>Transmogrify parser.
Ian Millington
odd line of Inform, and it occurred to me that I was thinking in
interactivefictionese.
> Close interpreter.
> Open file with notepad.
> Compile.
What do you want to compile?
> My game.
It isn't here, you may have to compile it first.
Which got me thinking: are interactive fiction authors (of whom I am not
one) prisoners of a rather nasty Sapir-Worf fueled self-catalysing reaction.
Using verb-noun parsers to build games that are destined to inspire parsers
that get better at parsing verb-noun phrases.
IF systems combine set-piece text in interesting ways depending on user
choices.
Do choices have to be verb-noun sentences, and does the alternative have to
be hyper-fiction.
My vision goes cloudy at this point and smoke begins to emerge from my ears.
Back to Trinity then: bloody good game...
Ian.
Andrew Plotkin
> Which got me thinking: are interactive fiction authors (of whom I am not
> one) prisoners of a rather nasty Sapir-Worf fueled self-catalysing reaction.
> Using verb-noun parsers to build games that are destined to inspire parsers
> that get better at parsing verb-noun phrases.
Well, sure. We're getting better at what we do. Why is this
controversial?
> IF systems combine set-piece text in interesting ways depending on user
> choices.
Only if you define "set-piece" very loosely -- the pieces can be
paragraphs, sentences, phrases, or words; whatever level of detail is
appropriate for that part of the game world.
--Z
"And Aholibamah bare Jeush, and Jaalam, and Korah: these were the borogoves..."
*
* Make your vote count. Get your vote counted.
Jeffrey F Pack
> Here, Ian Millington <0247...@blueyonder.co.uk> wrote:
>
> > Which got me thinking: are interactive fiction authors (of whom I
> > am not one) prisoners of a rather nasty Sapir-Worf fueled
> > self-catalysing reaction. Using verb-noun parsers to build games
> > that are destined to inspire parsers that get better at parsing
> > verb-noun phrases.
> Well, sure. We're getting better at what we do. Why is this
> controversial?
Well, I think it's the idea that parsers have abandoned English in
favor of Adventurese. Not that I believe an otherwise "traditional"
game which required alternate structures (no idea what this would look
like; perhaps a type-the-response conversation system that tried to
gauge tone as well as subject?) would be well-received; players are
too used to VERB NOUN.
Jeff
Andrew Plotkin
> Andrew Plotkin <erky...@eblong.com> writes:
>> Here, Ian Millington <0247...@blueyonder.co.uk> wrote:
>>
>> > Which got me thinking: are interactive fiction authors (of whom I
>> > am not one) prisoners of a rather nasty Sapir-Worf fueled
>> > self-catalysing reaction. Using verb-noun parsers to build games
>> > that are destined to inspire parsers that get better at parsing
>> > verb-noun phrases.
>> Well, sure. We're getting better at what we do. Why is this
>> controversial?
> Well, I think it's the idea that parsers have abandoned English in
> favor of Adventurese.
You have to own something before you can abandon it... text game
parsers have *always* been in Adventurese. I don't believe Colossal
Cave had any command range that isn't covered, pretty much, in the
standard modern parsers.
(Exception: I'm told that the original Adventure code supported typing
a room name (no verb) to mean "go to ROOM", which would only work if
you were near ROOM. That convention has been dropped. However, that is
even less like English than modern Adventurese.)
> Not that I believe an otherwise "traditional" game which required
> alternate structures (no idea what this would look like; perhaps a
> type-the-response conversation system that tried to gauge tone as
> well as subject?) would be well-received; players are too used to
> VERB NOUN.
(And VERB NOUN PRONOUN NOUN, and so on.)
My _Space Under the Window_ was a random stab at a different command
system.
It was received fine, but it hasn't been followed up on much. I'd say
that's because it's much more limited than Adventurese. The player can
guess what words the game might accept, but he isn't really
manipulating the game world in an intentful way. It's just "try a
word, see what wacky text results".
The point of adventure games -- as we know them, and this hasn't
changed much over time -- is to be presented with a fictional world,
to decide what to do in that world, and to be able to effect that
decision in a natural way. That is, the command interface must
*respond* to your decisions, rather than shaping your decisions.
Maybe that's fundamentally impossible, in a Sapir-Whorf sort of way --
but then what? There can be no response to that line of inquiry. "Give
up, adventure gaming is impossible without true AI." Whatever.
The Adventurese system works because, for most of what you want to do,
it's easy to compose a command that does it.
In some areas, notably dialogue, it *doesn't* work. And yes, I think
this tends to push authors away from writing games with dialogue.
(Certainly has for me.) But I don't see it squeezing out other forms
of dialogue interaction either. We've had *plenty* of games with
alternative dialogue interfaces. The problem so far is that none of
them are very satifactory either.
Ian Millington
> >> controversial?
I'm sorry - I don't have an agenda. I was musing is all. I love IF as it is
now, and am not trying to organise the revolution.
