've got a question. And this may be completely out of scope. And I may be uninformed too. So just bear with me, but. And I've, I've heard that the court. Recorders. Who are typing out the transcripts of the courts that they have. Types of input devices. And. That are not strictly keyboards and maybe they have phrases on them. I'm not sure.

[44:10] Dan Mailman

I haven't looked. To the best of my knowledge. Those are syllable based and they, I think that's a good avenue for future. Study.

[44:17] Chu-Ren Huang

Yeah.

[44:18] Dan Mailman

I think it would be for Mandarin as well. That if you, instead of using strict pinyin, you could, you would. And people have done this divide. The, the syllable up into. Initial and final. We might be able to even squeeze one more. Save gesture into that. but I went with these as sort of for prototypes.

[44:39] Audience member (Zoom label: Schwab iPad)

Right. Yeah, I understand you have to kind of hone it in to exactly what you can work on. But yeah, thank you for that.

[44:46] Dan Mailman

Sure, thank you.

[44:52] Yu Liang

Some other questions?

[44:54] Dan Mailman

Looks like Sarah has one.

[44:57] Sarah Tsiang

Thank you, Dan. That was great.

[44:59] Dan Mailman

Wow.

[45:00] Sarah Tsiang

You kept mentioning Mandarin Chinese. How much would it take to adjust what you've already done to other varieties of Chinese?

[45:09] Dan Mailman

I think, well, okay, so I have this theory that I really can't prove that if you do a really good phonemic analysis of a language, you can get it down to six gestures no matter what the language is. But so I'm going to go out on the limb and say, I can get it down to six, but it would take me a month or two to come up with a system that does it. Have I answered your question?

[45:34] Sarah Tsiang

Yes.

[45:34] Dan Mailman

Okay, and I can't prove that. It's just this intuition from the work that I've already done.

[45:42] Chu-Ren Huang

Dan, I think Sarah's question is basically, can you apply it to Cantonese, Taiwanese, Hakka and stuff, right? Yes, I think so.

[45:53] Dan Mailman

The normal systems that scholars and missionaries have developed for Romanizing those tend to give more letters than I think are actually necessary, even given the actual pronunciation. Of those languages, okay? And I'm only passingly familiar with Shanghainese or Cantonese. So I can't say that's true for all of the dialects, but it seems to me from the very little I've read of the other dialects, the same thing would be true, that you would come up with a system that gives recognizable cues leading up to the characters. And those cues would be recognizable

[46:35] Dan Mailman

by a speaker of Cantonese or a speaker of Hakka, or a speaker of, let's say, Sichuan dialect, I think. Again, no proof, but I have this intuition we can do it. We can engineer such a system that will result in six cues maximum for 99% of the characters.

[47:06] Yu Liang

Any other questions?

[47:08] Audience member (Zoom label: Schwab iPad)

Yeah, I was going to, I'm not sure if I can even quite put this into a question, but just something that kind of came to mind. And, you know, some of the limit to typing speed these days is also the layout of the QWERTY keyboard. And I'm sure you're aware there are other layouts that have been optimized for common keys to be in a different position that's easier to type to. So again, I'm not sure I actually have a question on that. It's just, you know, the ultimate, it seems to me the ultimate goal here is to make people input to the computer quicker. So maybe a question would be, how do your methods compare to like just changing the layout of the keyboard?

[47:47] Dan Mailman

Well, in fact, that's the reason I chose the letters as cues. So that's to separate the problem. So people can get faster or slower depending on the keyboard layout, perhaps, right? But an A is still an A and you know how to get to the A on the keyboard, right? So whether or not it's, you know, one layout or another, you can use the same cues for the same, targets for the same outputs. Right. Does that make sense? So separate, you know, it's at the engine, you know, divide and conquer. If you want to speed up your typing using a different layout, you can still use this software, right? Because an A is an A, but it doesn't matter where it is on the keyboard.

[48:32] Audience member (Zoom label: Schwab iPad)

Right. Yeah, that makes perfect sense. Yeah. Thank you. Surely.

[48:41] Yu Liang

More questions?

[48:46] Yingfeng Wang

So can committee members ask questions?

[48:50] Yu Liang

Later. It is the closed-door discussion. Yeah. You can keep your question for the closed door. Yeah. Discussion. Okay. Yeah. If no more question from audience. Yeah. Please just, yeah. Let's give a big hand to Daniel's great presentation. Thank you.

[49:08] Dan Mailman

Thank you. Yeah.

[49:09] Yu Liang

And committee member, please stay.

[49:14] Yingfeng Wang

Daniel, should you stop recording?

[49:17] Dan Mailman

Yes. Good for you. Thank you. I'm going to do that right now.

[49:21] Yu Liang

Okay. Thanks, everyone. Thank you. Yeah. Let's just wait a minute.