infryq

Yesterday I added some better error handling to the goodreads-clp app, had
a late lunch with V to catch up (she’s applying to PhD programs which is
good), finished resolving tests and committed code for both the parser and
the server for phase one of the new types system. P had an interview with
ISI which went well, and he’s starting to outrun his teaching spoons so
taking more walks through the park to help balance everything out. Dinner,
crossword, and I started reading No Man of Woman Born by Ana Mardoll, which
is much better than anything else I was trying to read before. Kinda want
each story to be a book in its own right, though. What happens next???

Yesterday I added a thing to the goodreads-clp app to look up the number of
copies and waitlist for each search result, and started a thing to
cross-reference titles against a complete series listing so I can display
results in order by series.

After lunch I called legume to confirm our reservation and dietary
restrictions for Monday, and the woman on the phone was very nice and not
scary at all. Take that, anxietybrain!

I explored a couple of options in the sempre type inference system to try
and prevent ContextFn from returning whole semanticframes instead of just
entity references like I want it to do, and found several ways that don’t
work.

I went to a group meeting to help F decide how to structure an MTurk study
she’s working on along with a PhD student from HCII. Very exciting. Also,
pilot studies are good, always do a pilot study.

After work I went on a small adventure to return library books. I’d planned
to also get a flu shot, but found I was underdressed for the weather, so I
just hopped a bus instead. P gave me a lift from the geagle.

Dinner, crossword, bed. It’s way too dark at not even 7:30.
adventures

Yesterday I fixed a minor pagination issue with my goodreads/clp app, and
let it run all day. Disappointing results (like, *nine* series matches out
of 500 books, and none of them the first in the series), but way better
than doing all those searches by hand.

At work I finished the first round of parser revisions with largely great
success: reduced the size of the lexicon file by 1/3 and reduced the number
of silly parse candidates by a lot. The remaining issue is that now there’s
no way for the server to tell the parser about coreferent vocabulary for
relational values like “her email address”. There is existing functionality
in SEMPRE for contexts, which would let us match “her” correctly without
the assistance of the server, but we apparently removed all that machinery
from our fork? I’m checking with a past group member to see if there’s a
non obvious reason for that, otherwise I’m gonna pull it all back in.

Lunch was better, and I ate a pile of sunflower seeds all afternoon.

After work, picked two beans in the garden, dinner, trash, crossword, bed.
Some spooky shoulder pain that seems to have faded overnight. Somebody
dropped a large metal object into a dumpster about 2am and proceeded to
make backing-up noises for twenty minutes. Not cool. Dreamt of helping my
grandmother organize multiple sets of linens. Also dreamt of trying to hide
from a grizzly bear in a scrambled version of my junior high school.