> That is, the command interface must
> *respond* to your decisions, rather than shaping your decisions.
But doesn't it shape it by not responding to most of your decisions
sensibly? (And that's not fixable - because it cannot conceivably cope with
the range of human intention).
> The Adventurese system works because, for most of what you want to do,
> it's easy to compose a command that does it.
I fundamentally disagree. This should read: for most of what you are allowed
to do, you might at some point want to try it. You can't do what you want,
and that structure is what makes it interesting and a tractible playing
experience.
> There can be no response to that line of inquiry. "Give
> up, adventure gaming is impossible without true AI." Whatever.
(Computer) game AI is my day job (www.mindlathe.com). We're not going to get
to 'true AI' anytime soon, and IF simply doesn't have the budget to have
anywhere near cutting edge AI (and may be better for it, in many cases).
But I don't hink it is neccesary. And this was what I was getting at.
The Space Under The Window wouldn't be improved if I could have my every
intention realised: far, far from it. The restrictions made it what it is.
What I do wonder is if there are a _different_ set of restricted grammars we
could impose on our self (for the sake of implementability), that would
provide an alternative reality. Not to replace, of course.
Off the top of my head: models on the basis of PERSON is EMOTION [with
SUBJECT], with a stageplay metaphor (rather than a VERB NOUN model with a
caving metaphor).
> S [-- an abbreviation for SETS]
The following sets are availabe: The lounge, The dining room, Joe's Study,
The Garden
> NEW SCENE IN THE LOUNGE
When?
> EVENING THE NEXT DAY
The dusty evening light ran in at a sheer angle through the lounge window;
peppering the plain carpet with shifting patterns .... <snip>
> JOE ENTERS AND SARAH IS PISSED WITH JOE
... <snip>
And it doesn't have to be open ended. The story gets revealled, you change
the states of the characters and their interrelations (as opposed to the
states of objects and their interrelations), and try to achieve a goal state
revealed as such by the story.
And the text is going to repeat (Joe always says that when he is pissed) -
but how many times do you read the same text in an adventure game.
I don't think you can get better than the best games around (of which Zarf's
are among my favourites), and as long as you wan the interaction to be
textual (which I do, wholeheartedly) you need parsers, and good ones. And
any change is going to mean worse implementations (a previous post argued
that modern parsers can do anything Collosal Cave could do - true, and,
importantly, they can do a lot more - hoorah!).
I don't wonder if there is a better way, I wonder if there is a different
way.
Ian
Mike Roberts
> Which got me thinking: are interactive fiction authors prisoners
> of a rather nasty Sapir-Worf fueled self-catalysing reaction.
> Using verb-noun parsers to build games that are destined to
> inspire parsers that get better at parsing verb-noun phrases.
Even if you thought the Sapir-Worf hypothesis held any water (which I don't,
but that's a separate topic), I'm not sure how you could really apply it
here. The point of Sapir-Worf is that language determines world view
because one can only see the world in the terms in which one's native
language puts things. In the case of IF, perhaps you could make the
argument that a particular game is trapped in a Sapir-Worf box for its own
VERB NOUN PREP NOUN (etc.) parser, whatever it means for a game to have a
point of view, but I can't see the same argument extending to authors or
even players.
Authors write games in a combination of Inform or TADS or some other
programming language, and English or some other natural language; their only
VERB NOUN etc. interaction is programming parsers for such syntax. I doubt
that even the most faithful Sapir-Worf adherents would argue that a linguist
studying a foreign language's structure would become trapped in the world
view of the subject language by virtue of that study, or by virtue of
writing a computer program that parses the subject language. After all, the
linguist/programmer still has her native language to fall back upon as the
bedrock reference frame. It seems more appropriate to Sapir-Worf to argue
that the author's world view is molded by TADS or Hugo than by VERB NOUN
etc. (I can only assume the Sapir-Worf argument has been made many times
over the years about any given programming language - that it locks you into
a way of thinking that you can't escape as long as you're using that
language. Which is at odds with one of my favorite nerd adages: you can
write a Fortran program in any language.)
There is one small way in which I think the Sapir-Worf analysis might not be
completely bogus, but only as a rough analogy. Authors of IF games
sometimes find that they've conceived a puzzle whose solution doesn't map
well to the usual command syntax, and they're forced to modify the puzzle to
make it fit better. But this is a conscious process, undertaken with the
clear intention of working within a set of restrictions and conventions that
have developed over the years. If Sapir-Worf truly applied here, the author
would simply not be capable of conceiving those ill-fitting situations in
the first place.
As for players, they see the game's world in English or whatever, and think
of things they want to do in their own internal mental model (which would
not be a VERB NOUN etc. model if we were to stipulate Sapir-Worf - it would
come from their native language). A player only resorts to VERB NOUN etc.
as the last mental step, translating his intentions into that syntax in
order to operate the human-computer interface. As evidence that this is so,
I submit the guess-the-verb and guess-the-phrasing puzzle: if the player
didn't have some internal mental model of her intentions in terms other than
the command syntax, how could she have trouble guessing the syntax?
