Emily Short -- Down with Parsers!

Emily has a very interesting discussion here of why she thinks the command prompt/parser has to go. Her argument, essentially, is that it promises to understand any input, but most input in fact is not understood, and this keeps new players away. The prompt doesn’t tell the player about the affordances of the system, as she puts it.

It’s very interesting and there’s a lot of discussion on it – you should go read it.

A few thoughts:

  1. I don’t have that much of a problem with doing something esoteric as a hobby – I play NetHack, and some of the CDs I listen to don’t even have their track titles entered in the database of that sort of thing. But it does seem sad to do something that only a few people can participate in, especially something with “interactive” in the title, which doesn’t exist without an interactor. So I’m all for greater accessibility; though we should consider the costs.
  2. But cutting off the free-form parser seems like it’d have some costs – there can be a great satisfaction of thinking of just the right thing to do as opposed to having the options laid out for you. Though Emily’s envisioning a system where you have enough choices that the way forward isn’t obvious.
  3. That can be a problem, too, though. The troll on the thread has a point that the usual IF paradigm of going around, picking up objects, and using them in a few different ways is on its face pretty constraining. Explicitly restricting the verb set to a few things would seem to just reinforce that paradigm even more – unless we make the move to USE, or to keywords that have different (and possibly contextual) effects, or menu-based choices. (Conversation games are the ones I can think of that most break out of this paradigm – and a lot of them are menu-based or keyword-based.)
  4. This would require some totally new frontends. The demo Emily posted, where you type a keyword and then it gives you some verbs to type, just doesn’t work for me at all (partly because the command window is too small, but I don’t think that’s all). You’d really want to click on the keywords and then use a dropdown menu. And then you’re practically doing point-and-click games with words instead of pictures.
  5. Point-and-click games are an interesting comparison in general. Why are they more accessible than IF?

Anyway, an interesting discussion.

I thought that article was interesting. I do feel the open ended parser stands in the way of bringing in a wider audience, but I don’t want to see it gone, nor do I think that a parser is essential for every piece of Interactive Fiction. I feel it’s an interface decision similar to choosing what conversation system to use, it depends strongly on the context of the game. Some existing games would probably have been stronger with a select-the-verb interface (essentially ones that don’t have a lot of choices to begin with), but that is by no means universal.

I do think it’s important to educate new players on the limitations of the parser interface. Saying “just type anything and it will understand you” is clearly a lie. It’s similar to a command line interface for your operating system – it’s going to have more “power” than point-and-click, it will have a steeper learning curve, and to some people it will be more “fun” to use. Personally, I don’t want to see it go away, but I think authors should definitely feel encouraged to explore other options. Interactive Fiction does not need a one-true-interface.

Uhm, what? There are already games without a parser. Emily can go write one of those; they already have been invented though so she can’t claim the honor.

Furthermore, why would I care about new players? If the parser goes away, then I won’t play it, but the new players will. Now how does that help me exactly?

Respectfully, I think the post is more about why she is moving away from the traditional parser herself, rather than a call for everyone to ditch something that obviously has a community of vocal supporters.

Arguably a point and click game game has just one verb - “click”. As much as the IF community likes to claim that “guess the verb” is an issue of the past, the fact remains that people still find the enormous range of potentially valid commands to be overwhelming.

The parser is not going away. People are still making silent films and hand-drawn animations in cinema - the same is true for video games. Artforms are rarely forgotten, and even then can be reinvented.

This kind of thing is of interest to people who want to create games with text output and input, but who feel that the IF parser pushes them to make a kind of game they don’t want to. It’s of interest to people who are, as Emily writes in the article, more interested in devoting their development time to providing nuanced responses to actions within a specific framework than worrying about what response the game gives to >TASTE FIRE.

To clarify this a little further: I have a couple of WIPs for which this wouldn’t be a possible strategy, so I’ll likely release some more standard-parser games myself. All the same, I think I’ve hit a wall in terms of how friendly I can make those, and if I want to create text-based games that I can offer to new players, I need to think about an alternative to the bare command-line prompt.

In that case, you are not trying to make text adventures more “beginner friendly”* but instead are simply moving away from the genre. A text adventure (or IF, if you like that term better) would not be a text adventure anymore if you take away the parser. For example, CYOA games are not text adventures.

  • Which IMO isn’t needed; they’re playable enough for everyone who’s interested.

Out of curiosity, are you responding to what I actually wrote, or to the summary of it?

