The following online article has been derived mechanically from an MS produced on the way towards conventional print publication. Many details are likely to deviate from the print version; figures and footnotes may even be missing altogether, and where negotiation with journal editors has led to improvements in the published wording, these will not be reflected in this online version. Shortage of time makes it impossible for me to offer a more careful rendering. I hope that placing this imperfect version online may be useful to some readers, but they should note that the print version is definitive. I shall not let myself be held to the precise wording of an online version, where this differs from the print version.

Published in English Language and Linguistics 1.187–91, nominally 1997 but actually 1998.


 

Roger Mitton, English Spelling and the Computer.  Longman, 1996, x + 207 pp.

ISBN 0 582 23479 4 CSD

ISBN 0 582 23478 6 PPR

 

Reviewed by:

 

Geoffrey Sampson

School of Cognitive and Computing Sciences

University of Sussex

 

 

Most of us nowadays use word-processor systems which offer a “spelling-check” facility that attempts to compensate for spelling errors.  Anyone who has tried using a spelling checker must be aware of how limited their performance is.  To see how it worked, I once ran the spelling checker included in my WP package over a report I had written for the local newsletter, about a visit to our village by a (then) cabinet minister.  Memorably, the system identified Hurd as a word not found in its dictionary, and offered to correct it to Turd.

 

Roger Mitton describes a research project he has been working on to produce a good spelling-check system.  The work has lasted over many years; some passages read as if they were written at a period (not so long ago, after all) when word processors even of a primitive type were not yet widely used.  For Mitton it is perhaps bad luck that mechanical spelling-check technology, which may have been an original idea when he first decided to work on it, has since become a standard part of writers’ equipment.  But that does not rob the book of its value, because the workings of commercial WP systems are not normally discussed at length in the public domain.  It is unlikely that they use techniques fundamentally different from those Mitton explores, but this book is the best source I know for a clear analysis of the problems and the methods by which a machine can attempt to solve them.  Mitton surveys others’ systems, where these are documented, as well as describing his own.  Furthermore, he pays special attention to mechanical compensation for the performance of really poor spellers whose writing is on the border of comprehensibility.  This is probably not a priority for commercial spelling checkers, because very poor spellers are not given the clerical jobs that provide the main commercial justification for this technology.

 

The system I used could not respond in a sensible way to Hurd because no electronic dictionary can include comprehensive lists of proper names – there are too many.  Another large problem area is where a mis-spelled word coincides with the correct spelling of a different word.  A typist might transpose form into from or vice versa; a poor speller might write wether for whether, and wether is a good word meaning a castrated ram.  If a spelling checker considers each word separately, mistakes like these are undetectable.  In principle they might be located by reference to their grammatical and/or semantic inappropriateness in context; but this is no panacea, because correctly-spelled prose contains plenty of grammatical and semantic surprises.  All these and other issues are examined by Mitton, in an admirably down-to-earth style which never commits the academic sin of retreating into opaque jargon to get past the difficult bits.

 

Mitton’s system naturally does not include the glossy user-interface features which, at a guess, account for half or more of the coding under the bonnet of a commercial spelling checker.  It would be futile for an academic researcher to try to rival those aspects of commercial systems; Mitton has focused his effort on the central task of identifying mis-spelled words and proposing likely corrections.  His achievement is impressive.  He tested his system against 1991 versions of six commercial spelling checkers, on two kinds of imperfectly-spelled inputs:  office documents, and writing by low-ability 14-year-olds.  With both kinds of input, Mitton’s system outperformed all six commercial products.  Mitton’s system was particularly good at proposing what was actually the correct word as the first possibility in the list of alternative corrections (as he rightly says, it is doubtful whether “poor spellers find it helpful to have the required word buried somewhere in a list … they would like it at the top”).  I was rather sorry on Mitton’s behalf to see in a footnote that a 1995 test on the new version of Microsoft Word gave results better than Mitton’s system on office documents.