Later:
> What I do wonder is if there are a _different_ set of
> restricted grammars we could impose on our self (for
> the sake of implementability), that would provide an
> alternative reality.
You mean like the point-and-click grammar of Myst, or the joystick language
of Flight Simulator, or the mouse-gesture command language of Warcraft III?
My point being that it's not as though we IF authors and players have never
heard of any other way of playing games; many (if not most) of us have seen
all sorts of other game user interfaces. I think the answer to your
wondering is pretty clearly that, yes, there are all sorts of other ways of
handling input; but I tend to think that the alternate reality comes from
the design of the game, and that the user interface follows from that,
rather than vice versa.
--Mike
mjr underscore at hotmail dot com
Jeffrey F Pack
> Here, Jeffrey F Pack <jf...@konichiwa.cc.columbia.edu> wrote:
> > Andrew Plotkin <erky...@eblong.com> writes:
> >> Well, sure. We're getting better at what we do. Why is this
> >> controversial?
>
> > Well, I think it's the idea that parsers have abandoned English in
> > favor of Adventurese.
> You have to own something before you can abandon it... text game
> parsers have *always* been in Adventurese. I don't believe Colossal
> Cave had any command range that isn't covered, pretty much, in the
> standard modern parsers.
Perhaps "abandon" was the wrong verb. What I meant was that, as I
understand it, the VERB NOUN convention (Old Adventurese?) developed
early, and for a while parsers got better at interpreting grammar that
looked more like English, until we've arrived at Modern Adventurese.
No, it's not a perfectly linear timeline, and I really don't remember
the state of the original Colossal Cave parser
> (Exception: I'm told that the original Adventure code supported typing
> a room name (no verb) to mean "go to ROOM", which would only work if
> you were near ROOM. That convention has been dropped. However, that is
> even less like English than modern Adventurese.)
>
> > Not that I believe an otherwise "traditional" game which required
> > alternate structures (no idea what this would look like; perhaps a
> > type-the-response conversation system that tried to gauge tone as
> > well as subject?) would be well-received; players are too used to
> > VERB NOUN.
>
> (And VERB NOUN PRONOUN NOUN, and so on.)
PREPOSITION, surely.
> My _Space Under the Window_ was a random stab at a different command
> system.
And was one of the main reasons I put that "traditional" in there.
Could that system be used with a more conventional room-and-object
setup to "zoom" in and out of descriptions more fluidly than the
LOOK/EXAMINE/SEARCH hierarchy? I don't know.
> It was received fine, but it hasn't been followed up on much. I'd say
> that's because it's much more limited than Adventurese. The player can
> guess what words the game might accept, but he isn't really
> manipulating the game world in an intentful way. It's just "try a
> word, see what wacky text results".
> The point of adventure games -- as we know them, and this hasn't
> changed much over time -- is to be presented with a fictional world,
> to decide what to do in that world, and to be able to effect that
> decision in a natural way. That is, the command interface must
> *respond* to your decisions, rather than shaping your decisions.
Well, I just finished a relatively large puzzle game (Heroine's
Mantle), where I found myself thinking in terms of what VERBS I could
use on all the available NOUNS. Basically, when I got stuck I'd go
into IF-Playing Robot mode, in which I EXAMINE, SEARCH, PUSH, PULL,
TURN, LOOK UNDER/BEHIND every noun I can find in the game text. Then
I've got a collection of items which can only be used in relatively
basic ways (though the combinations can allow for complex behavior if
the author so desires).
> Maybe that's fundamentally impossible, in a Sapir-Whorf sort of way --
> but then what? There can be no response to that line of inquiry. "Give
> up, adventure gaming is impossible without true AI." Whatever.
It's not a "give up on adventure gaming" so much as "should we be
trying to do more in this area, or is this enough?" I honestly don't
know what else could be done. Guess I'm entrenched.
> The Adventurese system works because, for most of what you want to do,
> it's easy to compose a command that does it.
Yes, as evidenced by the fact that I can't think of any
counterexamples off the top of my head.
> In some areas, notably dialogue, it *doesn't* work. And yes, I think
> this tends to push authors away from writing games with dialogue.
> (Certainly has for me.) But I don't see it squeezing out other forms
> of dialogue interaction either. We've had *plenty* of games with
> alternative dialogue interfaces. The problem so far is that none of
> them are very satifactory either.
Dialogue seems to be a completely different beast, or perhaps it only
seems so because what works so well for a room/object setup doesn't
work there. But that's a whole nother topic.
Jeff
Ian Millington
it as such (along with the metaphor of an autocatalytic reaction).
> A player only resorts to VERB NOUN etc.