I’ve talked to many, MANY people in the last couple years who said some variation of “I love the idea of interactive fiction, but I can’t stand the parser.” In my experience, that’s the standard response among gamers and designers who like story-heavy games but don’t play IF; it’s a much more common response than “oh, I don’t want to read text when I’m playing a game” or “I’m not interested in a game unless there is graphics” or “The only thing I want to do in a game is combat/racing/whatever”. So I don’t have scientifically collected data, but I’d estimate that the audience for a friendlier variation of the parser is at least an order of magnitude larger than the hardcore IF audience.

In any case, what I describe in the blog post isn’t about throwing away parser-style input entirely, but about providing enough help that it’s clear what you can do from moment to moment.

(I’m also not trying to claim credit for inventing anything. I do work on graphical games for a living these days, so I have plenty of outlets for that. But I think there are things that can and should be done with text that our current structure does not serve well, and that we could preserve many of the neat things about the parser while still engaging a broader audience and enabling some new kinds of work and new authors, if we adjusted our methods a bit.)

1 Like

Well, I think Emily wasn’t just talking about her personal taste but saying that the parser has objective problems. (Thinking of the comment “We’ve tried to train players.”) She can speak for herself, though. I’ll admit that my summary was kind of stark.

Arguably a point and click game game has just one verb - “click”. As much as the IF community likes to claim that “guess the verb” is an issue of the past, the fact remains that people still find the enormous range of potentially valid commands to be overwhelming.
[/quote]
I think that’s a little bit of an oversimplification for most (of what I’m thinking of as) point-and-click games. You can click a spot on the screen, click an item in the inventory to use it on a spot of the screen, double-click an inventory item or click a special box or some other interface to examine it more closely, do various things with the item once you’re in the examine screen (move it around, use another item on it, etc.) Some games have a special combine tool, so you click somewhere and then click on two inventory items – which usually is a big pain. But there’s at least a little variety in the verbs that can be executed by clicking; to say that you do everything in the game by clicking isn’t necessarily any different to say that you do everything in IF by typing. And that’s not to mention the ones where you click on an object and it brings up three actions for you (usually “examine,” “take,” and a joke it seems).

But! I think you’re right that the clicking interface is much more manageable than the typing interface. Point-and-click games naturally present a lot more information than IF and allow you simpler ways to act on it. Which is maybe why the standard complaint about point-and-click is “hunt the pixel” (where in this vast field of things can I apply my few modes of action?) while the standard complaint about IF is “guess the verb” (which of the many possible verbs can I apply to the few things that have been described to me?). And, coming to point-and-click cold, clicking on the most obvious things will get you a lot farther than typing the most obvious things in IF would, coming to that cold.

Still, point-and-clicks have their own barrier. If you don’t know that you have to click the sides and undersides of things to look at them, you’ll hit a wall pretty fast – let alone if you don’t know that you can act on things your examining. (Without even getting into the cliches of the genre.) I think the reason I found point-and-clicks more accessible initially is partly that it was just more engaging to follow walkthroughs and see what they’re up to in point-and-clicks; perhaps because you’re actually translating text into action rather than just typing the text that’s in the walkthrough.

Anyway, that’s part of what I was thinking of with the point-and-click comparison. (Also, the IF community likes to claim that guess-the-verb is a thing of the past? srsly?)

I’d say I’ve come to the conclusion that there are essential problems, that I think they might be addressable without completely losing the current benefits of a parser, but that I don’t at all expect everyone to agree with me.

I do wonder to what extent a game could just change people’s expectations of the parser without changing the thing itself at all. I’d love to figure out how to disable the libraries in an IF system and just write a game with three explicitly stated standard verbs (plus some easter eggs) - and everything else just says, “Sorry, try one of these three verbs:”

Yeah, I did re-read the post after I wrote that and then kind of regretted it. I think we both overstated a little. :stuck_out_tongue:

Well, not quite. Sometimes I see IF come up somewhere it doesn’t normally, someone complains that they never know what to type, and an IF veteran appears to inform them that this is called “guess the verb” and is now considered bad design, and good modern games don’t have it. And using the strict definition they have in mind - you have to pry the padlock off with the crowbar, but PRY PADLOCK WITH CROWBAR doesn’t work - they’re right.

What they don’t realise is that the players they’re talking to are referring to a situation where they’ve encountered the padlock and have a bunch of items in their inventory other than the crowbar, and know a bunch of standard verbs, and are thinking, “Is there something I could type right now to open this door? How would I know?”

I discussed Walker & Silhouette with some indie game devs while I was working on it, and this was what pretty much each of them said - some of them also referred to having tried to write IF but getting bogged down handling actions they didn’t care about.

