Attempto Controlled English Language, Tools and Applications
Exercises 2

The following assignments cover the ACE vocabulary, construction rules and interpretation rules.

Vocabulary

General

ACE vocabulary consists of function words and content words.

• Give at least two word classes that are function words and two word classes that are content words.
• Can function words overlap with content words? If yes, bring examples.
• How does ACE differentiate between prepositional complements and adjuncts?

Missing words

• Try to think of a content word which is not included in the lexicon of ACE parser. Try this for nouns, verbs, adverbs and adjectives.
• Try to think of a function word which ACE doesn't support.

Unknown word guessing

Imagine that ACE is not supported by a content word lexicon and has to guess the part-of-speech (noun, verb, adjective, adverb) of the content words on the basis of the surrounding function words in the text. E.g. a good indication for nouns is that they are preceded by a determiner ('every', 'a', 'some', ...). Come up with a few ACE sentence for which this kind of guessing doesn't work (i.e. it cannot be done with 100% certainty).

Construction Rules

Discuss

• How can simple sentences be combined into complex sentences?
• How can nouns be modified?
• How can verbs be modified?
• Discuss the differences between nouns and verbs in terms of modification, reference, argument sharing, PP-attachment.
• List the different ways a noun can be referred to anaphorically.
• What are adjuncts in the sentence "A rich man walks in the park."? What are complements?
• Delete a word from the text "A rich man runs in the park.". Did it stay syntactically correct? Why? Do this for every word in this sentence.
• How many complements must a transitive verb have?
• Can PPs be used in disjunction in ACE?
• Bring an English example that violates the "NP + V + Complements + Adjuncts" pattern that ACE enforces.
• In ACE sentences, every common noun must come after a determiner (a, the, every, no, ...). Discuss the benefits and shortcomings of this restriction.
• Discuss the difference between function words (or constructions) 'that' and 'and that'.

Constructing ACE texts

Construct...

• an ACE sentence with words: 'dog', 'cat', 'John', 'like'.
• a shortest possible sentence with the function word 'if'. (The length of the sentence is defined by the number of tokens that it contains.)
• a shortest possible sentence that contains a negation.
• an ACE text with as many types of function words as you can.
• an ACE text with one subject, one complement, one predicate and one adjunct.
• an ACE text with all types of ACE content words.
• an ACE sentence with a relativized subject.
• an ACE sentence with a relativized object.
• ACE sentences which use (1) NP negation, (2) VP negation, and (3) sentence negation. Then do the same for conjunction and disjunction instead of negation.

Constructing ACE phrases

Construct...

• a shortest possible noun phrase.
• a shortest possible verb phrase.
• a shortest possible prepositional phrase.
• a copula construction.
• a copula construction with a transitive adjective.

Which of the following texts belong to ACE?

Assume that all content words are known to ACE. In case you decide that the text is in ACE provide also the classification of content words into the classes of 'noun', 'verb', etc.

• Every man is a human.
• Tere hommikust!
• In the park a man runs.
• Buffalo buffalo buffalo buffalo.
• Men run in the park.
• Every man owns more than 2 cars.
• Every man with a degree is a scientist.
• Every man owns most cars.
• John likes someone if that person owns a car.
• If a farmer owns a donkey then he beats it.
• Colorless green ideas sleep furiously.

Translate from English to ACE

... and make the meaning of the sentences precise.

• Men are mortal.
• John will study harder if he gets a low test-result.
• The code on the card is invalid.
• John used to work for an incompetent non-profit organization.
• Every nonpacifist is a narcissist.
• John will only be admitted to the exam if he has also attended all the classes.
• A card is either valid or not acceptable, but not both.

Which of the following sentence patterns does ACE support?

• Noun Verb .
• Determiner Noun Verb .
• Determiner Verb Noun .
• Determiner Noun Verb Noun .
• Determiner Noun Verb Preposition Determiner Noun .
• Sentence and Sentence .

Interpretation Rules

General

Describe how ACE handles English ambiguity. (Hint: there are 3 basic principles.)

Discuss the following English sentences. In what sense are they ambiguous? How does ACE handle this ambiguity?

• John likes Mary and likes Ann or likes Sue.
• It is not true that a man waits and a dog barks.
• Every customer types a code.
• A customer enters a card that carries a code and falls on a floor.
• John knows a bank.
• John fills in a form.

What is the (semantic) difference between the ACE sentences?

• "Everybody sees John." and "John is seen by everybody."
• "Everybody sees a man." and "A man is seen by everybody."

Anaphoric references

• Add a shortest possible sentence to "She likes John." to make it an ACE text.
• Can a removal of a sentence turn an ACE text into a non ACE text?
• Can a removal of an if-then sentence turn an ACE text into a non ACE text?
• Can a removal of an every sentence turn an ACE text into a non ACE text?
• What is wrong with the text: "Every man likes a car. The car is fast."

Do the following statements hold?

• Removing a sentence from an ACE text can make the text syntactically incorrect. (I.e. APE would not accept it anymore.)
• Removing the last sentence from an ACE text can make the text syntactically incorrect.
• Removing all sentences that contain anaphoric references from an ACE text can make the text syntactically incorrect.

Describe how are the anaphors resolved in the following sentences.

• A cat sees a dog. It hates it.
• There is a blue ball and there is a red ball. John sees it. John sees the ball. John sees the blue ball. John sees the ball.

Come up with ACE texts which would cause the ACE parser to output the following error messages.

• Unresolved anaphor: he
• Unresolved variable: X
• Redefined variable: X
• Definite noun phrase introduces new object: man