 

While Mitton’s system rates very well against existing commercial systems, the other question is how well any of these systems performs relative to the task confronting them.  As Mitton would agree, the answer here must be “not well”.  On the children’s writing, for instance, the proportion of errors not even detected (let alone accurately corrected) varied with different systems between 36% and 40%.  (Mitton’s system gave the best figure, the new Microsoft system was equal-worst on this particular measure.)  As one incorporates increasingly subtle considerations into a spelling-checker algorithm, the resulting performance improvements are subject to rapidly diminishing returns.  Mitton’s book concludes with an expression of faith that, ultimately, the problems can all somehow be overcome:  “The spellchecker that is as good as a good typist is not yet a reality, but there is no reason to think that it is only a dream.”  To me it seems that Mitton’s own analysis gives good reasons to think that.  But this prospect leaves me less gloomy than it would Mitton, for reasons that relate to his other topic.

 

In his early chapters, Mitton discusses the general educational and social significance of poor spelling and the complexities of the English spelling system.  Mitton surveys the history of English orthography, he examines various projects of spelling reform and spelling education, and he discusses the virtues and vices of our existing spelling.

 

Some British and American linguists in recent decades have argued that traditional English orthography is more systematic than it seems:  part, at least, of the reason why our spelling is complicated is that it reflects complexities inherent in the spoken language.  There may be a measure of truth in this, but Mitton argues that as a general picture it is misleading:  English spelling really is full of arbitrary irregularities which stem from historical accidents of many different kinds, and correspond to no features of modern spoken English.

 

I am sure Mitton is quite correct in reasserting the common-sense view of English spelling as unsystematic.  It is harder to keep company with him, though, when he assumes that this has significant deleterious consequences.  Mitton’s position here is a good deal more moderate and reasonable than, for instance, the proponents of various spelling-reform projects; but Mitton does make it clear that he sees our irrational orthography as a factor hindering the acquisition of literacy.

 

Earlier in this century, when languages other than those of Western Europe ranked as mysterious exotica, it is understandable that people saw irregular spelling as a bar to literacy acquisition.  The orthographies of all familiar languages were visibly based, closely or more distantly, on a one-sound-one-letter principle, so it was natural to imagine that regularity in this respect mattered.  The same view is not seriously tenable in today’s smaller world.  To take the most extreme case:  Japan has one of the highest literacy rates of any nation, but Japanese orthography is in large part not merely not phonemically regular but not predictable from the spoken language at all (and Japanese writing is replete with complications apart from the fact of not being phonemic).  If one’s linguistic perspective includes Japanese, to call English orthography difficult is a joke.

 

One cannot plausibly explain failure by young Britons to conquer difficulties that would be trivial for Japanese by reference to higher Japanese IQ levels, because the situation in Britain has been changing fast.  Not long ago it was normal for Britons with less schooling than the current legal minimum, and modest socio-economic aspirations, to be able to write without spelling errors, needing to consult a dictionary at most for an occasional difficult word that they rarely used.  The present situation in which many university students cannot spell even very common words has come about well within my working lifetime (i.e. since the late 1960s); it is a consequence not of the inherent difficulty of the task or of children’s innate abilities, but of changing ideology in the schoolteaching profession.  In the 1990s, as documented by Melanie Phillips (1996), primary teachers using successful and long-established teaching methods often have to do so in semi-secrecy, for fear of black marks from superiors who advocate novel teaching styles that supposedly foster children’s creativity, but do not actually lead to them learning much.

 

To achieve anything worthwhile in this world, one needs not only an element of creativity but also a willingness to get innumerable petty details right:  that seems to be a universal (if, perhaps, depressing) truth about human life.  In most ages I believe it has been taken for granted that one function of schooling is to inculcate the habit of attention to detail.  Mastery of a moderately irregular orthography looks like an ideal initial training task from that point of view:  it requires some effort and care, and leads to an outcome with relatively obvious intellectual and social payoffs.  If teachers nowadays are in effect saying to primary pupils “Here is the way your language is written, but no-one really minds whether you get it right or not”, that sounds like an unbeatable way of turning Britain into a low-achievement society.

 

Consequently it is a matter for rejoicing rather than sorrow if, as Mitton’s book appears to show, good mechanical spelling-correction is unachievable.  While that remains true, there is no hiding the success or failure of the teaching profession at one of the most basic of all educational tasks.  If computers could compensate for bad human spelling, professional spokesmen could claim that early learning was working fine when it was not; the consequences of daft educational philosophies would become visible only at later ages, when it is much harder to change intellectual habits.  Long live spelling checkers which confuse Hurd with Turd.

 

 

REFERENCE

 

Phillips, Melanie (1996).  All must have prizes.  London:  Little, Brown.