Based on papers presented in Exeter in September 1993 and in Lancaster in August 1996. Prepared as an HTML document by John Higgins, June 1997, and updated January 2003. Click on this link to download a copy of the relevant software.

VERBALIST; ONE WAY TO EXPLORE ENGLISH GRAMMAR

John Higgins, formerly of CELT, University of Stirling

Exploatory programs

VERBALIST belongs to a family of programs that were christened 'exploratory' in Higgins and Johns 1984. The first two such programs were written by Tim Johns for the unexpanded Sinclair ZX81 and fitted into its 1K memory; they were S-ENDING which would add the 3rd-person -s to any entered verb, real or nonsense, and A/AN which would correctly select a or an to put in front of any noun phrase. The latter was written in one night in a Paris hotel bedroom, picking up a casual suggestion from the evening's conversation in time to present at the next morning's conference session. It was sensitive enough to distinguish on the basis of rules alone, not using a look-up dictionary, a uniformed man from an uninformed man, an honest opinion from a honey pot, and an 11-metre vault from a 100-yard dash.

Tim Johns and I experimented further, and several variants are discussed in our 1984 book. One thing we tried to do was to take the principle beyond morphology into functions. Tim Johns wrote LOAN, described and listed in the book, in which you enter a sum of money and a relationship, and the machine suggests a suitable way of asking to borrow that sum from that person. I wrote SORRY, generating an apology when the user selected an event (a family meal, a theatre visit, a job interview, etc) and entered the number of minutes of lateness, and THANKS offering ways of saying thank you for favours and presents when the user entered a relationship and a value. Meanwhile Martin Phillips extended the morphology set with -ING ENDING which handled the morphology of the -ing ending, ie deleting or inserting -e, changing -c to -ck, and doubling or not doubling final consonants. The last of these is by far the most difficult, and was handled largely through identifying the prefixes which affect stress in two-syllable words.

VERBALIST itself started life through an attempt to add the third inflectional morpheme needed by verbs, the past-tense marker. This, of course, needed some kind of look-up dictionary to handle irregular past forms. The stimulus to work on it came from a student who wanted to write a program which would 'teach' English verbs by asking learners to select a tense form (given the minimal context supplied by a sentence with a blank in it) and inflect the verb, and then rap them over the knuckles whenever they failed. I persuaded her that we could not computerise the selection task since we did not understand the process well enough. The time for that would be when we could write foolproof rules for encoding and decoding the meanings of tense and aspect patterns. Meanwhile, I suggested, why not write a program which could generate and correctly inflect all possible finite verb-phrases using any entered headword. Then it would be the human learner's task to explore the output and try to find plausible contexts for the computer's phrases. We went ahead, with me writing the program code and the student combing reference grammars for relevant lists.

The program would have immediate practical value as a demonstration tool for foreign learners of English with problems over spelling past tenses, provided we could equip it with a complete list of strong verbs and a comprehensive apparatus of rules for regular inflections.. We could then ask learners to browse through it, to enter all sorts of verbs, real or nonsense, in order to try out spelling rules. VERBALIST next acquired a dictionary so that the meanings of a verb could be displayed along with usage notes, such as the fact that we say try to do when try means 'attempt' but try doing when try means 'experiment'.

The first version of the program did not handle passive verbs at all. When I decided to incorporate passives, I had to create a system of code numbers to identify intransitive verbs, so that some form of warning could be displayed if the user tried to make one of these verbs passive, generating a sentence like *She was remained. The code distinguished 'always intransitive' verbs like vanish, remain, or exist, from optionally transitive verbs like walk or try. Applying the passive switch to the first type of verb triggers a red warning: "This intransitive verb is not normally used in the passive." while applying passive to the second type triggers a blue warning: "This verb is sometimes intransitive; be careful when using it in the passive." The same principle was extended to cover statives and impersonal verbs, any of which should trigger a blue warning (for unusual) or a red warning (for probably ungrammatical) when an inappropriate form was generated.

In practice there is so much likelihood of dispute that I have created and distributed a dictionary-making program so that I can carry out instant repairs to the database and so that other people can add their own words or amend my entries where they disagree with them. The dictionary now has all the irregular verbs in English (try to prove me wrong), but only about 900 other verbs, including a lot of the ones that begin with the letters A to G. In fact we will never be able to put in all the verbs of English, since people invent new ones every day.

What is VERBALIST?

