- gutvol-d - lists.pglaf.org

playing the italics game
by Bowerbird＠aol.com 28 Oct '12

28 Oct '12

here i finish up my thoughts on italics, by myself... as i said earlier, to get accuracy through the book, it's almost a necessity that more than one person go through the book. it would be silly to assume that a sole producer, no matter how dedicated or skilled or tenacious, had located _every_ instance. (i am not saying it cannot be done; i'm saying that it would be silly for us to _assume_ that it was so.) so, then, a different sort of question poses itself: how do you motivate additional volunteers to do the job the second, third, or fourth time around? it's usually easy enough to find someone to do it the _first_ time, because it's virgin territory, and every find is a new hit, and _feels_like_progress._ it's the second, and subsequent, times through it that the job gets unfulfilling, because most of the instances -- 80% or 90% -- are _already_ marked, so there are fewer cases where you get the "rush" of finding a new one. (and fixing a false-positive, although valuable, is rare, and even less thrilling.) *** but there _is_ a way to get the best of both sides. and that's to _hide_ the visibility of current italics, so the person can "have the fun" of finding them. then, because we do know about the presence of many of the instances, we can reward the person with _recognition_ that they "found" what we "hid". and we can also show them any that they missed. so we're both (1) rewarding their successes, and (2) training them, by showing any shortcomings. this keeps 'em motivated, and makes 'em better. *** moreover, we can give them _extra_credit_ when they find an instance that we didn't already know. after all, that's why we do the additional checks; so people find new cases we didn't know about... so that's what we're gonna give good rewards for. in my "italics game" demo, if the italics you clicked had a blue bounding-box, it meant, "hey, good job! you found one of the instances that we had hidden." in our game, this "find" might earn you one point. (or 1000 points, given the point-inflation today.) but if the italics you clicked had a _red_ bounding, it meant that you'd _discovered_a_new_instance_... (this isn't just something arbitrary that i made up; the words with red bounding-boxes are instances that were _actually_ missed by either david or jim, and usually both, in their digitizations of "huck". so, according to them, you _did_ find a new one.) so in our game, this "discovery" of a new instance might earn two points, or five, or 2000, or 5000. (a nice twist might be to base it on the number of rounds so far; it'd be 2 points in the second round, and 3 points in the third, 5 in the fifth, and so on. thus, the more people who'd missed it before you, the higher the reward you'll receive for finding it.) we could have other game-play considerations too. as one, of course, there would also be _penalities_ -- _deducted_ points -- for any missed instances. we could decide how heavily we wanted to charge people for such missed instances, but i'd say that it should be at least _twice_ as much as we reward them for "re-finding" a known instance, since the whole purpose of this exercise is to find new ones, so we want subsequent checkers to be _vigilant._ and we might want to base this penalty, too, on the round, to balance the increased "reward" for finding new instances. so you'd lose 2 points in the second round for misses, 3 in the third, etc. as another wrinkle, we might want to plant some "false misses" in the mix. here, we would _say_ the person had "missed" an instance, but we will then point to an _unitalicized_ word. this would be the equivalent of a "false positive", and thus we'll want the person to _appeal_ the judgment. that way, if round 2 incorrectly italicizes a word, then round 3 will be penalized for not finding it, but the player in round 3 will see that the word is _not_ supposed to be italicized, and _appeal_. and our referee will come and uphold the appeal, thus _removing_ the incorrect mark from round 2. (and yes, retroactively penalize the round 2 player.) so the false penalties teach the players to appeal, and that benefits us by removing false-positives... *** like any social game, you'd post the "standings", and give players a message-board to chat, boast, trashtalk, bellyache, and just generally have fun... *** so, would this kind of "game" thingee work? i dunno. don't care, really. i'm just a coder, building a tool to help automate a dull task... if you want an example of a non-game use of it, see: > http://zenmagiclove.com/misc/italicscheck241.html > http://zenmagiclove.com/misc/italicscheck258.html > http://zenmagiclove.com/misc/italicscheck271.html > http://zenmagiclove.com/misc/italicscheck303.html > http://zenmagiclove.com/misc/italicscheck313.html > http://zenmagiclove.com/misc/italicscheck320.html > http://zenmagiclove.com/misc/italicscheck326.html these are the same 7 pages i had set up as a game, but here we just have the interface of the tool itself. you'll see that each instance of italics is _highlighted_ by a colorized box in both the left and right margins. (lines with multiple instances display multiple boxes.) these highlights are created by analyzing the text for the page, and seeing which lines have italics marked, then putting the margin-highlights at the appropriate place on the scan where we computed that line to be. or -- if people think it works better -- we can also mark each of those words themselves, with borders, the same way we did with the "game" version earlier. (to see this, click the "turn them on" button up top.) what the checker is checking for, then, is congruence with an italicized word in a line and a highlight for it. any noncongruence -- italics without a highlight, or highlight without italicized word -- needs attention. also note that, new in this version, i've included a _button_ in the right margin, for each line of words. when a line "needs attention", you'll click its button. this will send a report that that line needs attention. this is for when the text is sufficiently correct that changes will be rare, and we would want 'em to be confirmed by someone before they are committed. *** now let's recall how i described the current workflow: > look at the pagescan and note > where there are italics, then > check each instance in the text, > and mark it in the event that it > hasn't already been marked. > and then, just to be safe, > make sure that every italics word > that _is_ marked in the text > is actually an italics in the scan. so there are 2 entities in front of you on your screen. one is the pagescan, and the other is the text-file... and most of what you're doing is shifting between one and the other, and then back to the first one... look at the scan, then the text, then back to the scan. repeat, repeat, repeat. like watching a tennis match... haha! my goodness: "like watching a tennis match." sometimes i'm so funny i crack _myself_ up! ;+) now look at how this tool makes that more efficient. first, the text-file is removed from the picture entirely. all edits could be registered automatically by the tool. (however, i suggest instant review by an administrator, so that false-positives will not even get a toehold, and to address anomalies such as don's partial-word italics. this is the "confirmation" that i was mentioning earlier.) still, a person who's _checking_ doesn't mess with text. this means there's no loss-of-attention involved in the task-switch from examining-the-scan to text-editing. moreover, it is unnecessary to look at the text-file to see if an italicized word is marked, because the mark is shown _right_on_the_scan_itself_, where it's needed. you _know_ if a word's marked, or if it still needs to be. so if you see an italic _with_ a mark, you know it's fine. if you see one _without_ a mark, you know to mark it... and if there is a false-positive, you can see that as well, because there is a mark, but no italic-word alongside it. so all that tedious back-and-forth-and back movement of your eyes and attention from the scan to text to scan is eliminated. your eyes and attention stay on the scan. this is the essence of a _huge_ boost in task-efficiency. and that's why we build tools in the first place. are there any questions? -bowerbird

