Comments on: Spelling

By: Tom Morris

Tom Morris — Mon, 16 Aug 2004 11:18:23 +0000

Spelling reform is great. And it works. Spelling reform by government is not great, and does not work. Spelling reform is a slow process taking generations for spellings to be accepted and formed in to languages via dictionaries, common usage and adoption by different people over time. Government have enough control over our life without trying to control our language too.

By: Another Damned Medievalist

Another Damned Medievalist — Sat, 14 Aug 2004 18:25:46 +0000

Ugh—never comment early in the morning. Des—I am not saying we should return to OE (which is, after all, a different language) or that language should be stagnant (although I will continue to fight for the subjunctive, at least in the case of reported speech and ‘if … were’). All’s I’m saying is that spelling reforms to make things easier are EEEEEVILLL, because dumbing down never helps and because I really believe that learning more about languages does more good.

By: MM

MM — Sat, 14 Aug 2004 09:27:59 +0000

des/another d. m. Are you aware that the German spelling reform did exactly this – return to the roots? Some German words whose spelling concealed their origin were given a spelling that showed the origin, e.g. Stengel became Stängel, Quentchen became Quäntchen, verbleuen became verbläuen. At the same time the opposite was done with foreign words, so you can write Delfin as well as Delphin (there are quite a lot of alternative spellings) or Pappmaschee as well as Pappmaché. Well, I suppose this isn’t about the German spelling reform now.

By: Des von Bladet

Des von Bladet — Fri, 13 Aug 2004 17:41:30 +0000

ADM: Or maybe it’s because you success has desensitised you to the difficulties others reallly do experience. In case you’ve forgotten, I oppose spelling reform (in the weak sense of thiniking it won’t happen and not caring very much) except for the isolated examples of articially retrofitted ‘b’s that I gave. (I don’t think, of course, that that tweakette will happen, either, but I am infinitessimally less sanguine about it.) But riddle me this: The “straw” in “strawberry” was, in the Old English form of the word, transparently connected with the OE form of “strewn”, and it is clear that it refers to the way wild strawberries are found as if strewn along one’s path, hoorah! But, alas, the ravages of time have rendered the compound semantically opaque! (Example swiped from Bloomfield Language 1933/5) Now, are you: * prepared to agitate for a change in the spelling to “strewnberry” (leaving the pronunciation alone, of course) to repair this tragic and untimely morphological disaster? * content that, notwithstanding this sad loss, the One True Spelling system in which we rejoice is the best of all possible spellings, even when its ways are difficult to fathom? * not really bothered? And why? (Answers on the back of a fascicle, please, to: The English English Academy of England, Behind the bike sheds, The Old Vicarage, Cholmondley Street, Gloucestershire)

By: Another Damned Medievalist

Another Damned Medievalist — Fri, 13 Aug 2004 16:36:42 +0000

Des— Maybe it’s because ever good language teacher I’ve ever had leaned towards philology, and maybe it’s ‘cause I’m a historian. All I know is that Spelling isn’t particularly difficult if you let rote memorization take over part of the workload. And rote memorization is not a bad thing—it can be a useful tool. I know that my students are always interested when I talk about word roots and relationships in regards to history, and those discussions seem to stick in their heads much better—a simple example is the burg, bourg, burgh thing—what the word meant originally and how it’s been applied since. If we were to all agree to drop the ‘unnecessary’ h at the end of burgh, or even spell it phonetically as borough or boro or burrah, that lesson would be lost. Sorry, but I find learning spelling to be a minor inconvenience compared to losing a lot of the richness of language.

By: Des von Bladet

Des von Bladet — Fri, 13 Aug 2004 14:18:05 +0000

Scott: I will look at spell-check technology, but my gimmick is precisely client-side spelling reform, as it were – I want persons not to care at all about the spelling and use a (correctable, for sure) inverse pronouncing dictionary-esque map. As for speech recognition, your remarks can be summarised (polemically, yes) as “speech recognition people don’t know anything about phonology”. Post-Chomskyan phonology is well worth not knowing much about, generally, so the blame is distributable but even so, it’s pretty desperate out there. (Thanks for the review, nonetheless.) I think several strands in pre-Chomskian phonology offer plausible leads for a way in to modelling phonology, which is what’s needed. (My day job isn’t AI, and amateur cynicisme doesn’t have quite the same appeal.)