> as the last mental step,
The player might, but what about the author? My point is the author limits
the availability of story elements to those that can be manipulated through
imperative clauses (except for those few exceptions, which don't).
You have assumed throughout your post that IF can only be played by issuing
imperative commands for an in-game avatar to carry out. If that is the case
then you are right on all points. But I could cheekily say that you only
think so because the constraints of the imperative parser have limited your
ability to think about interactive fiction in any other way (hence S-W).
The best way of describing imperatives is using a VERB NOUN sentence
structure. This is as true as it is tautological.
> Later:
> > What I do wonder is if there are a _different_ set of
> > restricted grammars
> You mean like the point-and-click grammar of Myst, ...
No, no, no. I said in the same post that I was refering to textual input.
[Although it is interesting that there is fairly widespread recognition that
CYOA interaction is valid for IF (if not for a full game, at least for some
elements of it such as dialog).]
With respect, don't assume an unfamiliar poster is naive, or that a critical
post is antipathetic.
I meant this: rather than a grammar whose dominant rule is of the form:
S -> V NP [PP]
etc..
you could posit all kinds of dominant rules of different forms:
S -> NAME is ADJ [my previous post]
or
S-> Q V NP [Q = Question indicator]
or many, many others.
And I wonder if the IF that would naturally map onto those would be of a
different quality. And I wonder whether that change in character would be to
more 'literary' IF, more social stories, or more shred responsibility for
story between author and player. Or if there is a way to produce a grammar
to meet a specific project's needs. Or whether the whole thing would end up
unplayable as the player battles with understanding the rules of the parser
alongside the rules of the game.
> alternate reality comes from
> the design of the game,
Absolutely: and it is the authors desire and commitment to this that makes
IF games are far less homogeneous, IMO, than the commercial video games I
work on.
> and that the user interface follows from that,
> rather than vice versa.
I disagree. Why are the number of interaction modes so few? The user
interface flows from the parser author. For good reason - very few of us can
spare the kind of hard slog you've put into TADS, and fewer of us have the
skills to do it in any case. And it is exactly that reason why we'll stick
with VERB NOUN grammars; because to make just that work is frighteningly
difficult, and folks like you have done it so we don't have to. Plotkin's
SUTW does it differently, but uses a far simpler grammar than most [S -> N],
so he could get away with implementing it and writing the game, and having a
life.
Anyway, I don't think I am advocating anything. I would have to be unhappy
with something for that. And even if I were, I am not writing IF and
providing free entertainment, so I would have no right to be!
Ian.
Mike Roberts
> > A player only resorts to VERB NOUN etc.
> > as the last mental step,
>
> The player might, but what about the author? My point is
> the author limits the availability of story elements to those
Sure - I pointed out that this does happen in my previous note. But as I
said there, this limitation is *intentional* on the part of the author, not
due to any assumptions that the author can't rise above. That's why I don't
think Sapir-Worf holds, even as a metaphor; consciously chosen limitations
are of a different character than unconscious ones. This is what I'm
getting at: I don't think IF authors are stuck in a blindered world view
where they'd be writing a different kind of game if only they could see out
of their little box. Many of us have indeed peeked out of the box to see
what else is out there, but we come back to this form anyway, because it's
what we like.
> You have assumed throughout your post that IF can only be
> played by issuing imperative commands for an in-game avatar
> to carry out.
I don't see how you inferred that assumption from anything I said. I'm
really pretty liberal by raif standards on what forms of games I consider
IF; my personal definition of IF is open to things without text and without
parsers, so I don't see how I would have implied that imperative commands to
avatars are definitive of IF.
> > > What I do wonder is if there are a _different_ set of
> > > restricted grammars
> > You mean like the point-and-click grammar of Myst, ...
> No, no, no. I said in the same post that I was refering to
> textual input.
>
> With respect, don't assume an unfamiliar poster is naive,
> or that a critical post is antipathetic.
I didn't mean to imply anything of the sort, and I was quite serious in
raising graphical adventures, RPG's, sims, and other types of games as
examples of other grammatical forms of interaction. My point was that you
don't have to wonder in a vacuum of speculation what would happen if someone
were to create a game that didn't have a conventional parser; you can just
look at one of the many other styles of games for some concrete examples.
Okay, so they're not text, but they still seem like useful data points if
you're wondering how a game's method of input is related to its design. I
mean, it seems a bit odd to ask: why aren't there any kinds of games except
this one?, and then not want to even consider a bunch of examples because
they're not a kind of game like this one.
> > and that the user interface follows from that,
> > rather than vice versa.
> I disagree. Why are the number of interaction modes so few?
> The user interface flows from the parser author. For good
> reason - very few of us can spare the kind of hard slog you've
> put into TADS.
Well, I really think this is reversing cause and effect. Authors choose
their media. IF authors are IF authors because they choose to write IF;
it's getting things backwards to think that they write IF because they're IF
authors. They don't choose to write IF the way they do because they can't
conceive of other ways of doing things; they write IF that way because they
like it that way. And no one hunts around on the web to see what
development tools are available before they even decide what kind of game
they want to write; people who find Inform and the rest pretty much always
know exactly what they're looking for.