1 0

don't fight human nature
by Bowerbird＠aol.com 28 Oct '12

28 Oct '12

don said: > I would like someone to try training someone > to use a subset of HTML markup, and use it properly, > and see how it works. does the person know any .html to begin with? > Even with a subset, there are lots of ways > of applying that subset improperly, and > having the display view look just fine on a PC. yes, that's a huge part of the problem with .html... and it's worse if the person knew any .html before. > I suspect users will be unable to > avoid the temptation to tell themselves > that "if it looks right, it's right,", interesting that you call that a "temptation"... > especially in cases where > the match between the markup > and the structure of the text > is unclear or complex. i guess you're saying that in such cases, it's not gonna be obvious "it looks right". *** your problem, in _my_ opinion, is that you are fighting against human nature, instead of using it to your _advantage_. my system adopts as a primary rule that "if it _looks_ right, it _is_ right". and, conversely, "if it doesn't look right, you need to change your input so it will." -bowerbird

5 5

Re: [gutvol-d] any thoughts on the game?
by Bowerbird＠aol.com 28 Oct '12

28 Oct '12

don said: > 1. It's an example of a way to enforce my recommendation > to proofread for italics only. as it was shown, yes. but that wouldn't have to be the case. > 2. The first page exhibits a flaw I found in another ZML text; > it only allows the selection of your definition of a whole word; > so "*I*'d" can only be displayed as "*I'd*". there's nothing inherent about that. z.m.l. can italicize just part of a word. and i did indeed do precisely that, several times in this book. the problem in doing that, however, is that it looks like crap. the typography in a browser doesn't handle that very nicely... anyway, back to the world series... -bowerbird

1 0

any thoughts on the game?
by Bowerbird＠aol.com 28 Oct '12

28 Oct '12

anyone have any thoughts on "the italics game" i posted? or on an italics-checking tool, in general? how about the world series -- reactions? what do you think about the presidential election? a possible resurgence of an "arab spring" next year? world peace? hunger in poverty-ridden nations? global warming? the coming mass-extinctions of wild animals? isn't "dancing with the stars" a hoot this season? who's your favorite "american idol" contestant? how about survivor? -bowerbird

