.
Grammar-checking software is more useful if its suggestions are viewed critically
THE PROBLEM OF GRAMMAR-CHECKING SOFTWARE
Unlike spellchecking software, computer programs that try to indicate a writer’s grammar errors can be unreliable, sometimes failing to find the errors and sometimes questioning correct usage. I think the writers of these programs know this because they now talk more of “proofing” than “error-correction”, and they phrase their observations more as suggestions than as categorical statements.
The weaknesses of grammar-checking software can have some fairly serious consequences for those who use it. A program’s failure to pick up an erroneous structure can reinforce the writer’s belief that it is possible or correct, while unjustified highlighting of correct structures can create uncertainty where previously there was none, thus undermining confidence.
A legitimate question in light of such problems is whether grammar-checking software will ever be perfected. I have my doubts. First of all, I wonder how aware computers are of what I have elsewhere called “invisible” grammar errors, where a writer produces grammatically possible language but it is the wrong choice to say what is intended (see 100. What is a Grammar Error?). Computers seem to base their analyses on what is objectively visible in a text – letters and words – rather than what the writer is trying to say (they cannot, after all, read minds). This practice must surely fail sometimes to highlight mismatches between what a writer intends to say and what is actually in the text.
Secondly, computers seem to have a frequent problem with finding the right pieces of a text to analyse for errors. Many searches fail, for example, to identify what that – a word with particularly flexible usage – is combining with. There is a good illustration in the following sentence from my earlier post 68. How Computers Get Grammar Wrong 1:
(a) Motivation may exist because of the possibility mentioned above that learners can enjoy reading aloud.
The computer advice here was to change that into those to match the following plural form learners – clearly nonsensical. It results from that having been incorrectly associated with the noun directly after it – often a correct association elsewhere – instead of with possibility before. The reason for the error, I suspect, is the fact that learners is positioned closer than possibility to that, the computer being programmed to look first at the closest words when seeking possible combinations. That can combine with possibility as well as learners because it can be a conjunction as well as an adjective-like “determiner” (see 153. Conjunction Uses of “that”).
Yet despite such problems and questions, current grammar-checking software is still useful, and could perhaps be made even more so if something could be done to reduce its undesirable side-effects. It is this possibility that the present post is about. I believe writers can be encouraged to look more critically at the software’s advice, and that the way to promote this is through extensive practice in analysing questioned wording and suggested alterations. As the title above indicates, this is not the first Guinlist post with this aim. For others, see 68. How Computers Get Grammar Wrong 1.
.
ANALYSIS OF SOFTWARE FAILURES
Below are some actual examples of correct English that my computer’s Microsoft Word has highlighted as possibly incorrect (their blue underlining shown in black). Readers are invited to think about why the indicated changes should be ignored.
Sentence 1
HOWEVER THE QUESTION IS POSED, THE ANSWER IS ALWAYS THE SAME.
MICROSOFT ADVICE: “After an introductory word or phrase, a comma is best”
A major problem here is the vagueness of “introductory”, since any first word or phrase in a sentence can be so described. What is really meant is initial adverb-like wording that has what I have elsewhere called a “sentence-spanning” function (see 121. Sentence-Spanning Adverbs). The comma after posed above is actually showing all of the words before it to have these features. The computer advice, though, is just referring to the underlined word however.
However above is not sentence-spanning: it is instead verb-related, linking specifically with is posed and expressing the meaning “in whatever way” (see 272. Uses of “Ever”, #6). This is a use that does not allow a following comma. The use that the software has confused it with is however meaning “nevertheless”, a sentence-spanning adverb of the “connector” variety (see 20. Problem Connectors).
Many other connector spellings are similarly usable in a non-connector way (though not all: see 121. Sentence-Spanning Adverbs, last section). Indeed, all adverb types that can precede a comma at the start a sentence tend to have this alternative use, given the right circumstances:
(b) First came the regrets.
(c) Yesterday was when it all started.
.
Sentence 2
IT REFERS TO A STRONG EMOTION OF THE SPEAKER’S.
MICROSOFT ADVICE: “Double-check whether a possessive is needed here”
Here, the “possessive” ending -’s is attached to a noun (speaker) located at the end of a sentence. This is, of course, not the most common position of nouns with this ending – they usually go before another noun (which sometimes names something possessed but sometimes does not: see 58. Optional Apostrophe Endings).
Nevertheless, the end of a sentence is a possible position of possessive nouns. They are usually separated by of from a preceding noun with “indefinite” meaning, as shown by either the article a (before countable nouns – a…emotion above) or a “zero” article (before uncountable nouns). This combination in the above sentence means “a strong emotion possessed by the speaker” (see 247. Exotic Grammar Structures 6, #4).