As for the tools, TADS and Inform and Hugo and the others work the way they
do because there are authors who want tools that work that way. I'll
happily concede that the majority of TADS users probably have to make
changes to their game designs, probably far too many, to accomodate the
limitations and idiosyncracies of TADS. But at the same time, I've made a
lot of changes and additions to TADS over the years because authors asked me
to change the system to accomodate their game ideas.
I certainly can't disagree that writing a parser is non-trivial, so I do
think you have a point that there might be some self-selection that goes on:
someone might have a great idea for a work of text-based IF with an
unconventional parser, but there are no tools to create it, so the work goes
unrealized. Sapir-Worf isn't quite the right metaphor here, though; this is
more like someone with a great novel in his head who can't put it on paper
because he never learned to read and write.
Cedric Knight
> > That is, the command interface must
> > *respond* to your decisions, rather than shaping your decisions.
>
> But doesn't it shape it by not responding to most of your decisions
> sensibly?
Yes, partly, when the player first learns that the standard method of
playing IF is by manipulating objects and any high-level task that can't
be broken down into such manipulation is probably not worth considering.
However, in another sense, that's something authors of individual new
works of IF try to avoid by (a) allowing a reasonable, though still
restricted, range of commands; (b) crafting the output text in such a
way that the player wouldn't want to enter commands the parser doesn't
support, both in room/object descriptions and HELP text.
> (And that's not fixable - because it cannot conceivably cope with
> the range of human intention).
It's probably not practicable for an IF author to make a program to do
so, no. I don't think it's inconceivable, however: I've previously
pointed out with the aid of Roget that the total number of verbal
concepts in natural language is not huge. We obviously still get a
problem with a huge number of noun phrases, though. In an origami
universe:
>FOLD CRANE. FOLD GONDOLA. FOLD KING CHARLES SPANIEL
is too free to be meaningully programmable or playable without a
restriction on the number of origami designs we know. But provided
there is an objective to the story that the player agrees with, this is
not a problem; there's no reason to ask the troll about quantum
chromodynamics. If there is no such motive, we have IF art, which I
don't think is what you're considering.
[snip]
> What I do wonder is if there are a _different_ set of restricted
grammars we
> could impose on our self (for the sake of implementability), that
would
> provide an alternative reality. Not to replace, of course.
>
There are altenative grammars, like SUTWIN, like the world model in the
Geisha section of 'When Help Collides' about which I raised a similar
question here in November.
>
> Off the top of my head: models on the basis of PERSON is EMOTION [with
> SUBJECT], with a stageplay metaphor (rather than a VERB NOUN model
with a
> caving metaphor).
>
> > S [-- an abbreviation for SETS]
> The following sets are availabe: The lounge, The dining room, Joe's
Study,
> The Garden
> > NEW SCENE IN THE LOUNGE
> When?
> > EVENING THE NEXT DAY
> The dusty evening light ran in at a sheer angle through the lounge
window;
> peppering the plain carpet with shifting patterns .... <snip>
> > JOE ENTERS AND SARAH IS PISSED WITH JOE
> ... <snip>
>
> And it doesn't have to be open ended. The story gets revealled, you
change
> the states of the characters and their interrelations (as opposed to
the
> states of objects and their interrelations), and try to achieve a goal
state
> revealed as such by the story.
I like it, and think it could work.
> I don't wonder if there is a better way, I wonder if there is a
different
> way.
I feel this is worth exploring, too, and would like to play more IF like
this if anyone wrote it. However, what Mike Roberts said, and
consideration of WHC, makes me pretty sure an author needs an
appropriate story and design before inventing a specific new grammar.
CK
Ian Millington
> > or that a critical post is antipathetic.
My apologies, I thought you were suggesting I was trying to trash the idea
of text games. I'll try not to assume an antipathetic respondent is
antagonistic!
So are you saying that you think that adventurese came about because it is
the best tractible grammar for the kinds of games that IF authors want to
write?
I can see that, and I respect that, since we have no evidence to the
contrary. But I suspect it might be the case that the kinds of games IF
authors want to write is more influenced by what the grammar can support. Of
course, there can be an infinite number of games written to any non-trivial
grammar, so there is still unlimited scope.
> no one hunts around on the web to see what
> development tools are available before they even decide what kind of game
> they want to write;
That's because there are no alternatives, not in the way I am thinking. Sure
the identifier resolution algorithms are different in Inform and TADS, and
the clarification mechanism is different, and there are minor differences in
the precedence of prepositional clauses. But the differences aren't in the
coarse Grammar that can be supported (s->V NP [PP...]).