2 1

don't get yer knickers in a twist
by Bowerbird＠aol.com 27 Oct '12

27 Oct '12

don said: > The alternative scenarios I can think of are: > > 1. The site has been hacked. But how could that be, > if it's still up? > > 2. Marcel, or someone else, has misrepresented > their authority to speak on behalf of the Corporation. > (And why is the attribution to the webmsster and > principle developer, without mentioning who that is? > Except to reinforce that this is published on PG's authority, > and not Marcel's authority.) > > 3. The PG site url has been hijacked, or actually belongs > to some US affiliate of PG that doesn't represent PGLAF. > > 3. PG is not who I thought they were. well, first of all, your numbering skills are off-kilter again. but more importantly, don't get your knickers in a twist... this is just marcello being marcello, overstepping his bounds, believing that _he_ is project gutenberg, just because he can post something -- whether authorized or not -- to the website. i'm guessing that greg never "approved" this, or even had the slightest idea that he'd have to tell marcello "you can't do that". like your 4-year-old making crayon drawings on the den walls. so as soon as greg gets drift of it, the "review" will be removed. and even if greg _did_ approve it, all we have to do is say "wha?", and he'll come to his senses. this is clearly not what p.g. does... but geez, don't get all hysterical... it's not that big of a deal. indeed, even if the "review" stays up, it ain't the end of the world. that's not what the mayans meant. -bowerbird

1 0

playing the italics game
by Bowerbird＠aol.com 27 Oct '12

27 Oct '12

nobody has thoughts on a tool to help mark italics? why does this not surprise me? at any rate, the reasoning will be explained later, but for now, go try "the italics game" -- if you feel like it. when you find an italics word on the scan, click it, so it's surrounded by a bounding-box, either blue or red. (the bounding-box toggles on and off if you re-click.) when you're done with a page, click "grade me now", and you'll get a report on how well you performed... > http://zenmagiclove.com/misc/italicsgame241.html > http://zenmagiclove.com/misc/italicsgame258.html > http://zenmagiclove.com/misc/italicsgame271.html > http://zenmagiclove.com/misc/italicsgame303.html > http://zenmagiclove.com/misc/italicsgame313.html > http://zenmagiclove.com/misc/italicsgame320.html > http://zenmagiclove.com/misc/italicsgame326.html this game is not timed, so take as long as you need on each page, but remember that this book has 366 pages, so 1 minute per page means this job will take _7_hours_. even 1/2 minute on each page still means 3.5 hours... to get it down to 2 hours, each page has just 20 seconds. (19 seconds if you want a 5-minute-break after an hour.) -bowerbird

1 0

when will they ever learn?
by Bowerbird＠aol.com 26 Oct '12

26 Oct '12

keith said: > "Apparently" a master format can work! yes indeed, it is rather mind-boggling that jim -- not to mention so many others on this list -- can still be "surprised" by that fact, even though i have been proving it with pudding for so long... "apparently" they will never learn. -bowerbird

1 0

polishing polishing polishing
by Bowerbird＠aol.com 26 Oct '12

26 Oct '12

still doing some last-minute stuff on huck finn... once again, as usual, i am struck by the required amount of attention-to-detail involved in polishing text for an e-book. even for an obsessive type like myself, it can overwhelm... (although, it should be said, most books are _not_ as hard as this one, which presents a variety of atypical challenges.) sometimes even something you think is going to be _simple_ ends up using more time and energy that you want it to take. just as an example, i did very little work on italics as i was cleaning up this text for "adventures of huckleberry finn", since i planned on splicing in the italics from jim's #32325. except now, in my confirmation work, i found at least _ten_ instances where jim might well have missed an italics word. (can't say for sure, because i don't know exactly what edition he used, but at least according to my edition, he missed 'em.) that's not a criticism of jim, or the quality of his work, since -- for all i know -- my file might be missing a dozen more... the whole reason i decided to import jim's italics is because i hate italics-checking, and it is hard, and time-consuming, and i miss lots of 'em, so the entire exercise is frustrating... of course, _part_ of the problem here is that i am trying to move this book as close to perfection as i can, when really, i should just say "ok, let's bring in the smoothreaders now." that "smoothreader" stage is an important one, because that's where the cost-benefit ratio achieves its primacy... smoothreaders will _not_ necessarily catch _every_ error. but, in a nutshell, if smoothreaders don't catch an error, then it's probably the case that the error is unimportant. the object, of course, is to move every e-book to perfection. but that can _cost_ more than the _benefit_ would be worth. if an error is so slight that it's not even _noticed_, then -- the odds are -- it is unimportant enough to matter much, so the act of detecting and fixing it simply _isn't_worth_it._ there are lots of cases, for instance, where it's hard to tell if something is a semicolon or a colon. but does it matter? probably not. especially since the typesetter didn't seem to be following any kind of consistent rule deciding between 'em, even in the cases where there is no uncertainty which one it is. sometimes it's hard to know, then, when to let something go... that's why smoothreaders can be so useful... -bowerbird