Not to be the guy that interjects smug literary pretension into an otherwise interesting discussion, but my perspective on this owes a great deal to Umberto Eco, who wrote in his Postscript to the Name of the Rose:

The context is an extended discussion about how his publisher had pressured him to overhaul the first 90 pages of that novel, which are replete with ornate description, literary allusion, medieval politics, and eschatological references: a far cry from the immediate gratification that so many commercial readers expect.

Despite a strong interest in the milieu and the structural parallels to the Sherlock Holmes tales of my childhood, it took me three tries before I was able to crest those first few chapters and set off into the story in earnest.

This is not dissimilar to my initial experiences with IF. Back in the 80s, I played Planetfall for ten unproductive moves and quit, never to return. It wasn’t until Anchorhead that I found a game compelling enough to draw me through the initial learning curve. After finishing that game, I had all the tools I needed to play any other IF game I cared to try.

I imagine that when books first appeared, when reading and writing were tools of the elite, illiterates must have regarded them the way modern gamers regard parsers. What could possibly be worth the countless hours of agony and frustration necessary to master the written word? Learning to deal with a parser requires the merest fraction of the effort children spend learning to read, and once mastered the ability can deliver commensurate joy.

The task for the IF author is to produce works of such brilliance as to compel reluctant players to engage the work. This is hard. For certain players, and indeed certain authors, it may be impossible. But it can and has been done; I hear Bronze cited repeatedly in this context, and Lost Pig at least as often.

How about (I7):

After reading a command: if the player's command does not include "verb1" and the player's command does not include "verb2" and the player's command does not include "verb3" and the player's command does not match "undo" and the player's command does not match "save" and the player's command does not match "q" and the player's command does not match "quit": say "Sorry, try one of VERB1 or VERB2 or VERB3; stop the action.

I haven’t tested this and it’s crude as all get-out, not to mention that if one of the verbs shows up somewhere else it’s a problem (which could be got around with regexes I guess), but it might work.

Yeah, I did re-read the post after I wrote that and then kind of regretted it. I think we both overstated a little. :stuck_out_tongue:
[/quote]
No worries.

Hee. That sounds familiar – I was actually thinking of editing the Nethack wiki article on text adventures to point out that you have to make an effort these days to write a game in which “pick up shovel” doesn’t work. And that random one-turn deaths are deprecated too. This will be the least efficient IF outreach ever, especially because nethackers probably like random one-turn deaths.

Mmmmyeah maybe. But if we’re going that route, I want to be really, really sure it’s worth it. Is the current raw unaided parser really giving us so many options that it’s worth sticking out for? Would the possibility and enjoyment really be completely undermined by something with a bit more support for the player?

When I stand back and look at this from the perspective of other kinds of game design I’ve been doing and/or studying over the past couple of years, the command line seems to stick out as an almost perverse bit of misdesign: it doesn’t lend itself to easy discovery, it doesn’t remind the user of what possibilities are available from moment to moment, it relies on expert-level familiarity with the commands, it makes a lot of work for the game creator, it’s very hard to make “juicy”, and it often creates moments of frustrating stuckness. Even in work by accomplished authors, I still sometimes run into moments where I can’t communicate a (correct) solution to the game because of guess-the-verb issues, dialect differences, etc.

I’ve had some great moments when I typed in something unexpected and it worked – I’m not trying to deny those experiences – but I’m wondering whether it mightn’t be possible to retain some of the parser’s freedom and potential for surprise while also making the experience more accessible.

In your blog post, you mentioned and even provided screenshots of games like “Gateway.” Those games did make themselves more accessible. Not sure if you know about this, but the “Legend” games were context sensitive; the list of verbs change according to the current situation. They’re not static. Another thing about them was that by pressing F4 (or maybe F5, it’s been a while) all of that went away, giving you a classic interface.

If I were to design a new system however, I would probably build upon where Legend arrived at. A graphical game with the illustrations being interactive (in Gateway and other Legend games, you could actually click on the objects in the graphics to open doors, take items, etc.) But I would never get rid of that “>” at the bottom :stuck_out_tongue: It just wouldn’t be the same without it.

Well, maybe it would be. Take the case of a game that is predominantly about investigation or exploration: Anchorhead, or King of Shreds and Patches, or Bronze. The enjoyment comes from not knowing what you will find. There is a certain mechanical unity here between game and parser, where you do not know what commands will work and the pleasure comes from finding one that does.

It may be possible to retain some vestiges of that joy with a different interface. I started playing Alan Wake the other day, and it seems cool so far. But the instant I stumbled across a manuscript page, I thought - how much cooler would this be as IF? Where finding those pages might require an intuitive leap, might force me to fully embrace the narrative, instead of merely undertaking an exhaustive exploration of a bounded 3D space?