After a while playing text adventures you think in VERB NOUN; you mentioned
the fact that guess the verb is a sign that players don't get caught in the
language. I would say that the fact they place guess the VERB is a good
indication of that (as opposed to guess the ADJECTIVE, or guess the AVERBIAL
PHRASE).
So after you've played a few adventure games, and you come to write your
own, you think in VERB NOUN automatically, and demand tools that are very
good at parsing VERB NOUN type sentences.
This, I believe, is why dialog is pretty poor in most IF: you can't express
qualia, or even beliefs, easily in imperatives. So you either end up with:
> ASK BOB ABOUT GRASS
which lacks the revelation of mental phenomena that makes speech
interesting, or you would have to say
> 'JIM TOLD ME YOU HAD SOME GOOD STUFF'
and write your own parser to get it.
> I certainly can't disagree that writing a parser is non-trivial, so I do
> think you have a point that there might be some self-selection that goes
on:
> someone might have a great idea for a work of text-based IF with an
> unconventional parser, but there are no tools to create it, so the work
goes
> unrealized.
I would agree but go one further: the ideas an author has might be swayed by
their knowledge of how it will be implemented. Its rare, for example, to
have IF explore the resolution of a person's inconsistent emotions. But this
is one of the most common themes of literary fiction. Is that because IF
couldn't possibly do it justice, or is it because exploration of it doesn't
map easily onto imperative statements?
> Sapir-Worf isn't quite the right metaphor here, though; ...
Okay, I concede that. It was intended as a throw-away.
Ian.
Mike Roberts
> So are you saying that you think that adventurese came about
> because it is the best tractible grammar for the kinds of games
> that IF authors want to write?
Interesting question. I'm sure part of it is historical; people liked the
ADVENT model, so they built other games using a similar model, which people
liked; and others built still more games based on those, and so on. But it
would be silly to suggest that it's pure inertia and emulation, that the
decisions made by Crowther and Woods were forever locked in for all
subsequent games; obviously, there were other games even in the ADVENT
timeframe that used completely different UI models, some of which took and
some of which didn't. I tend to think there's more of an evolutionary
selection sort of process at work, so the adventurese grammar we have today
came from a basically sound seed idea that's been refined over many
iterations. Seed ideas that didn't have as much appeal simply weren't taken
up in subsequent games and have gone by the wayside. I don't think this
makes adventurese the uniquely ideal solution; it's just one possible good
solution.
> I can see that, and I respect that, since we have no evidence
> to the contrary. But I suspect it might be the case that the
> kinds of games IF authors want to write is more influenced
> by what the grammar can support.
I think I've been failing to articulate the core reason I keep disagreeing
with this hypothesis. It's not the "IF is limited" part that I disagree
with; it's the "because of the grammar" part. The thing is, I think if
there's a limiting factor, it's the world model, not the grammar.
We've been referring to the adventurese grammar as VERB NOUN PREP NOUN - to
which I've been adding "(etc.)" to emphasize that it's not actually limited
to that form, even though that's the most common form; all of the popular
systems make it possible to add essentially arbitrary idiomatic forms as
needed. But even if we were limited to VERB NOUN PREP NOUN, we'd *still* be
able to express vastly more than the world models can actually represent.
In any given game, at any given time, there are all sorts of things -
meaningful things, perfectly sensible and logical things in the context of
the moment - that the human player can express with VERB NOUN PREP NOUN and
that the game won't accept. The game doesn't fail to handle them because
the author thought of them and intentionally chose to disable them, to
impose authorial control over the course of the story; the game fails to
handle them because the world model simply doesn't have any concept of such
things.
I think if we got to the point where the model model could handle every
possible VERB NOUN PREP NOUN you could throw at it, and had further
capabilities that VERB NOUN PREP NOUN couldn't tap, *then* we'd have to
start thinking about how to add more expressive power to the input language.
We're not even close to that point, though. For now, VERB NOUN PREP NOUN
has *more* expressive power than the world models can handle. This is why I
think your energies are misdirected when you talk about adding expressive
power to the input language.
> After a while playing text adventures you think in VERB
> NOUN; you mentioned the fact that guess the verb is a sign
> that players don't get caught in the language. I would say that
> the fact they place guess the VERB is a good indication of that
> (as opposed to guess the ADJECTIVE, or guess the
> AVERBIAL PHRASE).
Well, actually, as I said, there are guess-the-phrasing puzzles as well:
sometimes it's hard to figure out which object should be the direct object
and which should be the indirect object, or what prepositions to use, or
whether there should even be an indirect object, and so on. Even the simple
VERB NOUN PREP NOUN syntax makes a surprising number of combinations
possible. And there are occasional guess-the-noun puzzles; they're not as
frequent, but they do occur.
Even if were *only* guess-the-verb, though, I'd still claim that this is
evidence that the grammar is not the limiting factor. If there exist verbs
that the player can think of and that the game doesn't accept, then the
grammar is fundamentally more expressive than the game can handle, so the
grammar can't be what's holding back the game.