4 6

aloha-readers and polishing and italics
by Bowerbird＠aol.com 26 Oct '12

26 Oct '12

i've made references to "smoothreaders" lately, mostly because that's a term commonly used over at distributed proofreaders... but i probably need to use some other term instead, because i'm not really talking about the same thing that they do at d.p. over there, "smoothreading" is a download-the-text-only deal. sometimes it can be .html, not just plain-text, but the point is that it _does_not_ involve the actual _pagescans_ themselves... whereas what i am talking about here _most_definitely_does_ involve an actual comparison of the text against the pagescan, albeit if only in those cases where there's doubt about the text. but the interface that i am talking about _always_ has the text displayed side-by-side with the pagescan, for easy comparison. *** also, at d.p., "smoothreading" is considered to be the _end_ of the process of preparing the text. i can see that viewpoint, but that means that the participants are volunteers who're already _embedded_ within the distributed proofreader infrastructure. i envision people who would consider themselves to be outside of such a system. and rather than the _end-point_ of the process, i would think they'd consider themselves to be the "beta-readers" for the e-text as it was _beginning_ its journey out to cyberspace. in other words, it's less of a "graduation" than a "commencement". then again, maybe the best way to think about this stage is as _both_ a "graduation" _and_ a "commencement", so perhaps we should be using one of those hello-goodbye words like "aloha". so i propose we call them "aloha-readers"... *** further, since these extra sets of eyes would be of the most value to _independent_producers_, i'd suggest that project gutenberg itself should recruit and train and enable these "aloha-readers"... i don't know (or care) where it would fit in the workflow that now has whitewashers doing (at least some of) this kind of last-minute checking -- perhaps, in the spirit of "aloha" it could be _both_ -- i'm just saying that it will be extremely helpful to solo producers. *** don said: > In my own proofing the only way I can have > any confidence at all that I'm catching all the italics > is to scan through the entire text looking for nothing but. i think that's true of a lot of people. and even then, any "confidence" might be undeserved. because most people will _still_ miss 5% to 15% of them, and sometimes even more. i have proven it with myself. i have stared and stared at some pages, _knowing_ that it contains one or more italicized words that i'm missing, and _still_ been unable to spot any. it's just the way it is. so i feel the only way to gain a good degree of certainty is to have the pages examined by more than one person -- and i wouldn't _bet_ on 99% accuracy with fewer than three. but of course, if/when you have independent digitizations, you can use the _comparison_methodology_ i'm discussing, and i can guarantee you that it will jack up your accuracy... so far, with this copy of "huck" that i'm digitizing right now, i've used jim's italics _and_ david's _and_ some of the o.c.r., and i am very confident that i'm getting fairly good coverage, although even then you'll notice the hedge with "fairly good". (and i can also report that both david and jim missed "some".) further, i'm making my text available, so you are welcome to take on the challenge of finding any italics that i've missed... > As soon as I start noticing (and marking) anything else, > I start missing them. i can also say that that's true for me too. i need the focus. > It also destroys my accuracy when the text already has > italics markup in it and I must check for false positives. well, there should almost never be "false positives" in a file. not for italics. false positives mean your workflow is flawed. but you bring me to a good point i have been preparing... for the proofing of italics (and other styling) we have a need -- an _overwhelming_ need -- for tools to help us do better. the current methodology -- look at the pagescan and note where there are italics, then check each instance in the text, and mark it in the event that it hasn't already been marked. and then, just to be safe, make sure that every italics word that _is_ marked in the text is actually an italics in the scan. whew! this is tedious, back-breaking work. fatiguing. so, as you might expect, i've got some ideas for a tool. but i'm gonna let y'all think about it for a while first... so see if you can come up with anything, and then i'll dish... -bowerbird

1 0

Basic simple test case.
by don kretz 26 Oct '12

26 Oct '12

Here is a PG project that needs some formatting help. The HTML conversion would probably best be termed "fail". Yet it should be a simple case. How would you want proofers to submit it after "matching the scan"? What would you consider to be a properly marked up copy in the format of your choice? How should it get there from the previous step? http://www.gutenberg.org/cache/epub/14668/pg14668.html

11 43