yay! bug finally squashed: how many words is "gonna"?

[trochee]

My labmate S. has been after me to fix a bug for him for oh, six weeks.

i finally had a little time to tackle his bug (after a visit to the Ling department to turn in my fellowship applications) and was able to figure out what was wrong -- at last!

Here's what happened:

The code S. was using was code I developed to synchronize ten or so different annotations of the same speech corpus: one describes words, one describes syntactic structure, another annotates disfluencies and speech acts, another annotates prosody, etcetera etcetera ad nauseam (and I do mean ad nauseam).

Here's one of the gotchas:

syntactic annotation wants to treat a sequence like "I'll do it" as four words:

(S
  (NP I)
  (VP
    (Aux 'll)
    (VP (V do) (NP it)))

However, phonological annotation (speech recognition transcripts etcetera) would really like to treat it as three: I'll + do + it.

Both sides have their logic. To align these files, i had to make a hand-built list of syntax "words" that should be mooshed together with the previous word. Here's my list: n't 's 'S -s 'd 're 'll 'm 've '

But I'd missed one: na, as in "gonna". That meant that when I was trying to align the phonological words, I had an extra syntactic word (either "gon" or "na" depending on the mood of the dynamic-programming alignment when it reached that point) that had no corresponding phonological word. And there was much lamenting in the code.

Hooray that it's fixed now! Hoping that S. doesn't find any more of these.

Comments

[faithhopetricks.livejournal.com]

ICON LOVE

....ahem.

[trochee.livejournal.com]

credit is due to chr0me_kitten, who made it for me.

[boobirdsfly.livejournal.com]

No i think she was trying to offer you the riddles in the dark one !

am I right ?

Poetry

[boobirdsfly.livejournal.com]

And there was much lamenting in the code.

I love this sentence.
This sentence will be in my play which involves a crazy computer guy one day....

quizzical

[trochee.livejournal.com]

i have no idea what you're saying. :-)

Re: Poetry

[trochee.livejournal.com]

it was a distant allusion to "and there was much rejoicing", but I think insane computer guy works too.

and it's certainly believable that insane-computer-guy makes references to The Holy Grail.

Re: quizzical

[boobirdsfly.livejournal.com]

I am saying that I think redshoes was trying to gift you the icon she used !
But i could be wrong.

Re: Poetry

[boobirdsfly.livejournal.com]

I see the allusion. but that doesn't even matter...
the sentence itself is gold.

Re: quizzical -> clarity

[trochee.livejournal.com]

ah, I read it as "she was trying to offer you [the riddles in [the dark one]]", rather than "she was trying to offer you the [riddles in the dark] one".

No longer confused. :-) Not certain of the_red_shoes' intent either way.

Re: quizzical -> clarity

[boobirdsfly.livejournal.com]

Will she pop up and clear it all up for us ?
stay tuned...

[faithhopetricks.livejournal.com]

No, actually I was expressing helpless fangeeky love for Bobby Detective Goren Vincent D'Onofrio.

Re: quizzical -> clarity

[faithhopetricks.livejournal.com]

I have RETURNED!

I was lusting after Vincent D'Onofrio.

[boobirdsfly.livejournal.com]

I stand corrected.

Re: quizzical -> clarity

[boobirdsfly.livejournal.com]

Yes, aren't we all ?
*sigh*

[faithhopetricks.livejournal.com]

....wibbles over Vincent D'Onofrio some more

Re: quizzical -> clarity

[faithhopetricks.livejournal.com]

....he's just so....so.....GEEKILY HOT. HOTLY GEEKY.

Re: quizzical -> clarity

[boobirdsfly.livejournal.com]

*swoon*

Re: quizzical -> clarity

[faithhopetricks.livejournal.com]

((scoops pillow under your head before it hits the floor))

Re: Poetry

[faithhopetricks.livejournal.com]

I think my husband regularly comments laments in his code at work. Heh.

Re: quizzical -> clarity

[boobirdsfly.livejournal.com]

Ha ha, i got a really awesome visual for this one.

[evan.livejournal.com]

Write an automated test: a function that takes one of these parsed sentences and exports how many words it sees, then verify that all systems have the same outputs for sentences you'd expect to be the same.