By: Scott Martens

Scott Martens — Fri, 13 Aug 2004 14:05:51 +0000

Oh, and as for kana-based cellphone entry, I saw the system you’re talking about on Language Log a few months back. I’ve always wanted to see if you could make it work using a one-handed keyboard so that you could chord kana in. I’ve always thought that would be cool. And, if you extended it to Chinese using an attack-nucleus-coda model for each syllable, and then chording in the appropriate hanzi from an on-screen menu, you might be able to simultaneously solve the Chinese/Japanese data entry problem, establish a defacto Chinese phonetic alphabet, and figure out the best way to marry a keyboard to a PDA. That would be cool. And might make a ton of cash on the side.

By: Scott Martens

Scott Martens — Fri, 13 Aug 2004 13:58:17 +0000

Des, yes, spellcheckers reduce – or at least postpone – the problem. I would certainly agree that there are more pressing issues of social justice in the wealthy anglophone states than spelling reform. But I do want to highlight the political character and political consequences of the decision that there should be just one written English. This too was an unexamined element of the Ebonics debate way-back-when. I came across a paper a while back suggesting that English spellings may be the principle cause of dyslexia. I can’t remeber who wrote it, but I’ll see if I can dig it up for you. I don’t remember being too impressed with the conclusions at the time. As for the mechanics of spellchecking, you can use Levenshtien distance and Soundex to guess what someone is trying to spell even when they totally screw it up. If you use context vectors to guess which word they meant to say as well, you can get 95%+ accuracy in automatic spellchecking. My understanding is that this is the current dominant paradigm in spellchecker design. The major problems have to do with morphology errors, for which you need a local parsing model to make an automatic correction. The real problems are lazy spellcheckers who correct words incorrectly. As for the speech recognition problem, the problem is no longer phoneme recognition, since for isolated sounds the machines do that as well as people do. (Which is to say, badly.) The problem is in the internal language model used to prime the recogniser for the next sound and to correct errors. It seems people primarily use an internal language model – some system of expectations about what the next sound or word will be – to recognise speech. Getting the machines to do that is a bitch. Lernout & Hauspie blew countless millions on the problem, and managed to make some progress but never did quite “solve” the problem. (One of my grad school profs was – IIRC - the head of research there at one time.) There was somebody in the LSA community the last time I was in grad school who was getting good results using distribution matrices and singular value decomposition in lieu of a rule-based language model, and getting good results. I don’t know what has become of his work since. I did some experiments myself with spreading activation networks for the same purposes, but my results were mediocre and proved more useful in automated thesaurus development. There are suspicions that the problem is not fully solvable as conventionally understood. Too many extra-linguistic factors have to go into a language model sound enough to minimise error to the level of the human speaker. We can’t encapsulate a lifetime of human context knowledge into a piece of software, and we are not really any closer to doing so than decades ago when speech recognition was first dreamed up. AI people like me are professionally pessimistic. Our highest aspirations are to minimise error. Actually solving problems is not within our expertise.

By: Des von Bladet

Des von Bladet — Fri, 13 Aug 2004 13:07:22 +0000

Scott: I’ll look for Eboniciste studies; thanks for the tip. But as handwriting is progressively obsoleted by computer-mediated text, and spell-checker technology trickles down, then surely orthography ceases to be a barrier to entry. So long as my spell-checker can figure out what I mean, of course. In fact, a dedicated software front-end that mapped phonetic (or reformed) spelling to standard is not only doable, but very closely related to things I’ve been doing anyway. (Banish your spelling woes! With the New Von Bladet Wordiciser you can spell the way you speak and our software will take care of the rest!) (Of course, I want to get seriously disgustingly rich by finally solving the speech recognition problem, but this might be a nice way to start.) Have you seen/heard about Japanese phone tech, though? The numeric keypad is perfect for kana entry, and the subsequent kanjification is menu based, and everyone seems to like it.