> This, I believe, is why dialog is pretty poor in most IF: you
> can't express qualia, or even beliefs, easily in imperatives.
> So you either end up with:
> > ASK BOB ABOUT GRASS
> which lacks the revelation of mental phenomena that makes
> speech interesting, or you would have to say
> > 'JIM TOLD ME YOU HAD SOME GOOD STUFF'
> and write your own parser to get it.
Again, it's clear to me that the world model is the limiting factor here. I
don't know how familiar you are with the tads/inform/etc world models, but
one thing they don't attempt at all is a model of any mental phenomena of
the characters. I don't think that omission is just laziness on the part of
the library designers or game authors, either; I simply wouldn't know where
to begin to model something like your example.
The thing is, parsing per se isn't incredibly difficult. The techniques for
mechanically dividing strings of characters into lexical units and then
matching strings of lexical units to grammatical rules are fairly well
understood, and linguists have fairly exhaustively analyzed the grammars of
many natural languages. If that was all there was to it, it would be within
reach of modern computers to handle input as complicated as your example.
But the hard part isn't the tokenizing and structural analysis; the hard
part is inferring the meaning of those structures, because they are always
encoded relative to internal mental contexts that casual observation reveals
to be vastly, deeply complex. When we constrain the problem to a very small
and simple "mental context" like the standard IF world model, the problem
becomes tractable, but even then the part of the parser that extracts the
meaning is pretty darn complicated.
> The ideas an author has might be swayed by their
> knowledge of how it will be implemented.
I agree. I think your basic hypothesis - that the existing tools and the
existing style of IF impose limits on the kinds of games you can create - is
sound; but I think your identification of the input grammar as the culprit
is misplaced. I think the world model is the true limiting factor.
Cirk R. Bejnar
> You have assumed throughout your post that IF can only be played by issuing
> imperative commands for an in-game avatar to carry out. If that is the case
> then you are right on all points. But I could cheekily say that you only
> think so because the constraints of the imperative parser have limited your
> ability to think about interactive fiction in any other way (hence S-W).
>
> The best way of describing imperatives is using a VERB NOUN sentence
> structure. This is as true as it is tautological.
While there certainly could be (or are) other types of IF, I will
remain an advocate of the traditional forms because I want to become a
character in the story rather than interact with it in another way.
So that grammars with dominant rules such as:
> S -> NAME is ADJ [my previous post]
>
> or
>
> S-> Q V NP [Q = Question indicator]
>
> or many, many others.
simply don't intrest me. Certainly works that use them would be IF,
but they would not be adventure games, I think, in any sense. Games
that used them would be cut off from the dominant stream of IF in a
very profound way. This is not a bad thing necessarily but it must be
reccognised.
The current state of IF is not due simply to mindless emulation of
Adventure or Inforcom, but rather to concious choice to make games of
a certain type for which the VERB NOUN parser is best suited. The
other mades of interaction with the text suggested create
fundamentally different game dynamics because the elininate the PC.
The player is no longer a character ineracting with a game world and
living out a story but rather a director creating a story out of the
pieces that the world offers. This would be a new and intriging
design paradigm for IF, but because of its deep differences from
adventure gaming I think that it would be as if not more distinct from
current mainstream IF than most graffical games.
Cirk R. Bejnar
Ian Millington
> with this hypothesis. It's not the "IF is limited" part that I disagree
> with; it's the "because of the grammar" part. The thing is, I think if
> there's a limiting factor, it's the world model, not the grammar.
I see your point now, Mike. And its one I hadn't thought of but is a good
one. I might agree with you, I'll think about it.
> For now, VERB NOUN PREP NOUN
> has *more* expressive power than the world models can handle. This is why
I
> think your energies are misdirected when you talk about adding expressive
> power to the input language.
As a small defence, I wasn't thinking of *more* expressive power. There is a
long and futile history of r*if postings asking for better parsing of more
sentence types; I wasn't following that. I was thinking about *different*
biases. A parser, for example, that could handle sentences containing
declarative sentences representing the beliefs of characters; but may not be
able to handle all but the most trivial imperative clause (unless they were
'idiomatic expressions', as you say, and didn't require any complex semantic
reconstruction).
Your point presumably being that this is more about modelling beliefs and
interactions than it is about expressing them in a grammar.
After all, as you've said in other posts recently, the parsing isn't the
hard bit, its working out the meaning.
Ian.
Benjamin Fan
> Even if you thought the Sapir-Worf hypothesis held any water (which I don't,
> but that's a separate topic), I'm not sure how you could really apply it
> here. The point of Sapir-Worf is that language determines world view
> because one can only see the world in the terms in which one's native
> language puts things.