Or take the case of a game that is about performing a sequence of carefully planned actions at the correct moment. Where the thrill is that of choreography, or dance. I would put Spider and Web in this category, and All Things Devours. Both can be brutal and unforgiving, and the parser only makes it more excruciating. But once you know the sequence of commands, once you have grasped the point and the goal, the parser becomes your best friend. Because it will allow you to take your time, and it will always respond to the correct sequence in the expected way. Again there is unity between the mechanism and the game; precision and consistency are demanded, but they are likewise rewarded.

I finished Assassin’s Creed 2 the other day. I rather liked it, up until the end, but the emotional response that lingers is one of frustration bordering on rage. The game contains tombs which require increasingly more elaborate and sustained levels of finesse to solve. Even when I understood precisely what was necessary, it often took dozens of attempts to execute the solution correctly. The controller and my reaction time created an obstacle far more difficult than any IF “guess the verb” moment I have ever experienced, because they lacked the requisite precision and consistency.

I think most people find parsers and prompts accessible. Google does pretty well for itself, despite a default UI that presents users with a giant prompt tied to a parser complex beyond mortal comprehension. They are only unfamiliar in the context of games, and that lack of familiarity gives rise to discomfort and fear.

My son is two and he loves books. But he really only likes books with big pictures and very little text. Open a book with more than 50% text on the page, even one with a compelling cover that he picked out himself, and he pushes it away and finds a different one to read. Because the text is meaningless to him, and because reading means something more (and less) than words to him.

I meant to draw a parallel between the task of collecting journal pages in King of Shreds and Patches and the task of finding manuscript pages in Alan Wake.

The IF version demands intuition, logical deduction, and environmental manipulation.

The triple A title version requires only mechanical persistence.

If the parser is the price I have to pay for hands-down better gameplay, so be it.

Sure. But the adjustment I’m suggesting neither throws the parser away entirely nor gets rid of intuition. You still have to think how you want to manipulate things; the point is to whittle the player’s interactions down to those manipulations, instead of including also huge amounts of flailing.

Yeah but, Google’s parser is complex enough that whatever you type in is very likely to work. (And if you aren’t doing fancy stuff with quotation marks and the like, I don’t think its parser is that complex – isn’t it mostly glomp-the-keyword? It’s what it does with those keywords that’s amazingly complex.)

Now imagine that he had been interested in writing something for immediate gratification, or just something simpler or more poetic, or more succinct, that expected the reader to expand on it with their own imagery and research, and it was the publisher pressuring him to add those 90 pages. Would that be any better?

Even accepting your IF/literacy analogy, is it so terrible to create a spectrum of games from the elaborate to the more straightforward? Is it even terrible to accept that most people would prefer the straightforward?

Or to put it another way, if I told you you had to learn how to ride a unicycle before you could read my book, would you read it, or would you read something else? If you chose to just read something else, I don’t think it would reflect badly on you, or the book, it would just show you had different priorities for how to spend your time.

I find it hard to classify any creative act as terrible. As a parent, I cherish children’s books - but how many centuries went by before books targeted to and tailored for children’s abilities appeared?

I don’t much care for Harry Potter or Twilight, but undeniably the content is compelling for broad swathes of the population who might otherwise be completely oblivious to the charm of books. With apologies to Admiral Jota, is Lost Pig the closest that IF authors can get to that sort of content? Probably not, although it is clearly more accessible and more entertaining than most IF.

Setting aside the legal difficulties, I believe a Harry Potter fanfic IF that garnered critical praise would do more to raise the profile of IF than any number of fiddly interface improvements.

Playing IF is nowhere near as hard as riding a unicycle, learning to read, or mastering a foreign language. It isn’t even as hard as Assassin’s Creed 2, where I spent two hours on the third scene trying to beat someone in a pointless footrace. Only its glowing endorsement from my brother saved it from an immediate return. This from a commercially successful game with an enormous budget and players numbering in the millions.

I have tremendous respect for your games in particular, but I do think it is somewhat hubristic to suggest that people avoid IF because of the incredible complexity of typing stuff into a prompt.

It’s not as if people are born knowing how to use search engines. It would be one thing if the discussion centered around how to make IF accessible to the people who find Google too intimidating. But it’s quite another to take as gospel the word of people who are perfectly capable of making their intentions understood to a parser.

I don’t suggest that they are lying, simply that they can’t be bothered to make the effort to learn this new way of controlling games. It’s like that ridiculous Xbox 360 game with the full-sized skateboard controller. Its control scheme is very much an obstacle; I would repeatedly fail my dexterity check just trying to stand on it. But if instead of being wedded to a pedestrian skateboarding sim, it was somehow a means whereby I could navigate a Cthulhu Mythos story, I might be compelled to make the effort.