talk, pcru

Today was talk preparations (making a handout), pcru hacking, talk, more pcru, and some recipes (up to 820)

handouts

I think I've found the basic formula for making the handouts. I've never really been satisfied with the idea of just using my slides as a handout, because they are fundamentally different things. For starters, handouts show you everything at once, so you can afford to cut out a lot of the repetition, which you mostly put into slides so that people can follow along.

The key points to the Erician Handout (2008-01-10) are:

1. single sheet of paper
2. two-column
3. one section per column?
4. references at the end of each column
   
5. skinny margins
   
6. no title (but headers indicating title/author)

And here is the relevant code.

\documentclass[]{article}

\usepackage{color} % this comes with beamer; so here, we must explicitly specify
\usepackage[margin=1.3cm,columnsep=1cm]{geometry}
\usepackage{fancyhdr}

\pagestyle{fancy}
\setlength\parindent{0cm}
\renewcommand{\headrulewidth}{0pt}

\begin{document}

\lhead{\tiny \sf Surface realisation: ambiguity and determinism}
\rhead{\tiny \sf Eric Kow 2008-01-10}
\twocolumn

Blah blah blah

\end{document}

You may need to play with the margins a bit so that the headers/footers come out right. I guess if I knew more LaTeX, I could make this my own class or something. Using 'article' seems kinda weird :-)

Also, I believe the right approach would be to make the slides, copy the tex into a new file and turn them into a handout. Summarise the slides, and add in some extra value, for example, by putting down URLs which won't be very useful in slides anyway, or extra bib references. Put in the handout the kinds of things which people are likely to want to pore over, like diagrams, trees, numbers.

One thing I'll need to write is a set of sed scripts or whatever for converting beamer presentations into handouts, or maybe just find a way to configure beamer so that it produces a proper handout. Latter bit is doubtful, because I don't think I really want individual frames.

pcru

Last night, it struck me why I was stuck and bored with the pCRU stuff. Something was gnawing at me... basically the problem of distributing a nice example for the README. Basic thing was essential inflexibility, had two corpora that worked different ways. One provided start symbols (the segmented one) and one did not (was systematically TOP). Figured out exactly what I had to do on the way back, and implemented it this morning: make the scripts handle both situations! So now I have both the traditional weather-num example in Examples/SumTime-forecasts directory and the segmented version in Examples/SumTime-segmented. And the README actually makes sense.

Yay! Now stuck on one minor issue, which is that mixed generation doesn't really seem to work. Can select from 1 best, but not from say, 3 best. The original version doesn't appear to like it either (ran some scripts)

the talk

Eh. It was OK, I guess. I don't think I did a very good job answering people's questions. Anyway, I said I wouldn't feel settled until the talk was done. So... I guess it's done. So am I settled yet?

Chatting a little bit with JC. He mentions that there's some Japanese work approximating an HPSG grammar with a CFG. Seems directly relevant to what we're doing here. Should go look.

Paradoxically, the chatting which was very nice and pleasant, made me a bit homesick. I mean. I'm not exactly emotionally developed enough to latch on to that feeling, or for it to be a big deal, but it was interesting. Or maybe it was the fact of giving what may well be my last Nancy talk... dunno...

good luck!

Good luck finishing up your ACL submissions. I'm not putting one together. Maybe for something else in the next couple of months :-/. For those of us in the UK, that's 04:59 this morning, or 05:59 (even better!) for those on the continent.

Boston Cookies

1 cup butter
1/2 teaspoon salt
1&1/2 cups sugar
1 teaspoon cinnamon
3 eggs
1 cup chopped nut meat, hickory or English walnut
1 teaspoon soda
1&1/2 tablespoons hot water
1/2 cup currants
3&1/4 cups flour
1/2 cup raisins, seeded and chopped

Cream the butter, add sugar gradually, and eggs well beaten. Add soda
dissolved in hot water, and one-half the flour mixed and sifted with salt
and cinnamon; then add nut meat, fruit, and remaining flour. Drop by
spoonfuls one inch apart on a buttered sheet, and bake in a moderate oven.