By: Scott Martens

Scott Martens — Fri, 13 Aug 2004 12:06:38 +0000

Des, as for studies on literacy and spelling, there are some, but not where I can get to them quickly. A lot of the Ebonics research touches on this stuff and I suspect you could find what you’re looking for by sifting through their bibliographies. I’d gladly do it myself, but I haven’t the time right now or the easy library access.

By: Scott Martens

Scott Martens — Fri, 13 Aug 2004 11:58:35 +0000

Des, I’m not sure that we disagree about the likelihood of English language reform ever happening. Just because I think it wouldn’t be the end of the world if some of the worst bits of English spelling – especially the ones that I have trouble with – were to go away doesn’t mean I think I’m ever going to see it happen. The only plausible scenario I have where it happens in the next century involves English actually becoming an international vehicular language in a much wealthier world. In a world where 80+% of English users were non-native and accounted for the vast majority of the money earned by anglophones, I think it is just possible that they would demand a reform and have to power to do it despite resistance from English speakers. But I doubt it, on essentially political grounds. Learning English is, just like French in the 18th century, another skill that requires time and money to learn, and thus separates ruling classes from ruled classes. Simplifying English undermines that end even for non-natives.

By: Des von Bladet

Des von Bladet — Fri, 13 Aug 2004 09:16:16 +0000

Another Damned Medievaliste: I do think you're odd. Since I dout [sic], however, that you're quite odd enough to endorse a proposal for a spelling based on reconstructed proto-Indo-European roots, you presumably accept that the line has to be drawn somewhere. For me, etymologically-motivated retrofitting of silent letters not present in the source language of a loan word is the wrong side of that line. As a philologiste, you apparently find a writing system optimised for philologistes convenient. (Quelle, as they say in West Belgian, surprise.) The bulk of the persons who end up grappling with Engleesh orthography are not philologistes and never will be and there is simply no "maybe" about it. To add to what Scott says, it's an orthodoxy in linguistics that in practice lots of people need to be able to read, but almost nobody needs to write. That used to be true, but it increasingly isn't. My theory, which is mine, is that the standards of spelling and written grammar that are always said to be falling are doing no such thing, but rather a wider range of the competence spectrum is now on view. But Scott is also right that it's essentially a political issue, and correspondingly wrong (IMHO) in so far as he thinks that the political will to address it is likely to come when it's called.

By: Scott Martens

Scott Martens — Fri, 13 Aug 2004 08:47:18 +0000

Bza – one of the issues in language management is the notion that literacy ought to mean more than just being able to read and understand, it means being able to construct texts. Communication is a two way street. While eliminating unnecessary distinctions may not raise people’s ability to comprehend texts significantly, it can raise their ability to produce them by quite a lot. It’s not just being able to read books that matters. Linguistic empowerment also means being able to use print media to communicate what you want to say. With the rise of the ‘Net – and the proliferation of micropublishing fora like blogs – I think this element of language politics has become terribly important. In the past, relatively few people expressed themselves regularly in print. Now, tens – perhaps hundreds – of millions do every day. This problem is extremely acute in Chinese and Japanese, much more so than in English and French. But it’s present in those languages as well.

By: Another Damned Medievalist

Another Damned Medievalist — Fri, 13 Aug 2004 01:58:36 +0000

Des— i hadn’t realized that the b was added later, but you misunderstand me if you think I meant that it helped with Latin. What I meant was that that link to Latin Debeo, opens up a whole bunch of cool relationships in English—the relationship between owing and what one ought to do, for example. Maybe I’m odd in that I relate words and idioms back and forth in as many languages I can—I find it very helpful when I’m reading something (in English, Lating, or German, usually) from a different time period. But like I said—maybe other people don’t find it useful to know where words come from and how they evolve in other languages …

By: dsquared

dsquared — Fri, 13 Aug 2004 00:51:30 +0000

Will you be going to Loughborough, or just passing through?