I thought that Sapir-Worf was discounted in recent times and
that the Inuit don't really have 30 words for "snow". Anyway, back
on topic. I think that Adventurese is prevalent because it is the
grammar that makes the most sense for protagonist-driven IF. In
the English language it is either "(I) verb noun" or "(You) verb
noun" to describe an action. If anything, the box might be the
tendency for IF to be protagonist-driven. ("You are the hero of
the story. You make the choices. You affect the outcome.") The
first- or second-person narrative form is the natural result of
this.
It also could be that an author who uses an existing IF system
is boxed in by that system-- someone thinks of a new idea but does
not have the ability or wherewithal to modify the parser. But that
is probably a minor cause. If there were an alternate type of game
that worked well, then I suspect that people would want to write
these types of IF, and the IF systems would have evolved to support
these grammars (like the relatively new "ASK X ABOUT Y"
interrogative-driven IF).
In any case, I don't know Inform, TADS, or any other IF
language or system. I don't play much IF either, so I don't think
that I've been "molded" much by existing IF. But, when I think
about writing IF and when I wrote the JIGSAW parser, the imperative
grammar seemed natural. Does anyone have an idea for a short
proof-of-concept game with an alternate grammar (that pretty much
follows set rules)? If so, I would be willing to try implementing
it.
Ben
--
To get my current email address, concatenate these three strings:
1. "benjamin_fan" 2. "_2002a" 3. "@yahoo.com"
It will look a lot like: xxxxxxxx_yyy_zzzzz @ yahoo . com
Mike Roberts
> I thought that Sapir-Worf was discounted in recent times and
> that the Inuit don't really have 30 words for "snow".
Yes, I think the consensus among psycholinguists and anthropologists has
been for some time that the Sapir-Worf hypothesis is false. It's one of
those ideas that appeals to intuition, though, and I think outside of the
natural sciences it still has its adherents. (And Stephen Pinker does
report in his book The Language Instinct that the 400-words-for-snow thing
is an elaborate urban legend; the actual number, I believe, is 4. Pinker
traces the 400 claim to a series of popular and academic anthropology
articles from (I think) the early 20th century, each of which cited a
previous article and inflated the number it reported. This was another idea
that so appeals to intuition that the story became more plausible every time
the number became bigger.)
[Sorry, veered off topic a bit; I'll stop now.]
Jaap van der Velde
Hello,
I don't want to post a reply to any post in particular, but I've
been reading this thread for the past few posting and one subject
doesn't seem to have come up. Language in an IF environment,
particularly the commands entered by the player, is intended to
direct the actions of the PC. It is a command language, not a means
of communication in the usual sense.
So it seems to me the argument whether or not there is a better
alternative to Adventurese, should focus on command languages.
Is there a better command language than Adventurese?
How is:
> carefully take the queen's purse without getting noticed
Better than:
> use stealth. take purse from queen. act normal
Of course, this completely ignores dialogue in the game, but that's
a different discussion althogether. And a note on the Sapir-Whorf
discussion. Wouldn't such a discussion be more relevant if the
parser only returned feedback in Adventurese as well?
Imagine this:
> get chair
Get chair. See rat. Jump on table. Scream help. Look scared.
> examine rat
See rat. See big ears. See sharp teeth. Hate smell. Have fear.
An idea for a piece of IF, perhaps?
Just my thoughts,
JAAP.
---------------------------------------------------------------------------
"If a `religion' is defined to be a system of ideas that contains
unprovable statements, then Godel taught us that mathematics is not only
a religion, it is the only religion that can prove itself to be one."
-- John Barrow
Jeffrey F Pack
> Of course, this completely ignores dialogue in the game, but that's
> a different discussion althogether. And a note on the Sapir-Whorf
> discussion. Wouldn't such a discussion be more relevant if the
> parser only returned feedback in Adventurese as well?
>
> Imagine this:
>
> > get chair
> Get chair. See rat. Jump on table. Scream help. Look scared.
>
> > examine rat
> See rat. See big ears. See sharp teeth. Hate smell. Have fear.
>
> An idea for a piece of IF, perhaps?
Back when the Comp02 titles were announced, I was hoping "Jane" would
turn out to be a riff on the Tarzan stories, casting the PC as someone
for whom VERB NOUN grammar was perfectly natural. Still think it
would be a cute idea.
Jeff
Jdyer41
T-Zero latches on to whether specfic words are verbs or nouns, but not the
order. So you can type 'compass drop' and it'll interpret it the same as 'drop
compass'. However 'north go' won't work because 'north' is considered a verb
in itself and then 'go' will also be considered a verb so it thinks your
sentences
has two verbs. (Whereas if you type 'go north', the parser will assume north
takes the noun form.) On the whole the sum result is the same as
an imperative parser (although the game has some Nord & Bert style
puzzles where wordplay comes into effect).
There was a discussion on ifMUD a while back on 'chatterboxes'. These
are programs in the style of Eliza. What if someone tried to tell
an IF story in this format? It'd have to be a conversation with a confused
robot or a madman or the like, but it'd still be interesting.
Jason Dyer
jdy...@aol.com