The documentation describes VERBALIST as a verb-building robot. So what? you may ask; English speakers already know how to build verbs so why get a machine to do it for us? The answer is that if we succeed in teaching a machine how to do something, we come to understand better how we do it ourselves. VERBALIST is an attempt to capture in a program everything an English speaker knows about forming verbs and verb phrases. The user plays with it and tries to catch it out (as they will sooner or later). In the process they bring to the surface all the subconscious knowledge they have about how these aspects of the English language work.

VERBALIST exists in two forms, a DOS version and a Windows version. When you start the DOS version, VERBALIST displays a verb phrase on screen, eg: 

	I walk.
together with a definition. The nine function keys, f1 to f9, (radio buttons in the Windows version) allow you to change the forms. If you press f1, the present/past switch, the display turns into: 
	I walked.
Also a note appears giving the rule for making WALK past. If you now press f3, the perfect aspect switch, the screen shows the past perfect: 
	I had walked.
F2 switches in a modal verb (such as can or must) or the emphatic form do. Notice that the emphatic form is deselected as soon as you select perfect (f3) or continuous (f4), and it changes from do to get if you select passive (f7) in order to avoid the ungrammatical *We do have finished. or *She didn't be fired.

F4 is the switch for simple, continuous, and GOING TO, and f5 and f6 will switch in extra perfect and continuous forms only if GOING TO has been selected. (There is no switch equivalent to F6 in the Windows version. This is a limitation due only to physical constraints on screen size.) F7 is the switch for active/passive, f8 for affirmative/negative, and f9 for statement/question/question tag. As soon as any form is chosen which requires a 3rd person -S, an -ING form or a past tense or past participle, a relevant rule will appear explaining what the program has done to inflect the verb.

If you try every possible combination of the switches, you will have created 292 different forms of a single verb. If you combine these with all the possible pronoun and modal verb selections, you will have made 6132 different verb phrases from a single verb.

APPLICATIONS

VERBALIST has no structure. There is no beginning, middle or end; it is event-driven. (Programmers will recognise that this made it an obvious candidate for translation into Visual Basic when I came to make a Windows version.) Like a reference book it sits waiting to be consulted, and offers no prompts to carry out any action at any given time. It is not in itself 'teaching material'. To become so, it needs to be given shape usually in the form of worksheets, of which the following are examples. (The instructions apply to the DOS version of the program.)

1: Verb phrases and what they are called

Start the program by typing VERBAL and pressing the RETURN key (the large key with the broken arrow).

On the top row across the top of the screen you see f1, f2, etc to f9. These refer to the function keys at the top of your keyboard. Underneath each you see what the key will do. For instance the f1 key has the words Present and Past. The word Present is highlighted. Press function key f1 now and see what happens.

The phrase in the box changed from I walk to I walked and you saw a rule telling you how to make the past tense of a word like walk. Five words are highlighted in the top four rows. Reading from left to right, write down those five words.

	Past simple active affirmative statement
This is the grammatical label for a phrase like I walked. Now press f3 once and f4 once and write down the description and the phrase.

	Past perfect continuous active affirmative statement
	I had been walking.

Press f6. What do you see?

The f5 and f6 keys can only be used after you have selected going to with the f4 key. Try it now.

Try pressing different f keys and write down the descriptions and the phrases that you create. Altogether there are 292 combinations. Try to find ten different ones. Then compare your ten with those found by the other people in your group. Which of your phrases are normal and everyday? Which are strange? Are there any which you think are ungrammatical? Discuss those with your teacher or look up verb patterns in a grammar book.

2: Morphology; putting endings on verbs

Start VERBALIST by typing VERBAL and pressing RETURN.

The starting verb is always walk, but you can put in any verb you like. The program has a dictionary of about 1000 verbs. If you enter one of those, you will see its meaning and you may also see notes about how it is used. If you put in a verb which is not in the dictionary, the program calls it an 'unrecognised verb' but will continue to make phrases. You can, of course, enter nonsense verbs, such as verbs made up of your own name.

	I higgins ..
	I higginsed ..
	I may have been higginsing ..

Of course there is no verb to higgins in the dictionary, but you can always make up new verbs and if enough people use them they will become part of the language. When the computer makes a past tense or an -ing form from the name higgins, it also displays the rule:

	The rule for making HIGGINS past is: 
	add ED if the verb ends in two consonants 
	or two vowels plus a consonant.

