New-player transcript collection
ems...@mindspring.com
idea that's been rattling around in the back of my head for a while --
namely, that it might be useful to assemble a collection of
transcripts of IF being played by new players. We've had a number of
conversations of the years about how players of different
ages/backgrounds/experience levels attempt to communicate with IF, but
an actual data set would be especially useful. The phpZork logs I
looked at a couple of years ago for this sort of information have
vanished, and they were not the most convenient thing in the world to
use anyway.
Teachers presenting IF to classes would be in an especially good
position to collect this kind of information, if it didn't prove too
cumbersome or intrusive to do so. (Taking it a step further, I
suppose, would be to assign students to collect transcripts from their
roommates and friends trying IF, if the teacher is feeling especially
enthusiastic.)
I'd be willing to undertake the assembly and hosting of such a
library, so if this sounds interesting to you and you have a
transcript or two lying about, feel free to send 'em to me. Noting
the age of the player would be useful; names and other identifying
info obviously can be left off.
danie...@hotmail.com
interpreter which logs transcripts, throw in a few of the old standard
IF works, and spam the URL all over the place. I'd think anyone who
actually signed up in a class on IF might already have some experience.
After all, working with new players is like quantum mechanics: you
can't measure someone's first reaction without destroying their
capacity to give you another.
Alternatively, you could set up a realtime system in which someone gave
a command, and then someone else on the other end translated it into
standard IF and the response came back to both while the exchange was
logged. This would give the appearance of playing a laggy IF game with
a world-class parser, I guess.
ems...@mindspring.com
> Maybe we should have a java- (or flash- or whatever-) based zcode
> interpreter which logs transcripts, throw in a few of the old standard
> IF works, and spam the URL all over the place. I'd think anyone who
> actually signed up in a class on IF might already have some experience.
Maybe. If anyone wants to set that up, more power to 'em...
> Alternatively, you could set up a realtime system in which someone gave
> a command, and then someone else on the other end translated it into
> standard IF and the response came back to both while the exchange was
> logged. This would give the appearance of playing a laggy IF game with
> a world-class parser, I guess.
I've occasionally wondered what would happen if you hooked a webbased
chatbot into IF -- the chatbot would take incoming patterns, translate
them into IFese (so that, eg, "I want to kick the head repeatedly!"
would be stripped down to KICK HEAD), send them to a Zplet somewhere,
and then pass back the results -- and, of course, retain a log of all
this, as chatbots often do. It would perhaps be irritatingly slow,
but it might over time develop into a pretty good pre-processor for
English-to-IFese. Then again, maybe there are difficulties here I
haven't thought of.
I currently don't have the expertise or the space to do it, though, so
this is one of those useless ideas that just rattles around in my head
making noise.
danie...@hotmail.com
work of IF directly. I could code some of this up if I ever get some
time off of school; but it'd be really easy to add logging to a
preexisting java zcode interpreter, so maybe just email whoever made
that and ask them to set that up
ems...@mindspring.com
> Surely the chatbot's parser could just be extracted and put into the
> work of IF directly.
Maybe. There are a couple of reasons I didn't put it that way,
though:
1. One of the advantages of web-based chatbots like ALICE is that you
can check the logs, add new material, and keep running without having
to rebuild anything.
2. That sort of chatbot requires a very large file of template data,
which might be slow and unwieldy to incorporate into the gamefile
proper.
3. I was imagining winding up with a front end that could interpret
instructions, not necessarily for any imaginable IF game, but for the
majority of those with semi-standard command sets.
danie...@hotmail.com
customized for each game; note the heavily inconsistent implementation
of "turn" verbs -- do you "turn the nut on the bolt," "turn the nut
with the wrench," "turn the wrench around the bolt," etc.? Probably
the best sort of place for this thing is integrated into the zcode
interpreter, which a separate parser file to correspond to each z5
file. (This simplifies a whole bunch of interactions between various
subsystems and makes porting and updating a lot easier).
Then we could publish the metafile update and authors could release a
file along with their games that would allow them (the games, not the
authors) to accept more verbose input. In fact, it would even be
possible to patch Inform to output some files like this, describing the
various nouns and verbs in the game. If this was done in java we could
insert the logging provisions as well, and have one nice big chunk that
could be delivered as an applet or application to anywhere and send the
game transcripts back for analysis.