I can think of two possible reasons why Word has failed to recognise this structure: either it has not been programmed to do so, or (surely more likely) it has been “confused”, perhaps by the presence between a and emotion of strong. I have argued elsewhere that such “interruption” seems to be a common cause of grammar-checking software missing links between words (see 68. How Computers Get Grammar Wrong 1, #1).
.
Sentence 3
THE SECOND OF THE TWO POSITIONS IS NORMAL.
MICROSOFT ADVICE: “Double-check that you’re sticking to singular or plural”
The words “singular or plural” are the clue that are is being advised here instead of is. The very informal “sticking to” means “continuing with” (the programmers appear unaware that speakers of languages other than English – a major potential client group – are rarely familiar with such seemingly simple language). This idea of continuation is a clue that the plural are is being linked with an earlier plural idea, which must be the phrase two positions.
Thus, the software thinks the verb is does not “agree” with a preceding plural noun. Since verbs only agree with nouns that are their grammatical subject, two positions must have been taken by the computer to be one of these. However, it is not. The true subject is the second. Since this is singular, the verb form is is correct.
Key to this analysis is the preposition of. Verb subjects are rarely able to follow a preposition (see 12. Singular and Plural Verb Choices). I suspect that the software was more influenced by the fact that second is usually an adjective, so that it attributed this role to it and consequently saw positions as the only subject candidate. Second is not an adjective here but a pronoun. Adjectives cannot directly precede of + noun – they would need an intervening one(s) (see 263. Uses of “One” and “Ones”, #6).
.
Sentence 4
WHAT IS NEEDED IS MORE SOUND.
MICROSOFT ADVICE: “Double-check the way this adjective makes a comparison”
The software has taken the word sound here to be an adjective (meaning “in good order”), a use that would indeed necessitate revision of “the way this adjective makes a comparison” (i.e. its comparative form), since the correct comparative form is sounder.
However, in the absence of a context, an equally likely interpretation of sound is as the uncountable noun meaning “noise”. This would necessitate no change at all: the grammar would be fine. More would then be in a different grammatical class just as sound would: not the adverb that makes comparative adjectives and adverbs, but the adjective-like “determiner” associated with nouns (see 182. Structures with a Double Meaning 2, #3).
Spellings that are the same for different words belonging to different word classes are another very common cause of faulty computer analysis (see the end of 69. How Computers Get Grammar Wrong 2). However, I still find it slightly surprising that the software states so categorically that sound above is an adjective when its noun use is at least as common.
.
Sentence 5
ONE MEANING THAT THIS WORD CAN HAVE IS “HAPPY”.
MICROSOFT ADVICE: “Double-check the verb form after the helping verb”
I would not be surprised if this advice mystified its readers. To someone unfamiliar with technical grammatical terminology, the phrase “helping verb” must be almost as unclear as “auxiliary” (for some limited explanation of what it means, see 237. Auxiliary Verbs in Professional Communication). Moreover, even with an understanding of this concept, perplexity is likely to result from the fact that the underlined verb is follows two verbs (can and have), of which only the more distant can is a “helping verb” (“helping” have). Why, one might ask, is have not underlined instead of is?
The reason for the is underlining is that the software thinks have is a “helping” verb too. Of course, HAVE is sometimes an auxiliary (showing tenses before a “past participle”), and with that use above, a following been rather than is would be necessary. However, have here carries its non-auxiliary meaning of “possess”, which does not necessitate a following past participle. There is thus no problem with is.
.
Sentence 6
THE WATER HAVING BEEN EMPTIED, THE LAKE CONTENTS WERE REVEALED.
MICROSOFT ADVICE: “Both clauses can stand alone, so a semi-colon works best”
This says the words before the comma and those after it could each be separate sentences. In such situations, a separating comma is certainly incorrect, and a semi-colon would be acceptable (see 17. Colons versus Semi-Colons) – even though full stops separate stand-alone statements much more often.
The problem is that the above two clauses cannot both stand alone: only the second one can. The rule is that two verbs accompanied by a “joining device” must be in the same sentence, with either a comma or no separating punctuation at all (see 30. When to Write a Full Stop). The above sentence has the two verbs having been emptied and were revealed, and the joining device -ing (without BE) within having (see 52. Participles Placed Just after their Noun).
The Word software seems particularly likely to go astray when -ing is added to a passive form of a “perfect” tense (i.e. to HAVE before BEEN + -ed). For more about having participles, see 267. Participles and Gerunds with “Having”.