To enter a new verb, look at the bottom of the screen. You will see six words: VERB, PRONOUN, MODAL, RULES, OUTPUT, and QUIT. Make sure the word VERB is highlighted (printed light-on-dark). If it is not, press the left arrow key or the space bar until it is. Then press RETURN. Type in your new verb and press RETURN again.

Enter your own name. Make a simple past tense from it (press f1 but leave all the other f keys in their starting positions. What rule is printed for your name? Write down any real English verbs which would use the same rule as your name. Try entering those verbs and see if you are right. Now do the same with the names of other people in your group.

Press the space bar in order to highlight the word PRONOUN at the bottom of the screen. Press RETURN and select SHE or HE or IT. Press RETURN again. Now press the f keys in order to make a present simple active affirmative statement. What rule do you see for adding an -S to your name? What real English verbs would use the same rule?

Now select the word RULES at the bottom of the screen and look at the list of S-ENDING rules. Try and find real English verbs for all the rules.

Conclusion

It is often said that the best way to learn a subject is to teach it, and what we are doing in VERBALIST is, in a sense, teaching the machine by making it create phrases and seeing if they are right. We have given a new answer to the question Arthur Luehrmann asked in 1972: Should the computer teach the student or vice versa?

Appendix: the rules

The following is the complete set of spelling rules which the program uses and reports. The order of the rules is a rough reflection of the program's logic; imagine them being applied in sequence until a match is made with the base form of the verb. Thus the final -e in canoe will not be deleted by rule 10, since rule 8 will already have applied, giving canoeing. Syllables are counted by an algorithm which alternately hunts for the next vowel or the next consonant. This yields the correct result in most cases, but it will be fooled by words such as reappear, which will be analysed as two syllables. The program treats QU as one consonant rather than as consonant + vowel, and distinguishes consonantal from syllabic y.

S-ending rules

  1. change HAVE to HAS (irregular)
  2. double the last letter and add ES if the verb has less than 3 syllables and ends in one vowel plus S or Z
  3. add ES if the verb ends in S, X, Z, SH, CH, or consonant + O
  4. change Y to IES if the verb ends in consonant plus Y
  5. add S to verbs which do not end in S, SH, CH, consonant + Y or consonant + O

ING-ending rules

  1. add ING if the verb is only two letters long
  2. change final IE to YING
  3. add ING to verbs which end in EE, OE, YE or INGE
  4. reverse the last two letters and add ING if the word ends in consonant + RE
  5. delete the final E and add ING
  6. add KING if the verb ends in vowel + C
  7. add ING if the verb ends in two consonants or two vowels plus a consonant
  8. add ING to the verb if it ends in a vowel other than E or one of the consonants F, H, J, Q, W, X or Y
  9. add ING to verbs of three or more syllables
  10. double the last letter and add ING if the verb has two syllables, ends consonant-vowel-consonant, and begins with a Latin prefix 16: double the last letter and add ING if the verb has one syllable and ends consonant-vowel-consonant
  11. double the L and add ING to a two-syllable verb which ends consonant- vowel-L
  12. add ING to a two-syllable verb ending in R, N, P or T with stress on the first syllable
  13. double the final consonant and add ING if the verb ends consonant-vowel- consonant and the last consonant is B, D, G, K, M, S, V or Z

ED-ending rules (Irregular verbs are first checked against the dictionary loaded from the disk file.)

  1. reverse the last two letters and add ED if the word ends in consonant + RE
  2. add D to any verb which ends in E
  3. if the verb ends consonant-Y, change Y to IED
  4. if the verb ends vowel-Y, add ED
  5. add ED to a verb which ends with any vowel except E or with the consonants F, H, J, Q, W or X
  6. add ED to verbs of three or more syllables
  7. double the last letter and add ED if the verb has two syllables, ends consonant-vowel-consonant, and begins with a Latin prefix 27: add ED to a two-syllable verb ending in R, N, P or T with stress on the first syllable
  8. double the R and add ED if the verb ends in R preceded by a vowel other than E
  9. double the final consonant and add ED to a one-syllable verb which ends consonant-vowel plus N or T
  10. add KED if the verb ends in vowel + C
  11. double the L and add ED to a two-syllable verb which ends consonant-vowel- L
  12. double the final consonant and add ED if the verb ends consonant-vowel- consonant and the last consonant is B, D, G, J, K, M, P, S, V or Z 33: add ED if the verb ends in two consonants or two vowels plus a consonant