grr...

[trochee]

a week's work, including some non-significant but promising intermediate results, and the final results?

... nothing. Exactly the same BLEU score as when I started.

To quote the philosopher Cartman:
Screw you guys. I'm going home.

Comments

[boobirdsfly.livejournal.com]

sorry j~. that is the nature of the work I guess. and i guess it makes results very special when you do get them. :)

[http://users.livejournal.com/_dkg_/]

hey man-- i know it's discouraging, but you have to remember that the failure of a hypothesis is not the failure of the experiment.

Keep your scientist's hat on, not your engineer's hat, and consider what kind of knowledge can be gained from these results. i'm sure you don't need me to tell you these things, but sometimes it's good to just have them restated, especially when the "must publish soon" urge is present. Why were the preliminary results encouraging when the final results weren't so much? Can you pick apart any differences there?

[trochee.livejournal.com]

right on, man.

this morning there are some ideas I have about what the differences were generated from (my version does much better on unigrams, bigrams and trigrams, but the naive version does better(!) on 4-grams and up, and the improvements even out on BLEU score).

I'll be working on it to see what I can tease out. and you're a good angel-on-my-shoulder, _dkg_. In an engineering lab, it's easy to lose sight of the fact that I'm trying to get a science degree. Thanks!

[boobirdsfly.livejournal.com]

Good Brother. Good.

[beckyb.livejournal.com]

Man. Welcome to the field. The process, the quest, the search has to be part of the reward since the results themselves are so often not.